a /3. 10 *v?y /■ o* <£? 0F c °< \ NBS SPECIALVjJ^LICAT)^/ 494 U.S. DEPARTMENT OF COMMERCE / National Bureau of Stan lEIBfeM NATIONAL BUREAU OF STANDARDS The National Bureau of Standards 1 was established by an act of Congress March 3, 1901. The Bureau's overall goal is to strengthen and advance the Nation's science and technology and facilitate their effective application for public benefit. To this end, the Bureau conducts research and provides: (1) a basis for the Nation's physical measurement system, (2) scientific and technological services for industry and government, (3) a technical basis for equity in trade, and (4) technical services to pro- mote public safety. The Bureau consists of the Institute for Basic Standards, the Institute for Materials Research, the Institute for Applied Technology, the Institute for Computer Sciences and Technology, the Office for Information Programs, and the Office of Experimental Technology Incentives Program. THE INSTITUTE FOR BASIC STANDARDS provides the central basis within the United States of a complete and consist- ent system of physical measurement; coordinates that system with measurement systems of other nations; and furnishes essen- tial services leading to accurate and uniform physical measurements throughout the Nation's scientific community, industry, and commerce. The Institute consists of the Office of Measurement Services, and the following center and divisions: Applied Mathematics — Electricity — Mechanics — Heat — Optical Physics — Center for Radiation Research — Lab- oratory Astrophysics 2 — Cryogenics 2 — Electromagnetics 2 — Time and Frequency 8 . THE INSTITUTE FOR MATERIALS RESEARCH conducts materials research leading to improved methods of measure- ment, standards, and data on the properties of well-characterized materials needed by industry, commerce, educational insti- tutions, and Government; provides advisory and research services to other Government agencies; and develops, produces, and distributes standard reference materials. The Institute consists of the Office of Standard Reference Materials, the Office of Air and Water Measurement, and the following divisions: Analytical Chemistry — Polymers — Metallurgy — Inorganic Materials — Reactor Radiation — Physical Chemistry. THE INSTITUTE FOR APPLIED TECHNOLOGY provides technical services developing and promoting the use of avail- able technology; cooperates with public and private organizations in developing technological standards, codes, and test meth- ods; and provides technical advice services, and information to Government agencies and the public. The Institute consists of the following divisions and centers: Standards Application and Analysis — Electronic Technology — Center for Consumer Product Technology: Product Systems Analysis; Product Engineering — Center for Building Technology: Structures, Materials, and Safety; Building Environment; Technical Evaluation and Application — Center for Fire Research: Fire Science; Fire Safety Engineering. THE INSTITUTE FOR COMPUTER SCIENCES AND TECHNOLOGY conducts research and provides technical services designed to aid Government agencies in improving cost effectiveness in the conduct of their programs through the selection, acquisition, and effective utilization of automatic data processing equipment; and serves as the principal focus wthin the exec- utive branch for the development of Federal standards for automatic data processing equipment, techniques, and computer languages. The Institute consist of the following divisions: Computer Services — Systems and Software — Computer Systems Engineering — Information Technology. THE OFFICE OF EXPERIMENTAL TECHNOLOGY INCENTIVES PROGRAM seeks to affect public policy and process to facilitate technological change in the private sector by examining and experimenting with Government policies and prac- tices in order to identify and remove Government-related barriers and to correct inherent market imperfections that impede the innovation process. THE OFFICE FOR INFORMATION PROGRAMS promotes optimum dissemination and accessibility of scientific informa- tion generated within NBS; promotes the development of the National Standard Reference Data System and a system of in- formation analysis centers dealing with the broader aspects of the National Measurement System; provides appropriate services to ensure that the NBS staff has optimum accessibility to the scientific information of the world. The Office consists of the following organizational units: Office of Standard Reference Data — Office of Information Activities — Office of Technical Publications — Library — Office of International Standards — Office of International Relations. i Headquarters and Laboratories at Gaithersburg, Maryland, unless otherwise noted; mailing address Washington, D.C. 20234. 2 Located at Boulder, Colorado 80302. a MFPG Detection, Diagnosis and Prognosis Proceedings of the 26th Meeting of the Mechanical Failures Prevention Group, held at the IIT Research Institute, Chicago, Illinois, May 17-19, 1977 Edited by T. Robert Shives and William A. Willard Metallurgy Division Institute for Materials Research National Bureau of Standards Washington, DC. 20234 The 26th meeting of the MFPG and these proceedings were sponsored by the Institute for Materials Research of the National Bureau of Standards, Washington, DC 20234; the Office of Naval Research, Department of the Navy, Arlington, VA 22217; the Naval Air Development Center, Department of the Navy, Warminster, PA 18974; the Frankford Arsenal, Philadelphia, PA 19137; the Federal Aviation Ad- ministration, Department of Transportation, Washington, DC 20591; the National Aeronautics and Space Administration, Goddard Space Flight Center, Greenbelt, MD 20771; and the Energy Research and Development Administration, Washington, DC 20545. ^fAU Of U.S. DEPARTMENT OF COMMERCE, Juanita M. Kreps, Secretary ( £ Dr. Sidney Harman, Under Secretary a§ Jordan J. Baruch, Assistant Secretary for Science and Technology NATIONAL BUREAU OF STANDARDS, Ernest Ambler, Acting Director Issued September 1977 Library of Congress Cataloging in Publication Data Mechanical Failures Prevention Group. MFPG— detection, diagnosis, and prognosis. (NBS special publication ; 494) "Sponsored by the Institute for Materials Research of the National Bureau of Standards. . . [et al.]" Supt. of Docs no. C13. 10:494 1. Fracture mechanics-Congresses. I. Shives, T. R. II. Willard, William A. III. United States. National Bureau of Standards. Institute for Materials Research. IV. Title. V. Series: United States. National Bureau of Standards. Special publication ; 494. QC100.U57 no. 494 [TA409] 602Ms [620.F126] 77-14098 National Bureau of Standards Special Publication 494 Nat. Bur. Stand. (U.S.), Spec. Publ. 494, 296 pages (Sept. 1977) CODEN: XNBSAV For sale by the Superintendent of Documents, U.S. Government Printing Office Washington, D.C. 20402 Stock No. 003-003-01844-9 ■fr U.S. GOVERNMENT PRINTING OFFICE : 1977 O-240-848 FOREWORD The 26th meeting of the Mechanical Failures Prevention Group was held May 17-19, 1977, at the IIT Research Institute in Chicago, Illinois. The program was organized by the MFPG committee on Detection, Diagnosis, and Prognosis under the chairmanship of Henry R. Hegner of GARD, Inc. The committee, the session chairmen, and especially the speakers are to be commended for the excellent program. Most of the papers in these Proceedings are presented as submitted by the authors on camera ready copy with some minor editorial changes. Special appreciation is extended to, Mr. H. G. Tobin Assistant Director, Electronics Division, IIT Research Institute for inviting the MFPG to use the very fine facilities of the Research Institute for this con- ference, Mr. Edward P. Fahy, Materials Manager at IITRI and members of his staff for the meeting arrangements, and to Robert M. Whittier of Endevco for meeting publicity. Appreciation is also extended to the following members of the NBS Metallurgy Division: T. Robert Shives and William A. Willard for their editing, organization, and preparation of the Proceedings, Paul M. Fleming for handling financial matters, Todd Eudy for photographic work, Larry W. Ketron for drafting work, and to Marian L. Slusser for typing. HARRY C. BURNETT Executive Secretary, MFPG Metallurgy Division National Bureau of Standards III TABLE OF CONTENTS Page FOREWORD HI SESSION I: OIL ANALYSIS REVISITED 1. Statistical Analysis of Wear Metal Concentration 3 Measurements in Oil; Calculation of Significant Wear Metal Production Rates. Karl Scheller* and Kent J. Eisentraut 2. Effective Fluid Analysis of Oil-Wetted Systems 24 Through Proper Planning and Interpretation. R. K. Tes smarm* and G. E. Maroney 3. Oil Analysis/Wear Particle Analysis. Peter B. Senholzi 27 4. Application of Ferrographic Lube Oil Analysis to 33 U.S.N. Ship Systems. Gerald F. Rester 5. Effectiveness of the Real Time Ferrograph and Other 49 Oil Monitors as Related to Oil Filtration. Raymond Valori 6. Ferrographic Separation of Organic Compounds. 73 E. Roderic Bowen* and Vernon C. Westcott SESSION II: SIGNATURE ANALYSIS TECHNIQUES 1. Mechanical Signature Analysis as a First Step in 77 Quantifying the Characteristics of Operating Machinery. John S. Mitchell 2. Spectrum Analysis and Machinery Monitoring. 79 George F. Lang 3. Comparison of Vibration Signature Analysis Techniques. 82 John L. Frarey V 4. The Role of Signal Processing in Machinery Vibration 93 Analysis. James H. Hamilton 5. Diagnostic Techniques for Steam Turbines. R. L. 97 Bannister,* R. L. Osborne and S. J. Jennings 6. Experimental Determination of Radial Magnetic Forces 117 as a Function of Rotor Offset in a Large Induction Motor. Robert L. Leon SESSION III: NEW DETECTION, DIAGNOSIS AND PROGNOSIS TECHNIQUES AND EQUIPMENT 1. A New Chip Detector -- Reliable, Versatile, and 123 Inexpensive. Thomas Tauber 2. Another Look at Time Waveform Analysis. J. B. Catlin 133 3. The Advent of Sophisticated Fluid Power Systems and 138 Its Impact on Preventative Maintenance in the Military. Michael W. Wigton 4. Tire Degradation Monitoring. Wieslaw Lichodziejeswki 146 5. Use of Microprocessors in Analysis of Acoustic 152 Emission Weld Monitoring Data. R. N. Clark* and T. A. Mathieson 6. Pictorial History of the Development of Proximity 167 Probes for Use in High Temperature Liquid Metals Environment. Leo Hoogenboom SESSION IV: RAILROAD SYSTEM DIAGNOSTICS 1. Department of Transportation System for Train 191 Accident Reduction (DOT- STAR) . James K. O'Steen 2. Comparison of Vibration Analysis Techniques for 205 Railroad Roller Bearing Diagnostics. Warren D. Waldron VI SESSION V: LAND VEHICLE DIAGNOSTICS 1. Maintenance Management Through Diagnosis. 225 Richard G. Salter 2. Vehicle Monitoring System. S. C. Hadden, R. E. Hanson 238 and Mi W. Stewich* 3. Systemized Diesel Engine Diagnostics. Henry J. 249 Mercik, Jr. LIST OF REGISTRANTS FOR THE 26th MFPG MEETING 259 MFPG PUBLICATIONS 268 APPENDIX Remote Diagnostic Techniques Used in Viking Lander 273 Operations. Parker S. Stafford (Banquet Speaker) The following registrants gave presentations at the symposium, but did not submit a manuscript for publication: John Dunmore, "Updating Signature Analysis Techniques," Session II. Jerry Allen, "Bearing Vibration Crest Detection," Session III. Raymond Ehrenbeck, "Overview of Department of Transportation Sponsored Diagnostic System Development," Session IV. David R. Sutliff* and Nicholas J. Darien, "Development of Data Collec- tion System for Railroad Testing," Session IV. J. E. Johnson* and Lloyd H. Emery, "Nondestructive Testing of On-Vehicle Tires Using Random Mechanical Excitation," Session V * Indicates speaker when a paper had more than one author. VII ABSTRACT These proceedings consist of a group of twenty four submitted papers from the 26th meeting of the Mechanical Failures Prevention Group which was held at the I IT Research Institute in Chicago, Illinois on May 17-19, 1977. The central theme of the proceedings is detection, diagnosis and prognosis as related to mechanical failure. Papers are presented that discuss oil analysis, signature analysis techniques, new detection, diagnosis and prognosis techniques and equipment, railroad system diagnostics, and land vehicle diagnostics. Key words : Failure detection, failure diagnosis; failure prevention; f errography ; land vehicle diagnostics; oil analysis; railroad system diagnostics; signature analysis; wear UNITS AND SYMBOLS Customary United States units and symbols appear in many of the papers in these proceedings. The participants in the 26th meeting of the Mechanical Failures Prevention Group have used the established units and symbols commonly employed in their professional fields. However, as an aid to the reader in increasing familiarity with and usage of the metric system of units (SI) , the following references are given: NBS Special Publication, SP330, 1974 Edition, "The International System of Units." ISO International Standard 1000 (1973 Edition) , "SI Units and Recommen- dations for Use of Their Multiples." E380-76 ASTM/IEEE Standard Metric Practice (Institute of Electrical and Electronics Engineers, Inc. Standard 268-1976). Disclaimer: Certain trade names and company products are identified in order to adequately specify the experimental procedure. In no case does such identification imply recommendation or endorsement by the National Bureau of Standards, nor does it imply that the products are necessar- ily the best available for the purpose. VIII SESSION I OIL ANALYSIS REVISITED CHAIRMAN: J . M . PEREZ CATERPILLAR TRACTOR COMPANY STATISTICAL ANALYSIS OF WEAR METAL CONCENTRATION MEASUREMENTS IN OIL; CALCULATION OF SIGNIFICANT WEAR METAL PRODUCTION RATES Karl Scheller Southeastern Center For Electrical Engineering Education, Inc. 202-D Samford Hall Auburn University Auburn, Alabama 36830 and Kent J. Eisentraut Air Force Materials Laboratory Air Force Systems Command Wright-Patterson AFB , Ohio W33 The service life of aircraft engines and various other types of lubricated equipment in the Air Force is currently being monitored in the USAF Spectrometric Oil Analysis Program (SOAP) in -which concentrations of wear metal in lubricating oil in the parts per million range are determined spectrometrically . These measurements generally ignore the effects of oil consumption and subsequent make-up and pay only qualitative heed to the rate of increase of wear metal concentrations. SOAP laboratories are provided with Wearmetal Tables to aid them in evaluating the significance of the wear metal concentrations The Tables, particularized for each item of monitored equipment, recommend maintenance actions depending upon the wear metal concentrations of significant elements. It is not intended that the laboratory analyst use the Tables on an absolute go-no go basis. Recommendations are to be tempered in the light of the past SOAP and maintenance history of the equipment and its operating conditions, augmented by a healthy measure of the analysts SOAP experience and judgment in ascertaining the trend of the data. When this becomes difficult, it is suggested that a simple plot of wear metal concentration values against equipment operating time be prepared as an aid. For equipment such as the TF^l-1 engine, which powers the A-7D aircraft, the lubricant capacity is small (ca. 3 gal) and the oil consumption rates are not infrequently as large as one quart per hour, with post-flight oil additions ranging up to 25% of the total oil volume. Under these circumstances the effect of oil consumption on measured wear metal concentrations is appreciable and must be considered in a proper evaluation of the data. This report discusses methods that have been proposed to simultaneously account for the effect of oil consumption and engine operating time in a quantitative manner. Attention has been focused on the TF^l-1 engine because of its particular sensitivity to oil consumption effects. The techniques described and developed here are, of course, generally applicable to other engines and lubricated equipment. The SOAP Wearmetal Table for the TFUl-1 engine, shown in Figure 1, is typical of those for engines. Three different sets of guideline concentration values are given for the three types of instruments available at SOAP laboratories, Baird-Atomic A/E35U-1 and A/E35U-3 emission spectrometers and the Perkin-Elmer 303/305 atomic absorption spectrometer. Recommended action codes are listed to the left of the Tables and explained below them. For this engine, especial caution is urged in using the Table alone as a basis for evaluating the significance of wear metal data. Trend analysis in the form of a plot of analytical results against operating time is strongly recommended. Accordingly, typical SOAP records for iron in the oil from three TFUl engines are graphed in Figure 2. All are highly erratic in character and any trends deduced from them are subject to doubt. In a broad qualitative sense, one might assert that iron is increasing rapidly for Engine A, less rapidly for Engine B, and is substantially constant for Engine C. However, one must have misgivings about the repeated increases and decreases in metal content , particularly marked in Engine A. The immediate explanation that suggests itself for this behavior is the low precision of the measurements. Upon examination of this possibility, however, one learns that the instruments are required to have a repeatability (standard deviation) of lppm in this range of iron concentration and that they are checked frequently for conformance to this specification. Proceeding upon the premise that the data are within the required precision limits, the effect of oil consumption becomes the next likely candidate as the primary cause of the anomalous concentration variation. Figure 3 illustrates the effect of oil consumption on the measured wear metal concentrations of an engine similar in characteristics to the TFUl-1. Assuming the oil is absolutely clean initially and is sampled (and topped) at equal time intervals, the concentration will build up smoothly to an asymptotic limit of Tppm for the given conditions. In contrast, if the engine were not consuming oil, the wear metal concentration would increase in linear fashion until the engine failed or the oil was changed. Thus after 100 hours of operating time, for example, the measured concentration would be Tppm while the concentration indicative of the actual engine wear would be 35ppm, well above the guideline for removal. Of course, in a real case the sampling intervals are not equal and the measurements are only reproducible to ±lppm; thus the concentration values would tend to fluctuate about their asymptotic limit much in the manner of that shown for Engine C in Figure 2, in which a limiting concentration appears to have been attained. Since measured wear metal concentrations and their simple variance with time have serious deficiencies for use as engine serviceability- criteria, particularly in the case of oil-consuming engines, the use of ■wear metal production rates (WMPR) has "been proposed for this purpose. These may be calculated "by the various approximate and one differential (and presumably more exact) relationship given in Figure k, the latter described by Lotan (Ref. l) . All are derived from a simple material balance. The three approximate equations differ in their assumptions regarding the wear metal concentration in lost oil while Lotan' s relationship assumes a constant rate of oil loss at a variable concentration and a constant wear metal production rate. The various formulas are applied to SOAP data for a TFla-1 engine in Figure 5- Results are calculated following the customary noted practice. As might have been expected, the approximation using the average value of the concentration yields the same result as the differential relationship while the assumption that the oil loss occurs at the final or the initial concentration gives a WMPR that is high or low respectively. In the interests of the simplest arithmetic, Approximation 2 was used for all further wear metal production rate calculations. In certain situations, it may be of interest to calculate the wear metal production rate where the oil system is topped off several times between analyses. A method for accomplishing this is outlined in Figure 6. The relationship shown here is similar to the one found in Reference 1, though its derivation is different and much less elaborate than the latter. Though Air Force practice, sampling after each flight, virtually eliminates the use of this equation for its intended purpose, it has some utility in calculating average rates from a total elapsed time, a total oil volume added, a given number of fillings, and the initial and final analyses. Concentration-time data and derived wear metal production rates are depicted in Figure 7 for a fairly typical SOAP iron record on a TFUl engine. The average oil consumption over the time period covered in the plot was approximately 0.25 quarts/hr (moderate and representative of the engines for which oil consumption data was available). The wear metal production rate graph proves to be both disappointing and disturbing. At best it mirrors the trend of the concentration curve in greatly magnified fashion, with some minor deviations due to the effect of oil addition (notably in the range from 23-29 hours where the concentration is constant). Its worst aspects are the numerous negative values of the WMPR and its highly erratic nature, varying widely from point-to-point. The physical interpretation of the negative WMPR values is that the engine is cleaning itself, an obvious impossibility. Hence they must be ascribed to the slight (and expected) variability in the concentration measurements. Since the concentration appears to be sensibly constant , one is led to determine the average concentration and the standard deviation of the measurements. The results indicate that the concentration is probably constant over the entire operating time and that the precision of the measurements is comfortably within the specified limit of lppm. Reasoning that since the concentration is constant, the WMPR might well be also, the same procedure is applied to these values with dismaying results. The precision of the calculated WMPR's is twice the mean value. The relative difference in the precision of the concentration measurements as compared with the wear metal production rates calculated from them is strikingly demonstrated in Figure 8, where the 95$ confidence limits on "both values are plotted to the same scale. These results are not unique to this engine but representative of others analyzed. The variability of the WMPR's is approximately an order of magnitude greater than the variability of the concentration measurements, if one assumes that we are obtaining a very imprecise estimate of a constant value . As already indicated, the expression developed for calculation of the wear metal production rates for the case of n fillings between two analyses (Figure 6) may be utilized to derive an average WMPR. For this engine (SN 1670) , it is found to be 2.1+5 for the recorded operating time and conditions. An estimate of this sort is somewhat reassuring in that it masks negative values and point-to-point variation in the averaging process. However, its precision (theoretically of the same order as the concentration measurements) cannot be deduced from a single set of data and its accuracy is indeterminate, since the calculated WMPR (r) is so highly dependent upon the final analysis (Cg ). For a sufficiently large set of data points, the uncertainty of the estimate can be reduced by forming subsets of adequate size and number and obtaining the distribution of the rates calculated for each subset. In general, such a procedure is not very feasible, since an engine may well reach a critical condition before enough measurements are accumulated for a reliable analysis of the data. An alternative procedure for extracting wear metal production rates from concentration measurements as a function of operating time is to perform a linear regression analysis upon the data on the assumption that the WMPR is reasonably constant. This method will yield a trivial and incorrect result (r=0) when the concentration is ostensibly constant (as in Figure 7). In most instances, however, the concentration measurements exhibit an increasing trend with time. A representative example of such behavior is shown in Figure 9« The rather wide fluctuations in concentration displayed by these data are undoubtedly due in part to the effects of oil consumption, which were not reported for this engine. Linear regression analysis produces a rather satisfactory correlation of the data as measured by the correlation coefficient of 0-9^+1 as compared with unity for an exact fit. The wear metal production rate is deduced from the slope of the straight correlation line with the indicated standard deviation. By contrast, the graph of the WMPR follows its usual erratic course, magnifying the fluctuations in the concentration. It exhibits no discernible trend, is nowhere alarmingly large and even tends to decrease as the concentration climbs to a value high enough to signal maintenance action. Furthermore the average of the graphed WMPR values (2.1+5) is only 70% of that found from the linear correlation and its standard deviation is thirty times as large. It is interesting to compare the relative trends in concentration and WMPR for an engine removed for teardown upon the "basis of a SOAP laboratory recommendation. Such data are shown in Figure 10 for an engine used in Reference 2 to illustrate the utility of the WMPR concept. The concentration trends upward (with considerable fluctuation) rising from 6 to 12ppm in the interval between 5 and 38 hours of operation. In the next hour of operation it rises abruptly to l6ppm. The WMPR graph exhibits a magnified fluctuation but no discernible trend over the same time interval. In the last hour of operation, the WMPR jumps from 6 to 50 mg/hr. It is contended that this is a much clearer indication of engine deterioration than the jump in concentration, even though the prior WMPR history gave no indication whatsoever of abnormal wear. Before discussing this further, it is well to note that the entire range of WMPR values prior to the last hour of operation falls well within the 95% confidence belt (ca. -10 to 15 mg/hr) , a fact that raises considerable doubt regarding their interpretation. In order to assess the significance of a large increase in WMPR, it is necessary to obtain an estimate of the possible errors in the calculation. This problem is addressed in Figure 11, which outlines the procedure to be followed in estimating errors and tabulates their maximum possible values as a function of concentration. For the engine under consideration (SN 1175, Figure 10, operating interval of one hour, concentration of 15ppm) , the maximum error at the 95% confidence limit can be as large as 39 mg/hr. Thus there is at least a 5% probability of calculating a WMPR of approximately kk mg/hr on a random error basis and the significance of the value of 50 mg/hr actually obtained is open to question. Consideration of the large inherent error possible in calculated WMPR's and their inevitable erratic nature leads to the conclusion that they are not suitable indexes of engine wear and should not be used as a criteria for maintenance action. One must next confront the problem of evaluating the significance of a large increase in concentration. A method for accomplishing this purpose is illustrated in Figure 12 for the data on Engine SN 1175 used in Figure 10. Assuming a constant WMPR, linear regression analysis yields the best least squares fit line to the concentration measurements. The correlation coefficient, Rr0.92, indicates that the assumption is reasonable. Superposition of the 95% confidence limit belt on the linear regression line demonstrates that the final concentration measurement of l6ppm lies well outside these limits. Calculation of the probable deviation of a measurement from the least squares fit line indicates that there is less than a 1% chance of obtaining such a value as a result of random error. It is now justifiable to conclude that the engine has indeed begun to wear abnormally and to call for appropriate maintenance action. Though the technique just described provides a satisfactory assessment of the concentration measurements and statistically valid wear metal production rates , it does not account for the effect of oil consumption. Certainly the wear metal production rate of 1.89 found for Engine SN 1175 (Figure 12) must be too low. To compensate for the oil consumption and dilution effects a corrected concentration may "be calculated from the expression given in Figure 13. The corrected concentration is, of course, that fictitious value that would he measured if all the wear metal stayed in the system and only the oil were consumed. The concentration measurements for Engine SN 1175 9 compensated for oil consumption in accordance with this procedure are depicted in Figure ik together with their fitted linear regression line and superimposed 95% confidence belt. Comparison with Figure 12 indicates that the fit of the data has been noticeably improved (correlation coefficient increases from 0.92 to O.96) and the wear metal production rate is approximately doubled and of higher relative precision. The most important conclusion to be drawn from this exercise is that oil addition data is necessary for a valid and meaningful assessment of engine wear from concentration measurements on oil-consuming engines. To drive home further the importance of oil addition data, corrected and measured concentrations are presented in Figure 15 for Engine SN 1670 (previously discussed in Figures 7 and 8). The oil consumption effect is magnified in this instance by the condition that the contaminant level appears to have reached its limiting concentration. Correlation coefficients and derived WMPR's are given for the two sets of data, though the linear regression lines themselves are omitted for the sake of reduced clutter. As has been already stated, the virtually constant concentration results in a trivial and incorrect WMPR of approximately zero and a corresponding low correlation coefficient. Compensation for oil consumption dramatically increases the correlation coefficient and produces a plausible wear metal production rate. In summary it may be stated that : 1. Linear regression analysis of wear metal concentration measurements offer a great improvement over current qualitative trending methods in evaluating engine service life. 2. Oil addition data is indispensable for the interpretation of concentration measurements from oil-consuming engines. 3. Proposed point-to-point calculation of wear metal production rates will not serve as a suitable index of engine wear due to their inherent low precision. k. The techniques proposed here seem to account satisfactorily for the effect of oil consumption and engine operating time on wear metal concentration measurements. References : (l). Lotan, D. , SNECMA Document MFTM No. IO56U , "An Investigation into the Rate of Contamination of the Oil Circuit of a Turbojet Engine" (2). Cuellar, J.P. , Jr. and Staph, H.E., Southwest Research Institute Report No. RS-630, July 1975- "Task Report, Task 3 - Wear Metal Production Rate Compensation". 8 A/E35U-1 Fe Ag Al Cr Cu Mg A 0-13 0-1 0-2 0-1 0-7 0-2 K 14-19 2-3 3-4 2-3 7-9 3-4 T 2 0+ 4+ 5+ 4+ 10 + 5+ Fe Ag A/E35U 3 Al Cr Cu Mg 0-21 0-3 0-3 0-4 ' — 0-10 0-4 22-29 4 4 N/A 11-14 5-6 30+ 5 + 5f 5 + 15+ 7+ Atomic Absorption Fe Ag Al Cr Cu Mg A 0-10 0-2 0-2 0-2 0-5 0-2 K 11-14 N/A N/A N/A 6-7 3 T 15 + 3+ 3+ 3+ 8 + 4 + Codes A - Continue routine sampling. K - Submit redtagged sample as soon as possible. Suspect possible discrepancy due to increasing T - wearmetal trends, recommend maintenance inspection. If aircraft has flown since last sample, ground unit until results of this sam- ple are known. Ground unit, examine for suspected discrepancy. Advise laboratory of findings. Figure 1 9 GO LU 2 CD 2^ uu T — ^ ^— ^ i >- UL, 1— \— O < DC Q_ CD < UL. CJ _j ^ C^=D ^Z. CD < 1— s? _j MBBB i < 2! ^ CD CO OC k ^— 2 QC O GO CO CD CD o CD _, Q_ CD < E CD — GO CD CD CO ° rr CO "- I LU CD h LT3 o z h CD «3- < DC LU a. CVJ 5 \ •H CD CO LU z J* o z LU CD CM G LU CO GL < _ CD lAIdd NOI±VdXN30NOO ®d 10 LU EFFECT OF OIL CONSUMPTION (AND ADDITION) ON MEASURED WEAR METAL CONCENTRATIONS WITHOUT OIL ADDITION LOWEST GUIDELINE FOR REMOVAL (Fe) LIMITING CONCENTRATION WITH OIL ADDITION 25 50 75 100 125 ENGINE OPERATING TIME, HRS. Figure 3 11 o CO 5ti ■H CO 12 UJ I D. < go £ o 5 Z UJ DC o i *>' < CO CO CJ ^ < 42 r o- er oo m O csi H II £3 ro jo < CD o o •I N u o < ro m O fO lO 2 H II ui X ui > O CO < CL < o z < oc < o ice So is z< uj _ . £»£ ?| Ui w ooz uj O Oj Z_i ui - ccS UJ £C u. £ = ft- Q CO -2 si o * uj < ZH O N I I IO CM IO M (0 N a CVJ = M CM _l« 31" 4f 0) toC •H CO CSJ o CO 13 DC UJ Z 5 O z 5 o -J -I o u. ui O z < Ui UJ Ui > < z UJ UJ I- UJ CO Ui o Q O < Ui UJ UJ > -J < z UJ UJ UJ 00 o g u. = O H uj O > UJ P O ^ > <_ - > U. Ui II S8 uj o CC Ui O uj U. CC £\= 3 H« if ;? jT z o < s o 0C a. a. < oc o u. z UJ X ffi o § + c ffi o t I > 8 z UJ X > 3 J 1 CNJ J 1 vO 2 h < QJ U SQ •H 18 < LU Q GO |jj LU < < §o < P S a UJ oc < (A UJ 5 X (9 s 3§ ^^52? CO UJ Z UJ UJ 5 t- o< z z> ;• oc i w % oc UJ UJ O (9 s UJ u z o a oc o oc oc UJ M ■ 00 »(E UJ 00 e UJ N X 5 o z UJ X a? <■§ zx UJ r- M |jE O uj O-i cci- o< °=Z oc uj ui ^ I* SCC — ui < UJ 8 oc ° cc S £ UJ 3 < -I < > 2 < a > 3 < 2 oc X < o o £ < ^^ Q ^ UJ 8 iX > o UJ UJ o a. z (/> W a d •H Pn 5 X UJ Q. UJ oc > o < 2g l-O zee Uj — o z o u a. a. NifiN S2K s t 8S88 olcJoievi 19 o < CO SP LU ccuJ Zj « LU *■ o LU LU LO CO o > LT H- LU \ * >! I lAOf o> O — 1 —J lL o o LO en o CO v\ CO o LO CN CO s 1 \ % \ +1 oc CM — 1 LU o v5 op II o CM z LU U 00 LU Q oc a. 2 •H == o \ \ lAIdd 'NOIXVHXN3DNOO 20 CO CD CD CJ < H LU 2 _l UJ 5 X UJ H- I 1- 1- CO < z UJ -I a. X 1- a. a. O 1- < CO co CO cc UJ z is UJ —I a. < CO O u_ < _l O z UJ _i Q_ UJ _l Q Z < z ION WHICH. RICATINGS 1- z UJ _l < 1- < CO 1- Q. < z o z UJ UJ < 5 cc -J UJ UJ co CO 1- UJ 03 -J < > 1- UJ § ^ cc a < Z X UJ CO -J < > cc UJ h- z z cc < UJ LL u. UJ I 1- (J z o z o o UJ CO cc i- z UJ O z _ T "6 + c 3 CONCE CKTOT O z o z z o o o Q UJ 1 > CC UJ 1- z TIOUS EDBA < _l _l O LL 1- < cc D cH- z 1- CC cc 1- CO < + 1- CO co z J> z UJ en o < - o _l UJ o z UJ 1- co UJ o z 2 JZ <-> D 0) c o _i _■ ~ O o >- o z o £ "J O a CO o UJ ii II CO — rr J- n 1" ii "6" 11 6" • I 1- cc o u_ o c O c > < UJ 0FCOUI WITHT 1- z UJ CC UJ »8 UJ X X X Zj _l V- § h- H w •H 21 ENGINE SN 1175 a. a. 21 19 17 15 CORRELATION OF WEAR METAL CONCENTRATION WITH TIME INCLUDING EFFECT OF OIL ADDITION 13 Z g < 1 11 z o O 9 95% CONFIDENCE LIMIT 95% CONFIDENCE LIMIT LINEAR REGRESSION LINE R = 0.9613 CORRELATION COEFICIENT 3.61+0.59 (95% CONFIDENCE LEVEL) i I | L_ 15 20 25 TIME, HRS. Figure 1^ 40 22 o o O ~ o CO CO " LU O < fE — CD CD o £ CD CO ^ < CO to m CO CO oc UJ 2 o h CO m •H in eg CM lAIdd N0IXV«XN30N00®d 23 EFFECTIVE FLUID ANALYSIS OF OIL-WETTED SYSTEMS THROUGH PROPER PLANNING AND INTERPRETATION R. K. Tessmann and G. E. Maroney Program Manager Program Manager Fluid Power Research Center Oklahoma State University Stillwater, Oklahoma A major dilemma which has long faced engineers concerned with the prevention of mechanical failures is the early detection of such events. In the case of systems com- posed of oil-wetted components, the assessment of wearing processes is hampered by the necessity to contain the fluid. That is, the critical mating surfaces cannot be viewed directly to appraise deterioration. In general, there are three methods used for deducing the internal state of such components. The first approach and probably the most wide- ly used is associated with external performance degradation (changes in leakage, heat level, vibration, noise levels, etc.). The second procedure involves the dismantling of the system to "pull an inspection." The third technique and the subject of discussion here is fluid analysis. The use of fluid analysis for the detection and prevention of failures in oil-wetted sy- stems has undoubtedly been spurred by the cost factor associated with maintenance and down time. As oil-wetted systems have grown in complexity, the cost of their mainten- ance and down time as well as their initial cost has increased. However, to be effective, any diagnostic program which relies upon fluid analysis must be properly planned and interpreted. The planning phase involves such aspects as sampling location, sample extraction, and sample container cleanliness. These aspects must be carefully considered to be certain that a representative sample is acquired. Once the analysis has been accom- plished, it is necessary to interpret the results in terms of wearing processes occurring in the system components. Basically, most fluid analysis techniques used today actually evaluate some portion of the contaminants entrained in the system fluid. The contamination level in an oil-wetted system is a dynamic parameter. For example, it can increase or decrease, depending upon the amount of ingression from the external environment and the rate of genera- tion from internal surfaces. Also, the contamination level, in general, will be lower downstream of the system filter than it is upstream. Thus, it is important to consider these factors when obtaining a fluid sample. The sampling procedure which is followed is another critical factor in the sample acqui- sition process. Since most contaminants of interest in an oil-wetted system will settle when the fluid is permitted to remain stationary, the sampling procedure most often 2k recommended is called dynamic line sampling. This technique requires that the fluid be extracted from a turbulent region of the connecting lines while the system fluid is circulating. Reservoir sampling is also used effectively when care is taken to see that adequate mixing occurs within the reservoir. In addition, the location within the reser- voir is important with such a technique. Sample container cleanliness is very critical. All of the contamination in the container prior to placing the fluid in it will be added to that already entrained. Standard pro- cedures are available for fluid sampling and evaluating container cleanliness. If proper consideration is given to sampling location, sampling technique, and container cleanliness, a sample of fluid will be acquired which is representative of the system fluid. If oil-wetted systems are defined to include both hydraulic and lubrication systems, there are five fluid analysis techniques in widespread use: (1) particle counting, (2) gravimetric analysis, (3) pad comparison, (4) Ferrography, and (5) spectrograph^ analysis. These five methods can be divided into two categories. The first three tech- niques are concerned with the total particulate contamination present in the sample, while the latter two are normally used for wear debris evaluation. Thus, the interpreta- tion which can be made from the data obtained from each of these methods is depend- ent upon the category into which it falls. That is, it is very difficult to assess the concentration and particle size distribution of the wear debris present in the sample from any of the first three methods listed. Particle counting provides information relative both to particle size distribution and contaminant concentration. Obviously, a high concentration of very large particles will result in early failure of any oil-wetted component. Gravimetric analysis can only produce data relative to the total concentration of contaminant, while pad comparison relies upon the similarity between a membrane produced from the fluid sample and a standard membrane. Each of these three analysis techniques provides data upon which a diagnostic appraisal can be made; however, particle counting data are much more complete. The two analysis methods which are primarily concerned with the type and concentra- tion of wear debris provide acutely different information. Spectrographic analysis pro- duces data relative to the concentrations of various elements, while Ferrography provides morphological results in regard to the wear debris. Both of these techniques have been and are being used effectively. The point to be made here is that the two methods offer different types of information, and the interpretation must take this into account. The main objective of any diagnostic effort is to determine what, if any, corrective action should be taken. Replacing the system filter with one which is more efficient or reducing the rate of ingestion of external contaminants will reduce the contamination level of the system and, in general, extend the service life of the components. In addition, the severity of the operating conditions will directly influence the wear rate of oil-wetted components. Thus, if it is found through the process of fluid analysis that 25 the system is being subjected to conditions which result in excessive wear, there are two obvious courses of action: (1) modify the duty cycle or (2) change to better components (filters, seals, pumps, etc.). Modification of the operational duty cycle is usually not a satisfactory corrective action. Such a change means that the machine will not perform effectively at the intended pow- er level. For example, in order to reduce the operating severity on a bearing, the speed or load or both must be reduced. In the case of a hydraulic system, the pressure can be reduced, which means that the system cannot transmit as much horsepower. Thus, corrective action normally takes the form of component replacement. However, merely changing a worn component with a new one will not remedy the cause of the potential failure. If the component life was not adequate, then something must be done about the conditions which caused the premature failure. In many cases, providing better contaminant protection is the answer. This can be accomplished by more effective filtration or by improving the seals and breathers to exclude contamination from the external environment. However, no matter what corrective action is deemed necessary, some consideration must be given to the cause of the failure if longer component life is desired. 26 OIL ANALYSIS/WEAR PARTICLE ANALYSIS Peter B. Senholzi Naval Air Engineering Center Lakehurst, New Jersey 08733 INTRODUCTION Oil analysis has been applied for many decades as a preventive maintenance concept utilized to predict impending mechanical failure through examination of used oil samples. Initially, physical properties of oil were considered indicators of equipment wear. However, meaningful results were not obtained until examination emphasis shifted to the analysis of oil carried wear particles. The term oil analysis has thus become synonymous with oil particle analysis or wear particle analysis. The oil analysis concept is based on the premise that all mechanical systems experience the process of wear during operation. This wear process results in the production of wear debris or wear particles which are indicative of the respective wear surface condition. In the case of lubricated components these wear particles are picked up and circulated by the lubricant. Thus an examination or oil analysis of a lubricant sample can non-destructively reflect component surface wear condition. Several oil analysis techniques are currently being utilized for government and industrial applications. These techniques, however, lack a firm technical basis. They are based on assumed monitored parameters and are applied, with varying degrees of success, by trial and error. PROGRAM DESCRIPTION In 1973, the Naval Air Engineering Center, under the sponsorship of the Naval Air Systems Command, initiated an ambitious research program with the goal of placing oil analysis on a firm technical basis. This goal was to be accomplished through an intensive correlation effort relating wear particle characteristics to component surface wear condition. Such wear particle correlations serve to hi-light critical characteristics/parameters which are directly relatable to component surface wear condition. 27 The Naval Air Engineering Center Program consists of two phases: a laboratory phase and a system application phase. (1) Laboratory Phase The laboratory phase of the program involves the bench testing of prime oil lubricated components. The objective of this phase is to monitor both the component surface wear condition and the respective generated wear particle characteristics throughout the life of the component. Correlations can then be drawn between surface wear progression and particle characteristic/parameter trends. Trend analysis will serve to identify critical particle characteristics which reflect the wear state of oil lubricated components. (2) System Application Phase The second phase of the program involves the application of laboratory findings to "field" operating systems. The objective of this phase is to verify laboratory component test results and conclusions. Laboratory bench testing has been completed under the program and the system application effort is presently near completion. PROGRAM RESULTS AND CONCLUSIONS Oil analysis has traditionally been thought of as a one step, straight forward testing procedure. However, in actuality, oil analysis is a critical combination of techniques, test procedures, and analysis criteria. The total progression can be divided into five major elements; sample, detection, diagnosis, prognosis, and prescription. (1) Sample The sample element of the oil analysis process involves obtaining a timely, representative lubricant sample from a mechanical system. 28 (2) Detection The detection element provides a first cut or preliminary determination as to the health of a machine. (i.e. is the machine wearing normally or abnormally?) If no abnormalities are detected, no future analysis is required. If an abnormality is detected or suspect, the lubricant sample is transitioned to the next oil analysis element. (3) Diagnosis The diagnosis element serves to further clarify the machinery wear abnormality. It provides a determination as to what machine component/ components are wearing and proceeds to define what wear mode/modes are present. Based on these determinations the analysis is transitioned to the next element. (4) Prognosis The prognosis element of the analysis process serves to define the course of the machinery wear abnormality. It provides a prediction of the residual life/time till failure based on the wearing component and the wear mode progression. (5) Prescription The last element of the analysis process serves to define a course of corrective action. It provides maintenance recommendations based on residual life, wearing component, and wear mode. The sample element of the process is based primarily on technique. The elements of detection and diagnosis are based on monitored wear particle characteristics and/or parameters, and lastly, the elements of prognosis and prescription rely heavily on analysis criteria. Primary objectives of the Naval Air Engineering Center research program is to define critical parameters that would effectively implement the detection and diagnosis oil analysis process elements. For this reason, discussion in this paper will be limited to these two elements. 29 SUMMARY Detection As indicated above, the purpose of the detection element is to provide a determination as to the health of a machine. Based on the program testing, severe wear can be effectively detected by monitoring both wear particle quantity and size distribution. Particle quantity as well as the ratio of large (7-15 ;um) to small (2-5 ^um) particles increase substantially during an abnormal wear situation. Diagnosis Once the abnormality has been detected, it is the purpose of the diagnosis element to provide a determination as to what component /components are wearing as well as what wear mode/modes are present. Again, based on the program testing, further clarification of a severe wear condition can be accomplished by performing an elemental analysis and a morphology classification of the wear debris. It is obvious that elemental wear debris analysis can be directly related to wear component material. Wear particle morphology (i.e. shape, size, color, etc.) also can be related to a wear component and further serves to reflect the relevent wear mode. These morphology relationships have been thoroughly documented under the oil analysis research effort. Present oil analysis techniques are based on assumed monitoring parameters and applied by trial and error. The Naval Air Engineering Center has instituted a program with the goal of placing oil analysis on a firm technical basis. Oil lubricated components have been bench tested in order to formulate correlations between wear particle characteristics/parameters and surface wear progression. Critical parameters defined by bench testing and verified by field testing are: - quantity - size - elemental composition - morphology 30 is o H >H CO H W H M ss o co pi d H H Pi w tn >H w <; m S H a ph hJ H S Pi <; co w c^ <; h < u P M i-J o is Pi O" P W S < u 1 1 _ i CO CO w u o Pi Ph o M CO M o w p CO >1 CD d •H Ph 53 53 CO CO O o M M M M CO CO H H o O Ph U 53 53 M W o O Pi H CVI CVI LO LO 22,^ CD 0) u d bO •H CO ^" Lt ^3 qc CD O < (/> -I cc llO 25 CO ^ GC (/> o CO UJ to §*; ^ UJ coz IX) "> CO T in cs "f-5 Q 00 t— 0) QC 3 Sj uj GO "J 00 3 **- _ _ — I uj = u_ in ii 2 3 i *■ - 1— r~ u_ O O s s • •*» -8 • • • •• • ••« 3- ..•r— — o =3 CJ >- O UJ O LU cc s § s 0) u 3 M •H -I 1 1 IT o o o o o •- CM r> * u» e f o -i — i — i — i — i — i — i i i — i — r SSSSSSSSSSSS UQ 39UV1 k6 UJ (3 < < > Q < UJ Q < DC H >- i — CD Fo ZD a >- LU i o^ ■ ZD CC. LU ZC 1— 00 « 1— LU UJ LU LU j CO CD lu zc a ° 1 a cd dz 1 — CD LU 1 QC T CC CO CD . LU 5 LU 1— CD i i i 2 a LU oc CO 1 >- LU CO - LU CC LU Q_ X LU CD CC LU n t— CO CD CO NAVS ATING C IITY OF ^ LLJ cc CD z: >- a 1— CU Q_ CO *~ — ^ 1 1 1 — ! — ZD CO CO > CD f— LU Lt_ CD CD LLJ r^ CO < CO UJ O < DC £8 «3 S § q I & Si 8 |i «fi " (_ 85 ill* • • • 8 id S£p 1»S 8— CC CO CD 4> s *2 £2 t— CO • • • - h- UJ < > — UJ OT Ul < CD 5000 - 4000 M a> M . 3000 X a. < O O CE CC Ul 2000 - 1000 100 i- SCORING FAILURE TIME-HOURS TIME-HOURS FIGURE 5 COMPARISONS OF THE R.T. FERROGRAPH READINGS AND SOA FOR VARIOUS SLIDE-TO-ROLL RATIOS ON THE GRTM 60 1 CD r- Ge: iv Li- es z UJ CD UD DC Qd !H GO GO CD I— 5 CD "" CJ q: go — oo C_3 C3 ►— O CM [3QfUIN9VW 803 QN3331 33S] 90H H01IN0W 110 6l FIGURE 7 OIL MONITOR READINGS VS PERCENT RUNNING TIME FOR THE DISC SCORING TEST SERIES LEGEND NO FILTER 15juM FILTER A 40 p. M FILTER <•> 25 ju M FILTER < 10 mM FILTER NOTE: ARROWS INDICATE POINTS AT WHICH SCORING OCCURRED READINGS IN THIS PORTION TAKEN AFTER TEST COMPLETION AT VARIOUS FILTRATION LEVELS 10 20 30 40 50 60 70 80 90 100 TOTAL RUNNING TIME - PERCENT 100 80 60 40 20 NOMINAL FILTER RATING ,u M 10 20 30 40 50 60 70 80 90 100 TOTAL RUNNING TIME - PERCENT 100 80 60 40 20 NOMINAL FILTER RATING -Z/M 10 20 30 40 50 60 70 80 90 100 ' 100 80 60 40 20 TOTAL RUNNING TIME -PERCENT NOMINAL FILTER RATING ^UM 62 00 en CO O I— Ql, 0^ CO h- GO CD a: CD O co CO a CD u- CO o s o en co ^D 8 — s or: H ^^ fa "^ O z < III Ul o a u [lW00l/S313llHV- c H oo LU CD 00 o Q. >- 6U FIGURE 10 OIL MONITOR READINGS VS PERCENT RUNNING TIME FOR THE BEARING TEST SERIES NOTE: INITIAL FATIGUE SPALL OCCURRED WHERE ARROW POINTS TO PLOTTING SYMBOL LEGEND O NO FILTER O 75 M M FILTER QtO^M FILTER ^ 25 ^M FILTER READINGS IN THIS SECTOR TAKEN AFTER TEST COMPLETION AT VARIOUS FILTRATION LEVELS 10 20' 30 40 50' W 70 M 90 100 TOTAL RUNNING THE - PERCENT 100 80 60 40 20 NOMINAL FILTER RATING u M 10 20 30 40 50 60 70 80 90 100 1 100 80 60 40 TOTAL RUNNING TIME - PERCENT NOMINAL FILTER RA 20 TING-/iM 10 20 30 40 50 60 70 80 90 100 100 80 60 40 20 TOTAL RUNNING TIME - PERCENT nominal filter rating ]m 65 cc O CD O cc id - rv 3: LU a: LU H- CD i — i Q_ S U_ GO 1 — OO CD LU h- CD CD O CD CD CD CD CD CD OO UJ _ I O l — ac «x Q- CD CD CD -^ OO cc QC o CJ <_> o QC «t oc UJ UJ QC u. •a: ii-3 O UJ 1— (J3 t— « Ct GO - Q. QC < t— QC Z 13 UJ o q: u. cc O UJ u_ »— X • iO t— ■— 4 ce ^ . 8 o o o O o o o o ai CO r»- U3 LR «* m C\J o C3 o — o CO o o o O CO >_ Z UJ CJ o 1 — (X UJ CL. o 1 VO UJ s <-> LT> ►— CO z o z *r z ► 9 . o CO C3 68 o o (S) < GO o >- a: O d a. O Lul O 69 CO LU UJ — < *n», >- z H III 2 O (3 O £ o i *3" Z LU -— >- *■ OO LO OO oo oo «* oc r— LU LU LU ^^v LU >- >- >- z ►— =d l_L- 8 oo oo oo oo ! LU LU LU LU | >- >- >- >- ^ Q_ «t QC or O CO o Q_ ■^ CO I QC ^. O QC z QC o LU t" | o LU OO S CD o o f>-- . 1. Ml LO CM o CO < at i- Ul 2 ac oo oo -- LU LU ^ i^ >- >- z O LU 3E oo LO oo oo OO ^ uc QC LU UJ LU LU ^ o o oo >- >- >- ae o U. oo a 8 OO OO OO 00 LU LU LU LU CD >- >- >- >■ ^ o_ <* oc QC O CD O Q. ■ CO h— QC «t O QC ~^ QC CO LU »— O LU oo SE Lk_ QC CD 1 J^ c_j CO oo LU CC o >- < * 71 OsJ PQ ccr. CD l— CD U! I— LU <=? Q_ i— i cd CD CD I— CD UJ CD c_> Li_l LA CD LO CN CD CD J C£ fr- PQ O —I o3 O *-" i-t Q < ^ CD O i— i _l > CQ CO s- 86 2.5 5 54 KHZ DEMODULATED SPECTRUM i ' LOU j F R E Q. i SPECTRUI 260 * goo 260 FREQUENCY - HZ Figure 4 LOW FREQUENCY VIBRATION VERSUS DEMODULATED HIGH FREQUENCY 87 3600/10% 3600/20% \ 3200/40% \ \ N 3200/50% \ \ N. 4320/7100% Figure 5 BEARING LOCUS FOR A PUMP UNDER VARYING LOADS AND SPEEDS 88 N = 1 N = 4 N = 16 N = 64 N = 256 Signal Noise 0.5 1.0 Figure 6 TIME SERIES AVERAGING 89 Figure 7a Top Uncorrected Probe Output Runout— Bottom Memorized Runout Figure 7b To P Uncorrected, Bottom Corrected Signal at Low Speed 90 SI SPEED Figure 8 AMPLITUDE VERSUS SPEED CORRECTED AND UNCORRECTED FOR RUNOUT 91 l " = l/? a„ = ™ / f (0 cos Wnt dt bn= f f(t)SlNW n t dt AMPLITUDE = /a 2 „+ b* PHASE n = Tan" 1 (^) Figure 9 CALCULATION OF SYNCHRONOUS AMPLITUDE MACHINE 1 DRIVER DRIVEN TELEPHONE INTERROGATE SPEED & PHASE MONITOR | ALARM |- 1 ... TRANSIENT DATA 1 I 1 ...... SELECTOR 1 1 MINI COMPUTER RUNOUT CORRECTOR DISPLAY/PRINTER FFT ENVELOPE DETECTOR DISK STORAGE Figure 10 MACHINERY DIAGNOSTICS SYSTEM 92 THE ROLE OF SIGNAL PROCESSING IN MACHINERY VIBRATION ANALYSIS James H. Hamilton Naval Ship Engineering Center Philadelphia Naval Base Philadelphia, Pa. 19112 Machinery vibration analysis is an important tool in the diag- nosis and prognosis of machinery problems, monitoring of vibratory conditions, acceptance testing, and modeling of machinery systems. With the increasing complexity of machinery systems, the more string- ent noise criteria, and the continuing effort to improve operation, machinery vibration analysis is becoming very useful to the designers, troubleshooters, and users of such systems. The need to perform this type of analysis was pointed out by Messrs E. Weinert and R. Badgely in their article entitled, "Down with Vibration", when they reported that "between 20% and 40% of all casualties, forced outages and forced partial outages, are directly relatable to vibration". Statistics such as these certainly emphasize the need for more and improved mach- inery vibration analysis. Since machinery vibration is generated by forces within the mach- ine, proper measurement and analysis of the vibration permit the de- termination of vibration sources and their severity. As a result of this analysis, determination of the machine's health including mech- anical deficiences and degradation is possible. The machinery vibra- tion analysis technique used for accomplishing the above involves the integration of three major areas consisting of signal acquisition, signal processing, and data interpretation. Signal acquisition in- volves the sensing of the machine's vibration at specific locations using transducers which emit signals or waveforms representative of the vibration experienced. The signals are then conditioned using amplification, integration, or other techniques such that they may be processed directly or stored for later playback and processing. Sig- nal processing, which will be discussed in more detail later, involves the processing of the acquired signals using techniques capable of providing required data or information in a desirable format. The third major area, data interpretation, involves analyzing the data provided by processing and identifying the sources of the machine's vibration. This is accomplished by comparing the data to the calcul- able forcing frequencies of the machine based on design characteris- tics. Once the interpretation of data for sufficient locations is complete, the vibratory condition of the particular system is known. The role of signal processing in machinery vibration analysis is to provide the means by which we can obtain useful information or data about a machine from its vibration signals. These signals represent- 93 ing the machine's vibratory motion are complex in nature and are composed of many frequencies at different amplitudes related to the forces being generated by various components of the machine. In gen- eral, the processing will be performed such that the time varying sig- nals are transformed into the frequency domain. Since design charac- teristics of the machine permit us to calculate forcing frequencies associated with the machine at various speeds, processing in the freq- uency domain gives us convenient relationships to use in the interpre- tation of the data. Processing in the time domain can also be very useful and, as we shall see later, has applications in machinery vi- bration analysis. Processing of vibration signals will involve the basic functions of filtering, detecting, and displaying of the data. These functions may be accomplished using analog techniques, digital techniques, or a combination of the two. A concern of many people involved in machin- ery vibration analysis work is whether to use analog or digital tech- niques for processing vibration signals. The main considerations in making such a decision should be the particular processing require- ments involved and anticipated future requirements. In general, digi- tal techniques offer many advantages including decreased processing time required, ease of performing various operations, high resolution, good repeatability, and others. Processing systems are classified by the type of filtering used. The two basic types are Constant Band- width where the bandwidth is invariant with frequency and Constant Percentage Bandwidth where the bandwidth is variant with frequency. The specific type required will depend on the nature of the signals being processed and the intended application of the resulting data. Many signal processing techniques and methods are conducive to machinery vibration analysis. Although most signals will be station- ary and periodic in nature, many of the techniques are applicable to non-stationary signals as well. Primarily, most processing for mach- inery vibration analysis work is based on the Fourier Analysis concept or the representation of a complex waveform in terms of its various frequency components. Such processing will permit one to analyze the vibration amplitudes (displacement, velocity, acceleration) at speci- fic frequencies. This concept and its many variations has given us signal processing methods which are utilized extensively in today's machinery vibration analysis work. Some of the more common methods include: (a) Narrowband Analysis which provides vibration amplitude vs. frequency data using narrow filtering techniques. (b) One-Third Octave Analysis which provides vibration amplitude vs. frequency data using filtering with bandwidths equal to 23% of specified center frequencies. (c) Signature Ratio Analysis which provides vibration amplitude vs. orders of rotational data. This method has the advan- tages of compensating for speed variations of the machine and 9h of displaying orders directly which eliminates the need for many calculations. (d) Frequency Translation which provides a means of separating a selected band of frequencies from the vibration signal and then translating that band to a lower frequency band. This method is very advantageous in achieving increased frequency resolution and in the analysis of modulated vibration sig- nals. (e) Order Tracking which provides a means of determining the re- sponse of a particular rotational order as a function of the machine's speed. This method also permits one to determine excitation sources for system resonances. (f) Spectral Mapping including Time Spectral Mapping which pro- vides representation of Amplitude vs. Frequency vs. Time and RPM Spectral Mapping which provides representation of Ampli- tude vs. Frequency vs. Speed (RPM). A method of RPM Spectral Mapping which has become very useful in machinery vibration analysis is "Shades of Gray" processing. This is accomplished through the use of a computerized spectrum analyzing system with outputs displayed on a line printer/plotter. The spectra are condensed and plotted sequentially at selected RPM increments dur- ing the machine's speed varying or acceleration run. The RPM at which each spectrum was obtained is printed at the beginning of the spec- trum. The vertical span for each spectrum is limited to preclude overlapping of the spectra giving a clearer and more useful display than normally achieved with other spectral mapping methods. The shades of gray effect giving the necessary contrast is achieved by automatically darkening the area under the spectra curves. Some of the more beneficial characteristics of this type of processing in- clude: (a) Provides a quick look at a machine's vibratory characteris- tics throughout its entire operating speed range. (b) Provides easy identification of rotational orders and system resonances. (c) Provides means for determining where to concentrate process- ing effort for more detailed analysis. This includes infor- mation such as which speeds spectrum analyses should be per- formed and which orders of rotational should be tracked. It should be noted that this method overcomes shortcomings of some other spectral mapping methods by providing permanent printouts with- out using troublesome photographic methods, and by not limiting the number of spectra which can be displayed sequentially. Other signal processing techniques widely used as a result of developments in Fast Fourier Transform analyzers include time domain techniques such as signal averaging and correlation. Signal avera- ging is useful in signal enhancement applications since it tends to 95 eliminate or reduce non-related background noise signals. Correla- tion processing provides still another means of signal enhancement in addition to aiding in the determination of similarities between sig- nals. Other techniques using the dual channel capabilities of the Fast Fourier Transform analyzers include cross spectrum, transfer function, and coherence function analyses. These are very useful in the comparison of spectra and, particularly, in the area of structural analysis. Now that we have considered the role that signal processing plays and some of the applicable processing methods, it should be noted that the types of processing required for specific applications will de- pend on many factors such as characteristics of the machinery system being tested, the characteristics of the signals to be processed, and the specific requirements of the analysts or users of the data. Al- though this paper emphasized the more sophisticated processing meth- ods, the most basic of processing methods may be sufficient to satis- fy your requirements. Too often, those involved in processing work tend to "overprocess" and still fail to meet the requirements. In summary, we have taken a brief look at machinery vibration analysis as to what it is and why it is necessary. We have also de- termined that the role of signal processing is to provide the means by which we can obtain useful information about a machine from its vibration signals. Finally, we have considered some of the many pro- cessing techniques available and some of their specific applications. 96 ^ DIAGNOSTIC TECHNIQUES FOR STEAM TURBINES R. L. Bannister, R. L. Osborne, and S. J. Jennings Steam Turbine Division Westinghouse Electric Corporation Lester, Pa. 19113 GENERAL DESCRIPTION AND OPERATING CONDITIONS Diagnostic techniques discussed in this paper are for large steam turbines. An installed turbine-generator can be over 200 ft. in length and weigh over five million pounds. Several turbines, built to operate at specific steam pres- sures and temperatures, are connected in series to drive a generator which produces the electricity our society now takes for granted. Units built in 1900 could generate 2 megawatts of electric- ity, less than a single multistory office building requires today. Some of the turbine-generators currently being built will be able to generate more than 1300 megawatts, or enough electricity to supply the residental needs of over 4 million people. The basi used for two cate gas), or steam fl do not m fossil b lb of st a large 17 x 106 supply, 3500 psi original the f i n a in. o f m c d e s i g steam g o r i e s nuclea ow rate ake ste oi lers . earn per nucl ear lb of inlet s and ex vol ume 1 row o ercury , n of t genera namely r f i s s s than am at A fos hour steam steam team p pand t as it f blad absol he tu t i o n . , fos ion . foss tempe sil f to a turb per h ressu he st flow es , i ute . rbine is determined by the fuel Normally used fuels fall into sil (i.e., oil, coal or natural Nuclear turbines require higher il units because nuclear reactors ratures and pressures as high as ueled boiler may deliver 7 x 106 large fossil steam turbine, while ine, Fig. 1, will require our. Depending upon the steam res can vary from 1000 psi to earn to more than 1000 times its s past the blades. When it leaves ts pressure has dropped to 2 to 3 For large units, reliability is of utmost importance in the economics of electrical energy generation. Modern turbine- generators operate with an availability of over 90 percent; that is, they are available to generate power 90 percent of the time. Unscheduled down time can be expensive as shown in Fig. 2. Using 1976 information, replacement power for a downed 600 MW unit could cost between $200,000 to $300,000 97 per day for make-up power from another unit within a utility Replacement power purchased from a neighboring utility would cost more. TYPES OF OPERATIONAL PROBLEMS Steam turbines generate static and dynamic signals which are indicative of its operating performance. Measurements of temperature, pressure, force, flow, strain, displacement, velocity, acceleration, phase, etc, can be used to estimate the usable work and integrity of the machine. To develop a diagnostic technique that can be used to minimize unsched- uled outages, it is necessary to understand some of the following operational problems: • Rotor Unbalance Rotating parts cannot be perfectly balanced during manufact- uring, due to geometrical variances and the lack of perfect homogeneity of materials. Parts may also change their dim- ensions under operating conditions. Identifying frequency will be at the turbine's running speed. • Rotor Bow When the turbine has been shutdown, the rotor can bow if uneven cooling occurs within its cylinder. A distance detector can be used to measure the rotor's eccentricity when the unit is on turning gear and operating at minimal speed. • Rotor Whirl Uneven leakage of steam through or past seals, at certain load conditions, can cause an energy input to the rotor. This causes the rotor to vibrate at its own natural freq- uency, which is usually less than half of the turbine's running speed frequency. • Rotor Impacting Impacting exists when rotating and stationary parts come into contact. The identifying frequency will be the sub- harmonic of the turbine's running speed which is the closest subharmonic to the rotor's fundamental natural frequency. • Bearing Instability This phenomenon occurs when the oil film between the journal and its bearing is not tapered as it should be for normal 98 operation. A thicker wedge of oil acts to drive the journal within its bearing at less than half of the turbine's running speed. • Rotor Non-Uniformity If the flexibility of the rotor is not equal in ions it will have different moments of inertia, rotor will have a double moment of inertia when deep enough to affect the shaft's flexibility. al 1 di rect- A cracked the crack is Identi fying frequency will be twice the turbine's running speed. • Short Circuit Torque Electrical transients imposed on the generator will cause the turbine rotor to vibrate at 120 Hz. If the rotor has a torsional frequency near 120 Hz, large vibration amplit- udes will occur. • Thermal Expansion When steam is admitted into a turbine, both the rotating and stationary parts will expand. Because of its smaller mass, the rotor will heat faster and expand further than the casing. Axial clearances between the various turbine parts are provided to allow for differential expansion in the turbine. Contact between rotating and stationary parts may occur if the allowable differential expansion limits are exceeded. Displacement measurements on the rotor, casing and a differential between the rotor and casing can be made. Blade Failures When natu are arou exha blad rota runn of t exce whic redu blad i s e steam ral sou always nd the ust, ex e s whic te. Bl i n g s p e he type p t i o n t h can o ced flo e group xci ted . f 1 ow i rces o some n circum tracti h caus a d e v i ed or modes o this ccur t w and ' s thi s inc f exc on-un f eren on si e a v b r a t i at th exci rul e o the high rd na rease i t a t i i form ce re ots a a r i a b ons n e noz ted a is a 1 ast exhau tural d and th on cause i t i e s in suiting nd the v 1 e 1 o a d i ormal ly zl e pass re shown phenome stage o st press f requen e turb blade the d from t a r i a b i ng on occur i n g f r in F i non ca f 1 ow ure . I cy, a i ne i vi br i s t r i he st lity the b at a equen gs . 3 lied press n thi torsi s 1 oa a t i o n b u t i o earn i of st 1 ades harmo cy. ( and blade ure b s cas onal ded, s . There n of flow nl et , the at i on a ry as they n i c of Exampl es 4) . An f 1 utter 1 ades at e , the mode , 99 • Water Induction Severe internal damage to the turbine can occur as the result of unwanted moisture or water being emitted into the machinery. Slugs of water can be carried over from the boiler to the high pressure turbine or from heaters through the extraction lines to the turbine. Static water can also back up through the extraction lines from reheaters and feed water heaters. Damage ranges from broken blades to rubbing, due to differential expansion between rotor and cylinder. When the problem first starts to occur there is no audible characteristic frequency which identifies the problem, but a temperature change will occur. TURBINE SUPERVISORY INSTRUMENTATION Turbine supervisory instrumentation is used to generate alarms and/or trips in the event of the turbine approaching certain defined limits which are indicative of turbine mal- function and/or failure. Measurements are compared with predetermined limits based upon either experience or calculations. Vibration and eccentricity measurements include displays of phase angle as well as detection of the peak-to-peak magni- tudes of the displacement. Since a rotor will tend to bow due to uneven cooling, rotor eccentricity is monitored continuously while the unit is on turning gear and up to 600 RPM. Bowing of the rotor will appear as vibration over 600 RPM. When specified, each measurement can be arranged to be scanned by a computer. Information could be displayed on a cathode ray tube as shown in Fig. 5. AVAILABLE MEASUREMENT AND ANALYSIS TECHNIQUES Since the advent of the first commercially available vacuum tube, electronic devices have been used to sense, condition, record, store and display static and/or dynamic measurements Rapid advancement in electronics over the past decade has provided equipment on a commercial basis which is a definite assist for making on-line analysis of static and dynamic data . 100 The oldest type of detector is the ear and the oldest type of analyzer is the brain. Some mechanics can listen to an engine's airborne or structureborne sound (via a listening rod) and determine the malfunctioning part, however, this skill is more an art than a science. Many types formance o contacting displ aceme these sens mer for me in a m a g n e crystal wh and a capa up. In ad to determi types of s ment stage radiation , of se f rota or no nt , ve ors ar a s u r i n tic f i i c h p r ci ti ve d i t i o n n e d i s ensors , u t i 1 and a nsors a ting ma n-conta 1 oci ty e a 1 i g s t a t i eld who oduces , indue , a fib pi aceme , some ize u 1 t tomic r re av chine c t i n g or ac near c or se ou a sig t i v e , er op nt ut of wh rason a d i a t ai 1 a ry . typ eel e vari dyna tput nal or tics i 1 i z ich ics , i on . ble fo These es tha ration able d mic mo is ve propor eddy c or ph ing re are st x-ray r detec sensor t are s Exam i f f eren vement , 1 oci ty t i o n a 1 urrent otosens f 1 ected ill in s , RF n ting t s may e n s i t i pi es o tial t a coi s e n s i t to ace proxim or can light the de o i s e , he per- be either ve to f some of ransf or- 1 moving i v e , a elerati on i ty pick- be used . Other vel op- infrared Once dynamic movement has been sensed, it must be analyzed so that the observer or control system, can take the appropriate action. A sound level meter or a voltmeter are examples of instruments that will indicate the gross signal level. If the signal happens to be of a single frequency, the reading is quite meaningful. When required, there are various size filters available to make a discrete frequency analysis as shown in Fig. 6. A popular instrument for making analyses in real time is a time-compression type spectrum analyzer. In this device, the analog signal is first passed through a low-pass filter after which it is sampled at a rate that is determined by the selected frequency range of the analyzer and then converted to a digital signal. In addition, data reduction systems employing only digital techniques are available for performing Fourier transforms yery quickly. Such instruments combine high speed proces- sing ability with the flexibility of a mini-computer capable of many mathematical operations, to produce a variety of different analyses. Probably, the greatest advance in experimental procedures over the past decade is due to the use of computers. The ability to reduce massive amounts of data in minimum time has made it possible to obtain a better understanding of the characteristics of complex machinery during various operating phases. 101 EXAMPLES OF DIAGNOSTIC TECHNIQUES USED ON STEAM TURBINES Blade Failures Tuned turbine blading is designed so that resonant frequen- cies of the blade do not coincide with multiples of the machine's running speed or the nozzle passing frequency. During its design stage the resonant frequencies of the blades are typically displayed in a Campbell Diagram, as illustratedinFig.7. Final product verification can be accomplished in a custom- er's plant using radio-link telemetry. Miniature ratio transmitters, operating in the commercial FM band, are mounted in environmental, protective capsules in the balance weight holes in the low pressure end of the turbine. Resis- tance strain gages are cemented to the blading at specific locations to measure the strain amplitude and its' frequency. To date, telemetry has been used numerous times to evaluate the performance of low pressure turbine blades in the field. The strain gages and wire paths will survive at temperatures up to 500°F and centrifugal loading of 20,000 g's. The transmitters and power supply will operate satisfactorily at temperatures up to 300°F and centrifugal loads of 8600 g's in the presence of moisture. The possibly does exist for using telemetry to extract infor- mation off the rotor for supervisory instrumentation. Long- term telemetry usage is currently limited by erosion from water droplets and the power supply. Another example of detecting a blade problem is shown in Fig. 8. This spectrum illustrates the airborne sound levels obtained near a low pressure turbine during normal operation and after a fast transient condition which was a blade failure. Using normal supervisory instrumentation, the turbine did not shutdown automatically as a result of the blade breakage. The increased magnitude of rotor unbalance did not raise the vibration level at one or more of the unit's bearings high enough to produce an alarm or trip condition. Use of external airborne or structureborne sound measure- ments to detect and decipher blade vibration is not as easy a task. It is certainly difficult to distinguish which row of blades may be vibrating when the frequencies of the various modes are similar. The important point, however, may not be which row of blades is vibrating, but that a frequency has appeared in the spectrum that could be critical to further operation . 102 CI earances Stationary and rotating parts sometimes come into contact within a steam turbine. Normally, a rub is not detected until the machine is opened up for inspection. In several instances, Westinghouse has been successful in pinpointing rubs associated with excessive clearance which allow the inlet connection between the inner and outer cylinders of a high pressure steam inlet to vibrate. From data, similar to Fig. 9, it was possible to decide when the bell seal used on these machines had excessive clearance. The turbines were dismantled and the clearances restored to normal. This action minimized the possibility of a future mechanical problem within the inlet cylinder. When the turbine returned to service, the characteristic vibration disappeared. Rotor Thermal Cycles and Cracks A change in blade-path steam temperature will produce thermal stresses in the rotor. Heating of the rotor surface followed by an equal cooling constitutes a thermal cycle and imposes on the rotor a cycle of alternating stress. Cracks will ultimately develop after a number of cycles which depends on the severity of the stress and the rotor's material. The relationship between alternating stress and cyclic cap- acity is a known material property. Therefore, it may be possible to predict the number of stress cycles necessary to initiate a rotor crack. Cracks usually propagate slowly and then have a exponential growth, with a final rapid acceleration of growth to a burst rotor. When the crack becomes large enough, the rotor will have a double moment of inertia which will produce a second harmonic of vibration. Of four field case histories of cracked rotors (not Westinghouse), the second harmonic was reported in two instances. In another case, the propag- ation of a thermal crack was accompanied by an increase in transient vibration during temperature decreases over a period of three months. During a normal turbine outage, ultrasonic and other techni- ques can be used to inspect rotors for cracks and thereby eliminate the probability of a rotor burst on machines which have been in service for a number of years. 103 Water Induction Water induction has been a major cause of forced outages of large steam turbines. Presence of water is first detected by changes in the measured temperature of the affected parts, as shown in Fig. 10. Use of metal temperature thermocouples to determine the presence of water or cool vapor in the turbine cylinder is an effective means of detection, but of no use in the prevention of water or vapor injection. Therm- ocouples placed at strategic locations within the piping system, however, might provide the necessary warning to the operator. Westinghouse is currently investigating other methods of more rapidly identifying impending induction of water into turbines under an Electric Power Research Institute contract. The main objective is to sense the presence of water rapidly enough to close valves or take other action to prevent water from entering the turbine. DIGITAL ELECTRO-HYDRAULIC CONTROL The increased complexity of todays steam turbine generators and the significantly greater investment in each power plant demands a control and monitoring system, Fig. 11, which can maximize the protection and efficiency of the units oper- ation. By combining the flexibility of the digital computer for handling complex logic decisions with the speed and high forces of hydraulics for valve positioning, the Westinghouse Digital Electro-Hydraulic (DEH) meets these needs. The DEH automatically increases the turbine generator speed or load at a rate based on calculated critical rotor stresses, as shown in Fig. 12. All other critical turbine generator variables are also checked. Transducer outputs are monitored for failures. The speed channel has two out of three redundancy checking, with automatic channel switching in case of a contingency. During synchronizing of the generator to the line and while loading the unit a set of control valves modulate steam flow to the unit under command from the digital computer. During initial loading even heating of the turbine inlet is obt- ained by opening each control valve the same amount. This is called single valve operation. Once the turbine is heated, higher unit efficiency is obtained by sequencing these valves to minimize pressure losses across the valves. If a control valve fails to respond, the digital computer will automatically sense this, switch the unit to single valve operation and compensate for the malfunctioning valve. 10U During operation, if there is a contingency in the digital computer the system automatically will switch to manual allowing uninterrupted power generation. Also incorporated in this portion of the system, is the overspeed protection controller. A second, redundant overspeed trip system also protects the unit, During all phases of operation, the computer constantly protects the unit, not only by scanning the actual values of existing parameters, but also by anticipating future values. Corrective measures are taken to avoid alarm or trip conditions. A modern centralized display provides communication between the turbine-generator and the operator. Through the use of control and display panels, a cathode ray tube, and a type- writer, the operator can digitally disptay information such as temperatures, pressures and vibrations. A continuous monitoring and alarm system assures effective and reliable operation by supplying the operator with timely information EXPERIMENTAL FACILITIES The Westinghouse Steam Turbine Division has an array of laboratory and experimental facilities that are engaged in developing, analyzing and verifying steam turbine design. Some of the facilities available to help develop new diag- nostic techniques include: • Rotor Dynamics This facility is a single mass rotor supported by two jour- nal bearings, with precise speed control up to 7000 RPM. Bearing geometry, oil viscosity and rotor unbalance can be varied. The facility has been used to analyze response to rotor unbalance and study instability phenomena such as oil whip. 105 For dynamic simulation of control system design and plant dynamics, advanced computer techniques are used in this facility, Figs. 14 and 15, to analyze frequency response, hysteresis and operating and stability characteristics. Prototypes of new control systems, operated in conjunction with computers simulating plant dynamics are tested for reliability, durability and dynamics to assure trouble- free operation in the field. • Data Acquisition and Reduction Several facilities are available for precise and responsive data acquisition, on-line data reduction, recording and storage for further analysis. Automatic traversing of probes can be computer-directed in preprogrammed steps. Pressure, temperature flow, speed, torque and frequency data can be printed out for performance monitoring. Vibration stress and amplitude levels, frequency and phase can also be printed out on a routine basis from a dynamic test. Using a Fast Fourier algorithm technique, one system proces- ses the dynamic data using standard calculation techniques and presents the results in terms of linear or power spect- rum, auto-or-cross-correl ati on function, or coherence. This system has also been used to generate orbital plots of bearing data, decay curves of damping tests and analyses of acoustical data. FUTURE DIRECTIONS The high capital investment and need for extreme reliability in electrical power generation dictate expansion of the turbine generator monitoring and diagnostic funtions. Areas such as analysis of the vibrational characteristics of the rotor-bearing system, continuous monitoring of the chemical 106 purity of steam , added water induction protection and increased long term recording and trending of machine behavior are prime candidates for future systems. Such systems could be incorporated into the existing DEH or structured in individual subsystems for retrofit to existing machines. Completely automated actions will be in general reserved for those cases where the operator does not have sufficient time to react. Emphasis will be placed on giving the operator increased information relative to decisions he has to make. 107 Figure 1. 1300 MW nuclear steam turbine generator, 200 400 600 800 1000 RATING , M W 1200 Figure 2. 1976 utility make-up power cost, 108 ^^v ^^- ^Sis. ^^^ ^©fcv 1st MODE * \^^^^ 2nd MODE 3rd MODE Figure 3. Typical first three modes of a group of last row blades. Figure 4. Hologram of iast row blades showing deflection pattern at 4783 Hz . 109 Z H 2 V & s o 00 o rH CN II « II U H w Eh Q < « o oo II «* CN Pi 00 U z II H fa ^^ fa ^ ti CO CD H ffi ^ Z Q H P CM CM O H II II K ffi X X W W «3< o in iH oo II II CM CM a s H H in O H O «=tf O CM rH II II Pi Pi ffi EC Eh Eh *— * -^* H co fa Cm *^— " ' ' CO CO CO CM W Pi w CM H CO <: p < u Eh in m o Pi Pi CQ CQ H CQ EC H > Cm H fa HI Eh Q P Q P O P Eh O CM oo ro r- t-- to cu tr> cu o e O fO JH o -p rcJ U Q) CM rH •H -P fl CO CU £1 w (d ■h m cu Cn ftS to co cu 6 m CN CN fa Cm Pi r-i ro in II CQ H CO Pi • Eh CO H Pi w CO U P CO in O CTl a w < P • ro CM fa p Eh H in m CN CN u o H 2 EH H «i Eh Eh o D s p • r~ CN H O a o • O O < rH •sT CN CM CM Pi P O P CQ II CO H Z CO > CM 00 <£> VT, w w H • CTl CTl CO Eh u Pi Z Eh ro o rH H CM H H 2 Eh Pi CO W H "—' CM ro •^r CO o m 2 It • • H Pi H W o <£> P CQ > Eh >1 cd H CM CO •H CU & -p >1 rd U CU o ■d -p rd u in cu U tn ■H fa 110 500 1000 FREQUENCY (Hz) 1500 Figure 6. Analysis of sound signal using different bandwidth filters. 480 420 360 i J i / 1— BLADE 30 ° FREQUENCY Hi 240 1 y 180 ,!<-"■ 120 -1— — 60 HARMONIC 3 2 1 1000 2000 3000 ' 4000 SPEED - R.P.M. 360 ° Figure 7. Typical Campbell Diagram for a group of low pressure blades. Ill 100 90 NORMAL AT BLADE FAILURE 100 W^fjM^W^^^^^J^J^^^^^^^ 200 300 400 500 FREQUENCY (Hz) Figure 8. Airborne sound spectrum before and at a low pressure blade failure. S~ 100 « lUUr I to 5 90h 80 w 70 l i..JuJumkkJoLk 100 200 300 400 500 FREQUENCY (Hz) Figure 9. Airborne sound spectrum of a bell seal clearance problem. 112 800 700 f- 600 500 400 300 h 200 100 h © 340 MW TRIP © TURNING GEAR ON © TURNING GEAR TRIP © TURBINE DRAINS OPEN © TURNING GEAR ON I INNER CYLINDER COVER " INNER CYLINDER BASE 2 4 6 8 1012 141618 202224 26 HOURS FOLLOWING INITIAL TRIP Figure 10. Abnormal high pressure - intermediate pressure turbine cylinder cooling due to water being induced accidentally into cylinder by flooding from a feed- water heater. Unit could not be placed on turning gear until there was an acceptable temperature differential at point 5 after a trip occurred at point 3. 113 u •H iH 3 CO U •3 >i £! 1 • U e -p CD U +> o W H >i CD CO rH H (0 o -p 5-1 ■H -P tr> C •H Q U CD M •H 111+ Figure 12. Automatic turbine control-speed/load profile. Figure 13. Industry's largest steam turbine test facility. 115 Figure 14. Controls laboratory test stand hydraulics. m \ Figure 15. Computer and simulation area in control laboratory. 116 EXPERIMENTAL DETERMINATION OF RADIAL MAGNETIC FORCES AS A FUNCTION OF ROTOR OFFSET IN A LARGE INDUCTION MOTOR Robert L. Leon Franklin Institute Research Laboratories 20th and The Parkway Philadelphia, Pennsylvania 19103 Hundreds of motors, used to drive compressors in gaseous diffusion plants, have been uprated from 2000 HP to 3300 HP. Bearing failures have been experienced on some of these reworked motors and the Franklin Institute Research Laboratories (FIRL) was asked to assist in finding the cause. Since the flux density in the gap had been substantially increased, the radial magnetic force, which acts in the direction of the minimum gap, and is proportional to the square of the flux density, became a prime suspect. Preliminary estimates from the motor rebuilder indicated that a radial force of 5,400 lb would result from only a 10 mil radial offset of the rotor within its stator. Computer analyses indicated that a radial force of less than half that estimated value could lead to bear- ing instability or to minimum film thickness failures depending upon its direction. A test was run on one uprated motor to experimentally determine the radial magnetic force acting on the rotor, and to see how this force varied as the gap was changed. The test served to check not only the estimated magnetic force levels, but the resulting bearing behavior as well. To determine the magnetic force reactions at the bearings, the horizon- tal and vertical displacements of each bearing housing were measured upon the sudden removal of the magnetic force. This was brought about by cutting the excitation to the motor at full speed. In order to maximize these displacements, FIRL reduced the stiffness between each bearing housing and its mounting flange by machining away material so as to leave four supporting columns, two vertical and two horizontal. Optimum vertical and horizontal stiffnesses were targeted at 10 7 lb/in. This value is more than 5 times the maximum film stiff- ness of the bearing, thus not affecting rotor-bearing performance. Yet it is compliant enough to yield ly in. displacement for each 10 lbs of force. With each displacement sensor able to resolve less than 2y in., force resolution would be 20 lbs. Since the anticipated forces were to be in the 2,000 lb range, this resolution was felt to be more than ade- quate. The actual stiffness values as determined by calibration tests on site were 1.4 x 10 7 lb/in for the horizontal stiffnesses, and 0.5 x 10 7 lb/in for the verticals. The displacement sensors were noncontacting eddy current probes. They were affixed to the mounting flange and they "looked at" copper target 117 buttons bonded to the bearing housings in the plane of the mounting flange. Copper was used due to the higher sensitivity so obtained, 500 mv/mil vs. 200 mv/mil for steel. The price paid for doing this is re- duced linear range and smaller gap setting, neither of which was a problem in this case. A Brush 6-channel strip chart recorder was used to plot the change of the vertical and horizontal displacement components of each bearing housing that resulted from the loss of magnetic force when the excita- tion to the motor was cut. In order to eliminate the high level vibra- tion portion of these four signals, they were passed through sharp cutoff (48 dB/octave) low pass filters set at 1 or 2 Hz. The filters also had 20 dB gain, thus raising the sensitivity to 5 volts/mil or 5 mv/u in. The fifth channel of the strip chart recorder displayed the analog speed output of the motor so that coincidence with motor cutoff could be noted. Other sensors mounted on the motor included orthogonal shaft probes in each bearing for journal position and orbits, and horizontal and vertical accelerometers on each bearing housing. Another proximity probe was used to sense motor rotational speed by looking at the passing of a keyway. Still another probe was used to sense the rotational speed of an oil ring by looking at the passing frequency of a steel insert. All the sensor outputs were recorded with a 14-channel FM tape recorder allowing for relative amplitude and phase determinations during subse- quent playback. Two dual-channel oscilloscopes were used for continuous observation of both journal orbits. Also, a single channel FFT spectrum analyzer was used to view various frequency spectra and time domain wave forms, and these were recorded for later analysis using a digital cart- ridge tape recorder. The moving of the bearing housings to vary the air gap was accomplished with jack screws mounted to the end bells. Flats were provided on the mounting flanges for the jack screws to act upon. Measuring shims were used to measure rotor to stator air gaps at the quarter points at each end. The flange mounting bolts were loosened and using the jack screws, each bearing housing was moved till the measuring shims indicated the rotor to be approximately centered. The bolts were then retightened, the air gaps rechecked, and plastic tipped locking bolts, which were used to hold the bearings tight in their seats, were snugged down. In addition, using a telescoping gage and micrometer, the gaps between the jack-screw mounting blocks and the flats on the mounting flanges were measured and recorded. This entire procedure was to be repeated each time the rotor to stator air gap was changed. The motor was now ready for its first run (rotor centered), following calibration which was accomplished by applying a 100 lb force in the vertical and horizontal directions at each bearing. All the initial tests were made with the motor sitting on rubber pads on the test floor, running with no load, for various rotor to stator air gap conditions. Later a coupling would be put on and the motor moved to 118 a test stand where it was coupled to a load generator and run under full load. The sequence of testing was as follows: Rotor "centered"; 10 mils low; 20 mils low; 10 mils high; 20 mils high; 10 mils right (looking at brake end of motor, having CCW rotation); 15 mils left; 15 mils left under full load. Figure 1 illustrates the resulting horizontal and vertical magnetic force components for each of these conditions. The forces generally follow what would be expected. That is they act in the direction of the rotor offset (minimum gap direction) and they increase as the offset increases. However, the magnitude of these forces is far below what had been predicted by the motor rebuilder, approximately 20 lbs per mil-of- rotor-offset vs. 270 lbs per mil, both of these on a per bearing basis. For the full load condition, the very high vertical couple is believed to be the result of coupling misalignment and represents actual forces under load due to misalignment. When the excitation is cut, the cou- pling no longer transmitting torque experiences sudden tooth to tooth freedom and these misalignment forces disappear in rapid fashion just like the magnetic force in question. Thus they show up as apparent magnetic forces. This contention is supported by the spectrum of the coupling end vertical shaft sensor which indicates a 2X running speed displacement component 10 dB higher than the IX running speed component while running under load. Measured journal orbits indicated little effect from gross (20 mil) rotor offsets, and no discernable shift in the position of these orbits upon the sudden removal of the excitation to the motor. This supported the conclusion that the magnetic forces were small. While this test was showing that the magnetic force was not the culprit, it indicated that the bearing oil rings were. The rings were seen to operate erratically and far too slowly at best. The worst were new perprint rings, as had been put in the uprated motors. New, better operating, redesigned rings are now being installed as the motors are reworked. This technique of filtering displacement signals to determine low frequency or step application forces on "spring" structures in high vibration environments can probably be used to good advantage in many diverse applications. 119 ROTOR POSITION RUN 1008 CENTERED RUNS 1001 1002 10 MILS DOWN (UNDER LOAD) 20 MILS DOWN KEY RUN ¥ 1004 COUPLING END BRAKE END 10 MILS RIGHT 500 LBS 250 LBS APPROX. SCALE tl Figure 1. Magnetic Force Components at the Bearings 120 SESSION I I I NEW DETECTION , DIAGNOSIS AND PROGNOSIS TECHNIQUES AND EQUIPMENT CHAIRMAN: R . A . COULOMBE NAVAL SHIP ENGINEERING CENTER A NEW CHIP DETECTOR - RELIABLE, VERSATILE, AND INEXPENSIVE Dr. Thomas Tauber Technical Development Company, 2k E. Glenolden Avenue Glenolden, Pa. 19036 For over a decade, electric chip detectors have been installed in virtually all helicopter transmissions and engines and in many pro- pulsion and drive systems of military fixed-wing aircraft. They con- tain a magnet which attracts the ferrous debris shed into the lube system by failing components and retain it for later visual inspection and failure verification. The debris particles which fall across the gap between two electrodes act like a switch and activate a chip warning light (see figure 1). Electric chip detectors are surprisingly reliable indicators of in- cipient failure. For example, during a two-year period studied by the Army, chip detectors on a part of the UH-1/AH-1G fleet scored 75 "hits" (indications followed by component removal) against 9 "misses" (component removals without chip warning) for an overall success rate of 88%. For the transmission chip detectors, the success rate was even higher with kO hits and one mi ss , corresponding to 97.5% success. This indication reliability is comparable to ASOAP experience overall and outscores it considerably for the transmissions. Unfortunately, as is widely recognized, conventional chip detectors are subject to a large number of "nuisance" indications which are not caused by failure-related debris. The same study shows that in addi- tion to the 75 correct indications, there were 529 nuisance indications; 305 or 58% of them were reported as being due to "wear fuzz" and 130 or 25% due to electrical failure of the chip detection system. The remaining cases were the result of moisture in the lube system {h%) , carbon and other metal contamination (6%) and 1% were due to unknown causes. Such statistics have made chip detectors controversial over the last few years, despite the high reliability of failure indication mentioned earlier. In order to eliminate the nuisance indications and make chip detectors more dependable, their causes must be understood more clearly. It goes without saying that the electrical failures which accounted for 25% of those indications are avoidable by using state-of-the-art connectors and test circuits. Clearly, the "wear fuzz" indications are the most important category to be eliminated. A model for the development of the debris spectrum reflecting a typical 123 fatigue-type failure (bearing or gear spall ing) is shown in figure 2. Under normal operating conditions, a background consisting of appreci- able quantities of particles usually less than 30 microns in diameter are continuously being produced by normal wear processes. In the diagram, the particle production rate in mg/hour is plotted on the ver- tical axis rather than the accumulated total. Therefore, the debris spectrum would remain unchanged along the time axis if no failure occurred. As a spall develops as a result of a local fatigue crack and then prop- agates across the load-bearing surface, more and more debris particles of larger and larger size are being produced. This causes both the spectral maximum and the large-particle threshold to move to the right. Simultaneously, the production rate of small particles also increases, permitting indent i fication of the failure by spectrographic analysis. The production of debris particles of approximately 500 micron and over leads to failure indication by the chip detector. In lube systems with very fine filtration (e.g., the newer 3 micron- absolute filters), most debris particles of the 10 micron class are removed as quickly as they are produced. With coarser filtration, these particles slowly accumulate in the oil and, being mostly ferrous, are deposited on the chip detector. Gradually, a conductive path forms across the chip gap and the chip warning light comes on. This process seems to be the origin of most "wear fuzz" nuisance indications. In addition to other strong evidence, this is indicated by the fact that during the development program of the T-700 engine, which has a 3 - micron absolute filter, no such "wear fuzz" indications were reported. Unfortunately, many chip detector installations, especially but not exclusively in engines and transmissions of older design, add greatly to this problem. Figure 3 shows a typical lube system schematic of an engine, with the chip detector installed in the main scavenge line. For any given size of debris particle, there is a finite probability e] that the particle will be transported to the vicinity of the chip detector (point A). For a large number of particles, this probability becomes the debris transport efficiency of the lube system. In the system shown, this quantity would be zero for particles larger than the openings in the pump inlet screen. Once the debris is transported to the vicinity of the chip detector, some provision must be made to either slow down the oil flow so that it can settle out or a screen must be used to capture, retain and deposit it on the chip detector. In many installations, neither is the case and the result is a low debris capture efficiency e2- Finally, the indication efficiency e$, the probability that a particle, once deposited on the chip detector, will be indicated, corresponds to the sensitivity of the chip detector and is largely determined by its gap size. If e|.e2 is low for failure-type debris, the chip detector 12 k sensitivity must be high to compensate for this. In practice, this means a small gap size which, in turn, leads to more nuisance indica- tions. This is due to the fact that the transport efficiency of any lube system is practically equal to one for particles smaller than 50 microns. It becomes obvious that a conventional chip detector cannot successfully cope with this problem, at least not in lube systems with coarse fil- tration. The reason is clearly that a chip detector responds with an indication each time the chip gap is bridged, regardless of whether the gap consists of large quantities of small or small quantities of large particles, or even of a fine, long sliver. This problem can be resolved by sending a strong current pulse through the chip gap every time there is electrical continuity. The pulse is generated in a small capacitor network which can be located in the chip detector (see figure 4) or elsewhere. It causes a local melting of the kind of particles that are responsible for these indications which interrupts the current path. Since this process occurs within a few milliseconds, local lubricant temperatures and flow conditions have no influence on this process. On the other hand, debris particles with more substantial cross-section conduct the current pulse unharmed and the continuity remains. This method can be applied in several different ways. Firstly, the current pulse discharge can be initiated automatically every time conti- nuity occurs. The chip light is therefore not activated, unless the debris survives unchanged the effects of the pulse. The result is a chip detection system with a greatly reduced tendency to nuisance indications. For example, the Hughes YAH-64 helicopter, whose trans- mission is equipped with such a system, experience no "wear-fuzz" type indication through 500 hours of flight tests. The Sikorsky S-76 helicopter is also equipped with such a system. A full-flow chip detector with screen located in its transmission is shown in figure 5. Secondly, the current pulse can be initiated by the pilot ("manual" system). In this case, the chip light comes on each time debris bridges the chip gap, regardless of its cross section. The pilot ini- tiates the pulse by pushing a panel switch and makes the usual pre- cautionary landing only if the chip light stays on. If the chip light goes out but comes on again after a short time, this indicates the continuing generation of fine debris particles. This may be associated with an early stage of a bearing spall, with a gear-scoring type of failure or with excessive spline wear. If the current pulses are con- tinued, this type of debris eventually saturates the chip detector and the chip light can no longer be extinguished. This point corresponds to the onset of the chip indication in the automatic system discussed earlier. If desirable, it can be set at an arbitrarily low threshold by selecting gap size and capacitance for high sensitivity. 125 With this method, prognostic information can be obtained about the developing failure. This can be made more practical by incorporating a counter into the chip detection system which records each successful current pulse. This technique can be used for both automatic and manual systems. Figure 6 shows a four-channel chip detection system for a helicopter installation (only one of the chip detectors is shown). The panel switch is activated when a chip indication occurs and if the current pulse successfully removes the indication, the corresponding counter advances. The pulsed chip detection system is very flexible. For example, it is equally adaptable to full-flow and to sump-type lube systems and is easily retrof i ttable into aircraft with existing chip detector instal- lations. It also retains one of the most important positive character- istics of most conventional chip detectors, the quick access to the collected debris for visual failure verification without lubricant drainage. Furthermore, operational experience in several flight test programs has shown that the pulsed chip detector offers greatly improved dependabi- lity and reliability over conventional chip detectors. Its compara- tively low cost therefore qualifies it uniquely for future real-time, airborne debris detection requirements. 126 o o ZLl ceo h o ®-||| — II'" -j g. o I o 2 a: uj a, o cc i UJ Q £L X o UJ tr 127 2 Z> a: t- o ui a. to UJ _i o h- 2 m UJ o CM UJ a: O 128 ^FAILURE OCCURS HE RE X' "A" ■ ® T — ' \ ■ • aajHf ,8, N SCAVENGE PUMP ' 1 1 | ■ f ~ PUMP INLET SCREE TTghip DETECTOR TRANSPORT EFFICIENCY £, = PARTICLES TO POINT "A" PARTICLES GENERATED CAPTURE EFFICIENCY ^2 - PARTICLES ON CHIP DETECTOR PARTICLES TO POINT "A 1 INDICATION EFFICIENCY £ 3 - PARTICLES INDICATED PARTICLES ON CHIP DETECTOR FAILURE INDICATION PROBABILITY e-e,- CD Q CD o -a •<- CD -C N CD ro o o CD 6iupeay uol^cpgj68q pazLLeuuoN 1A9 to 1 -o o» cc C\j o O) C7) ra s- a> > to S- to CO E •i — -a rO 0) a: E o •1 — +-> to n3 01 "O fO S- s: CJ1 a) "O Q 0) N C •i — O ( •i — m +-> F ra <^ ■ O 3 o CM S- 13 C7> O 6uLpeay uo.L^epeu69Q pazLLeuuoN 150 o Q 151 USE OF MICROPROCESSORS IN ANALYSIS OF ACOUSTIC EMISSION WELD MONITORING DATA* R. N. Clark and T..A. Mathieson GARD, Incorporated 7449 North Natchez Niles, Illinois 60648 Abstract: The use of Acoustic Emission as a nondestructive technique for real time weld evaluation creates a need for supporting analysis equipment. This equipment is needed to extract the information required for setting parameters on acoustic emission weld monitors as well as exploring the application of Acoustic Emission to various welding materials and processes. The equipment requirements are fast analysis turnaround time, simplicity of operation, and maximum information extracted from the weld. The authors are developing a system for this analysis using wideband tape recordings from a modified video tape recorder and a microprocessor-based computer. Analysis is made of amplitude frequency content and pulse-realted information of the acoustic emission bursts, and presentation of the analysis is designed for sim- plicity and efficiency in the operator-computer interaction. The paper presents current results of this procedure, a description of the ana- lyzer system and associated equipment, and possibilities for future work using microprocessors in this and in other areas. Introduction The study of Acoustic Emission Weld Monitoring at GARD began in 1971 with an effort aimed at real-time NDE of railroad tank car welding. Our goal was to minimize radiography, particularly the shooting of multiple radiographs of the same weld. We hoped to do this by performing "on- the-spot" inspection of welds as they were completed to allow immediate repair and correction of welding problems. Acoustic Emission is by its nature an "on-the-spot" phenomenon, and was determined to be a suitable solution to this problem. Many researchers have shown that useful information may be obtained acoustically from stressed metal specimens. 1 The welding of metals produces a unique situation where stress is generated by the thermal contraction of the cooling weld bead. As the weld cools, stress causes the growth of flawed areas, particularly cracks. Since the release of _ energy in a growing flaw is mechanical in nature, it is also acoustic in * Some of the data presented in this paper was acquired as a direct of indirect result of work done under U. S. Nuclear Regulatory Commission Contract No. AT(49-24) -0187. 152 nature, with the frequencies of interest ultrasonic and generally lying between 100 KHz and 1 MHz. Many problems are involved in Acoustic Emission Weld Monitoring, how- ever. These are generally due to extraneous and interfering noise. This noise is due to a multitude of sources, including the noise of the electrical arc, cooling slag, and mechanical scraping or pounding noises. These noises tend to mask the useful information which comes from the weld. GARD has demonstrated that Acoustic Emission Weld Monitoring can take place even in the presence of interference, by the use of signal pro- cessing techniques . ^ > 3 The first such techniques were discovered by trial and error, and involved many hours of manually adjusting para- meters and replaying taped data. Our experience is now sufficient to allow a more sophisticated approach to data analysis, and we felt that future use of Acoustic Emission Weld Monitoring was sufficient cause to develop an automatic data analysis system to fulfill our needs. It would allow us to efficiently study new welding materials and new weld- ing processes, to establish parameter settings for Acoustic Emission monitoring for particular applications, and to improve settings for systems in use, such as at our tank car manufacturing facility. To do this economically, such a data analysis system would have to: 1. Use either live or tape recorded weld data as an input. 2. Have the capability of holding large amounts of data obtained during one weld and to conveniently store data on all welds observed for subsequent reprocessing. 3. Have sufficient flexibility in formatting data input and output both to the convenience of the developer and for efficient operation of the processor. 4. Offer simplicity of programming for efficiency in altering the way the data is processed. 5. Be capable of gathering data in an uncontrolled environment, such as weld lab or production floor. We would have been forced to use a large-scale computer as the core of our system if microprocessors hadn't been developed. The small size, low cost, and programmability provided by microprocessor systems enabled us to more than meet the above requirements of Acoustic Emission analysis. They have allowed us to develop a small, on-line, programable monitor which will be discussed later. The introduction of the microprocessor to the electronics industry ranks as one of the most exciting things to happen in recent years. The 155 ies development of Large Scale Integration (LSI) manufacturing technique allows the reduction of all of the necessary functions of a computer system to the size of a few tiny integrated circuit "chips". In addition, mass production of the chips can and has brought the cost down dramatically. For example, the Motorola 6800 microprocessor, which has more computing power than many early computers that cost hundreds of thousands of dollars, can be bought for $35.00. This low cost allows the equipment designer to consider applications which he may otherwise have rejected, or for which he would have used fixed, hardwired logic elements . The microprocessor may be thought of as a universal logic element whose function may be changed by means of a stored "command". A series of these commands is called a program, or simply "software". The program- mer may remove an instruction, and insert another in the program and easily change the function of the microprocessor. Thus modifications are easily made, and mistakes easily corrected. In an application such as ours, turn-around time on a large-scale machine is, at best, a matter of minutes. At worst, it can be up to a half a day. Microprocessors allow one to conveniently set up a proces- sor that is dedicated soley to ones' project and can reliably give turn- around times in seconds. Large scale machines are still the data storage champions, but with mass storage peripherals (floppy disc, cassette) microprocessors are able to retain more data than we can presently conceive of gathering. In the formatting of data exchange, the microprocessor has a definite advantage. To suit many diverse users, large scale general purpose machines are programmed to accept input according to a limited number of formats. Each user must therefore organize his data toward the processor's preference rather than to his individual situation. The microprocessor allows one to easily construct his own computer and to wire and/or program it to accept data in a format convenient to his individual requirements . Programming a microprocessor is no more difficult than programming a large scale machine. Our processor was easily programmed in assembly language, but systems are on the market that allow programming at higher levels, such as BASIC, PL/1, etc. An area in which microprocessors are better for our applications than a large-scale machine is in their greater mobility and tolerance to environmental variance. Large-scale machines are set in air-conditioned rooms. You go to it, it doesn't come to you. Microprocessors are portable and capable of being run anywhere a human will be comfortable. 15^ Finally, when a large-scale machine is programmed, any further use of that program must be done on that or a similar machine. When a micro- processor is programmed, the job of building a dedicated processor is virtually complete. The computer can now exist as a field-portable electronic device by hardening the programs into PROMS. The present configuration of the acoustic emission data processing system is shown in Figure 1. The following discussion of its operation refers to the block diagrams in Figures 2 and 3. Data Gathering and Preprocessing The function of this portion of the analyzer (Figure 2) is to: 1. Acquire acoustic emission information from the weld, 2. Break the analog information down into characteristics of interest and 3. Put these characteristics into a digital format compatible with the digital processor. Information is gathered from the weld by means of (A) a piezoelectric transducer or transducers sensitive to surface mode vibrations in the ultrasonic range. Our experience has shown that the range of frequen- cies of interest is about 100 KHz to 1 MHz. Although information is probably present above 1 MHz, it becomes very difficult to get useable real world data in this region. We are continuing with our investi- gation in the >1 MHz region. The analog signals are amplified after leaving the transducer (s) , and may be tape recorded at this point (B) . We use a Sony Video tape recorder, AV3650 which is modified to defeat the servo system on the rotating head deck, but other equipment may be used. An audio control track is added at this time to facilitate location of various areas of the weld, or changes in welding conditions. This information is put in by voice to allow addition of pertinent comments. The signal from either "live" weld or tape recorder is then amplified and is simulta- neously broken down into frequency and energy characteristics. Frequency characteristics are determined by (c) a set of 8 band pass filters which break the band of interest into equal percentage portions. Each portion is rectified, filtered, and the average peak value deter- mined. A packaged "data acquisition system" (Burr-Brown MP7208) is used to multiplex the analog signals, digitize, and feed them to the proces- sor system on command. 155 Energy characteristics of the burst are determined by a common acoustic emission technique called "Ring Down Counting". The signal bursts of interest are approximately exponentially damped sine waves. It can be shown mathematically 4 that if a threshold level is set, and if the number of cycles of an acoustic burst exceeding that threshold are counted, then the energy contained in the burst is approximately related to the count total by the following relationship. Log (energy In burst) = K- (total of counts) This method of analog to digital conversion is ideal for acoustic emis- sion bursts, because it tends to average out fluctuations due to extra- neous resonances, etc. The ring down counting equipment (d) includes appropriate amplifiers, LKHz-lMHz bandpass filters, and gain adjustments and consists of a modified Dunegan/Endevco (model 301) totalizer. Additional circuitry is employed to supply the totalized ring down count to the processor, along with information as to our currently used thres- holds for energy levels for purposes of display selection, (e.g., we may select all bursts exceeding the current low energy threshold) . The information is fed to the processor on command. Data Processing and Display After the data is preprocessed, it is presented to the microprocessor system. This system (Figure 3) does the following: 1. Accepts data in both compressed and multiplexed modes and processes it. 2. Continuously displays 7 types of processed incoming data in real time, if desired. 3. Allows changes of display type, scale, or grouping without dis- rupting the data acquisition. 4. Simultaneously stores incoming data to mass storage, if desired. 5. Recalls data from mass storage, if desired. 6. Prompts the operator in everyday English through the various data acquisition and display options available to him. In essence, the data processor used for acoustic emission analysis is programmed to operate as an on-line, operator-adjustable organizer of data derived from acoustic emission waveforms. At the heart of its operation is the Motorola M6800 microprocessing unit (MPU) . This MPU, which contains all the arithmetic and dec is ion -making capabilities of 156 the system, is completely contained in one 40-pin IC chip. It is this compactness that makes the microprocessor system so portable and so environmentally tolerant. The MPU responds to and controls the system by means of a 16- line address bus, an 8 -line data bus, and a number of control lines, represented in Figure 3 as one line. The set of instructions which the MPU uses to operate the system as a data organizer is stored in several Motorola M6810 random access memory units (RAM) . Other items stored in RAM are the English language text by which the system and the operator communicate, the processed data, the current display, and results of calculations directing the system through its operations. A key feature of this system for our application is the aforementioned English text the system uses to prompt the operator and the operator uses to alter the system's operation. This was implemented so that data analysis could be done by someone who had no understanding of computers. An example of a system operator exchange is shown in Figure 4. This "dialogue" is conducted using a CRT with keyboard or a TTY unit receiving and transmitting data to a Motorola M6850 asynchronous inter- face adaptor (ACIA) . This adaptor, and the Motorola M6820 peripheral interface adaptor (PIA) discussed below, are the keys to the micro- processor system's ability to adapt to the data formats desired by the operator. The physical structure of these devices is sufficiently undefined so that an appropriate command from the operator, through the microprocessor, can configure it to accept and/or transmit in a variety of formats. These devices are different in that the ACIA receives and sends serial data whereas the PIA receives and sends parallel data. The CRT-keyboard combination used in our system trans- mits and receives serial data in a preset format (RS232) . Thus, upon operation initiation, our program instructs the ACIA to operate in an RS232 format. The ACIA can then present keyboard text to or receive CRT text from the data bus, thus creating an operator -MPU information exchange . The preprocessed acoustic emission data are put on the data bus via two modes. The multiplexed frequency band data (A) is placed directly upon the bus upon command sent along a control line from the MPU. The ring- down count, threshold, and energy criteria data (B) are presented to the bus in parallel form through a peripheral interface adaptor (PIA) instructed by the. MPU to be a 16-bit data acceptor. In addition, this PIA is instructed to initiate, upon receipt of a control signal from the pre -processor, a high-priority temporary interruption of the MPU's normal display and dialogue routines so that the MPU can receive, process, and dispose of the incoming data. The way the MPU disposes of the data in our system illustrates the versatility of the PIA. 157 If desired by the operator, the processed data can be displayed as it is received, thus giving a real-time analysis of the current weld acoustic events. This display is accomplished by presenting display coordinate data to a PIA instructed to be a data transmitter (C) . The output of this PIA is put through a digital -to -analogue and operational amplifier circuit and presented, at the operator's selection, to either an oscilloscope for online previewing of event trends, or to an x-y plotter for the creation of hard copy of an analysis. Simultaneously, if so desired, unprocessed data can be stored in mass storage for further retrieval and analysis. This data is presented to a PIA instructed to operate both as an 8-bit data acceptor and an 8-bit data transmitter (D) . This PIA, by use of its four accessory control lines and an additional PIA for 12 -bit control line capacity, directs the traffic of digitized data to and from a floppy disc mass storage system. Thus, the versatility of our acoustic emission weld data analyzer is based on the versatility of the microprocessor and its family units. Some Typical Results To demonstrate the type of output the data processor gives, the results from three weld specimens are presented in Figures 5. The type of flaw induced in these welds was that of incomplete penetration. Dye pene- trant was used to provide immediate confirmation of these flaws. The processor generated histograms shown in Figure 5 are those of the number of acoustic events of a given ringdown count plotted on the vertical axis against the value of the ringdown count plotted on the horizontal axis. In effect, the analysis corresponds to that of an energy distribution. This method of analysis creates patterns that allow the eye to pick out subtle variations as well as continuous emission sources such as electronic or weld arc noise. Figure 5b shows the analysis display acquired during weld number 1, a weld in which incomplete penetration was very slight and intermittant. Very little acoustic emission activity is present with a widely scat- tered energy distribution. Incomplete penetration in weld number 2 (Figure 5c) was nearly 100%. Note the dramatic increase in activity. Weld number 3 (Figure 5a) was generated as a control having good pene- tration, and the analysis shows the least activity of all. Figure 6 shows variations in the frequency spectra of the above three welds. The differences in frequency characteristics are also obvious to the eye. We are presently at the stage where similar analyses are being done on a mass production basis to create a large acoustic emission data base. Summary To summarize, the use of microprocessors in general allows one to construct a dedicated processor according to his specific needs, take 158 that processor anywhere he goes, and use it with a minimum of turn- around time. In addition, programming a microprocessor amounts to the "wiring" of an electronic device dedicated to a desired process. In our specific application, we are able to use the processor to display distributions of ring-down count (energy) and frequency bands during a weld or the play-back of one. Through operator interaction, we are able to search for a promising flaw- candidate pattern and follow it through sequences of good and bad welds. By this means, we can not only find background noise and flaw patterns, but are able to ascertain what adjustments of monitor parameters are necessary to exclude or enhance these patterns. This will result in a better "signal-to-noise ratio" in the detection of weld flaws. In addition, it is our hope to be able to tune the monitor parameters finely enough to allow not just detection of flaws, but indications of flaw type and extent. In addition to the analyzer system that has been described above, GARD has constructed a new acoustic emission weld monitoring system for particular field applications which also contains its own microprocessor system. The development of this system followed exactly the same lines as the development of the data processor but, whereas the processor had been developed up to the software development stage, this system has been carried through to the stand alone stage. Besides allowing easy adjustment, through software, of monitoring parameters as determined by the analyzer system; the microprocessor also allows this system to per- form other functions in its idle time. The microprocessor determines, by means of difference of arrival time, the location of an AE source along a weld length; it controls the display of such information, and allows the system to run on totally unattended operation. Figure 7 shows the system. References 1. Jolly, W.D. (1968) "An In Situ Weld Defect Detector, Acoustic Emission", BNWL-817. 2. Prine, D.W. (Dec. 1976) "Inspection of Nuclear Power Component Welds, by In-Process Acoustic Emission Monitoring", NDT Interna- tional . 3. Prine, D.W. (Nov. 1974-Mar. 1977) "Inspection of Nuclear Reactor Welding by Acoustic Emission", (Data Report), U.S. Nuclear Regu- latory Commission NUREG-0035-1, NUREG-0035-2, NUREG-0035-3. 4. Qno, K. (Sept. 1974) "Acoustic Emission and Microscopic Deformation". 159 Figure 1. THE ACOUSTIC EMISSION DATA PROCESSING SYSTEM IN OPERATION 160 JQ O LU X O > az z o — u x a X en uj Xuiui oujui ai- — CC (J UJ — h- XXX ZK UJ (- uj uju s: OK I ~\ < uj O CO Q Z UJ O — ■V^ TTTTTTTT \ S o o _i CO C!J CO CO LU o o <3 S a) ® ^ 161 RINGDOWN COUNT LOW THRESHOLD EXCEEDED HIGH THRESHOLD EXCEEDED J ENERGY CRITERIA- MET ADDRESS DATA P!A y- ® PIA ACIA RAM MPU } PIA DAC CRT 8 KEYBOARD \ X-Y PLOTTER DISPLAY \ / OP AMP OSCILLOSCOPE DISPLAY } MASS STORAGE DEVICE Figure 3. DATA PROCESSING AND DISPLAY BLOCK DIAGRAM 162 READY FDR OPERATION: TO RUN KEY R TO RESTART KEY S . . . R DATA RECALL FROM FILE NUMBER: 01 STANDARD 1 0-16-21 DB 00500 NUMBER HISTOGRAM DISPLAY ON OSCILLOSCOPE 001 X SCALE FACTOR 001 LINES--CELL TO CHANGE STATE? KEY 'C ... C ERASE DISPLAY SUSPEND DATA INPUT MASK BAND CHANGE SCALE CHANGE CELL SIZE CHANGE DISPLAY MODULE CHANGE TYPE HISTOGRAM CHANGE FILE NUMBER CHANGE PROCEDURE CHANGE KEY E KEY M KEY B KEY •\- KEY c KEY D KEY H KEY F KEY P Figure 4. AN EXAMPLE OF SYSTEM-OPERATOR EXCHANGE 163 a) Control Weld 1000 J b) Slight LOP 1000 to ■M C > i- 3 Ring Down Count c) Severe LOP 1000 Figure 5. ACOUSTIC EMISSION ANALYSIS SYSTEM PLOTS - ENERGY DISTRIBUTION 161* o I— I X .1 MHZ a) Control Weld 1MHZ 1 MHZ b) Slight LOP 1MHZ CT> S_ c > •r— -t-> or *— ' L c) Severe LOP 1 MHZ Frequency 1MHZ Figure 6. ACOUSTIC EMISSION ANALYSIS SYSTEM PLOTS - FREQUENCY SPECTRA 165 o oc I— o 1-1 (/} Z OO O uj ^ o o Q CX. —I Q. UJ O 3 Q£ o o :e »— i 1/0 o oo o •-I CO uj a: CJ> UJ © 2 OO o s: < oo • <_5 0 •i- t_> ; :> : :;V"h ; " 166 PICTORIAL HISTORY OF THE DEVELOPMENT OF PROXIMITY PROBES FOR USE IN HIGH TEMPERATURE LIQUID METALS ENVIRONMENT Leo Hoogenboom Mechanical Technology Incorporated 968 Albany-Shaker Road Latham, New York 12110 Introduction In the development and condition monitoring of machinery, it is common practice to use any number from a large variety of sensors that pro- vide operational and/or monitoring information. Fortunately, in many applications the physical environment in which the sensors operate tolerates the use of materials of moderate physical properties with regard to temperature and contact exposure capability. However, often in those instances where the internal environment of machinery is hostile to many engineering materials, the need for data is most urgent. The designers of such machinery and the designers of sensors for use in the hostile environment present inside the machinery, face much the same problem. Their design freedom is severely limited by: 1. Lack of choice of materials, types of construction and fabrication methods. 2. Need for long life, due to difficult access and replacement. 3. Lack of information on long term physical properties of composite structures (e.g., electrical connections, metal-ceramic components). Nevertheless, as the need arose, such sensors as proximity probes, pres- sure transducers, accelerometers, strain gages, microphones and others were developed and used. Almost without exception, these were for a long time laboratory devices, due to the very limited demand for field installations. Of those sensors that have become available commercially very few if any are shelf items. The commercial quantities made of the various sensors are generally very limited. Overall quality ranges from poor to excellent. Prices are up to five times higher than those paid for comparable devices used in less severe environments. Continuity of availability of sensors from commercial sources has been good. When at Mechanical Technology Incorporated (MTI) development work on sodium lubricated bearings began, it turned out that there was no com- mercial source for suitable proximity probes or pressure probes. As a result, MTI developed those sensors for its own use as well as offering them commercially. For one of these sensors, the inductive proximity probe, the development is traced as it can be seen in photographs of 167 sections of actual probes. Before presenting the picture exhibit, the principle of operation of this class of probes is described and the major design parameters are identified. The inductive proximity probes are separated in two groups: 1. Eddy Current Proximity Probe . 2. Magnetic Reluctance Proximity Probe Eddy Current Proximity Probe The principle of the inductive proximity sensor (eddy-current type) is shown in Figure 1. This sensor consists of a coil powered with alter- nating current, resulting in an alternating EM field around the probe. If a conducting (nonmagnetic) surface is brought near the coil, the relationship between E and I changes. This change is the result of alternating currents (eddy currents) generated in the target by the external EM field of the probe. The eddy currents in turn generate an EM field which opposes that of the coil. The opposing effect of the eddy current field causes the total field inside the target to become weaker with increasing distance from the surface. The depth at which the intensity is about 60% down from that on the surface is defined as the depth of penetration. The effect of the target is to reduce the inductance (L) and to increase the AC series resistance (R) . If part of the gap is filled with a conducting medium, the field of the coil is attenuated in the same way as by the target. If the field attenuation in the partially filled gap is small (< 10%) , the probe can still function. This is of importance when using the probe in a corro- sive environment where hermetic sealing is necessary. For that appli- cation, the probe is mounted in a hermetic envelope, typically of aus- tenitic stainless steel. The thickness of the stainless steel in the gap must be such that no excessive attenuation of the field occurs. Thus, it is important to know what factors determine the depth of pene- tration. The depth of penetration is inversely proportional to the square root of the electrical conductivity (p), the magnetic permea- bility (y) of the target material and the frequency (f) of the AC field, as is also shown by the graph in Figure 3. At a given operating frequency, the depth of penetration can vary two decades between the nonmagnetic resistive materials and the magnetic materials. Figure 3 can be used as a guide in selecting a usable com- bination of target material, probe envelope material and operating frequency. Magnetic-Reluctance Proximity Probe The principle of this probe is illustrated in Figure 2. The field gen- erated by the alternating current through the coil causes magnetic flux *EM = Electro-Magnetic 168 in the ferromagnetic core. The flux density with a core is from two to three decades higher than without a core. The flux passes from the core through the gap back into the core. The air gap constitutes a major part of the magnetic reluctance of the probe. The typical increase in probe inductance when placing a ferro- magnetic target in contact with the probe is 50 - 100%. The effect of a nonmagnetic target on this probe (Figure 2) is similar to that on the probe in Figure 1. That is, a nonmagnetic target causes a decrease in inductance (L) and an increase in the AC series resistance of the probe with the core. when operated with a nonmagnetic target, the probe with core is an eddy current probe. When operated with a magnetic target, it is a magnetic-reluctance probe. Discussion of Section Photographs Liquid metal bearing research and development requires reliable measure- ment of shaft orbit. Without exception, the orbit measurement is made with noncontacting probes. All probes are of the inductive type, of two classes: 1. Magnetic Reluctance Probe 2. Eddy Current Probe The reluctance probe consists of a pot core, with a coil inside it. The eddy current probe consists of only a coil. The construction of both types of probes is a compromise between the need to electrically insulate the coil windings (and the entire coil) and the need to mechanically secure the coil to the housing either directly or first to the core and then to the housing. The need for a practical and controllable assembly sequence of the various parts, including making electrical connections to the mineral insulated lead, and sealing the entire envelope hermetically, leads to additional compromises in the entire design. With the foregoing as an introduction, a number of actual realizations of high temperature probe designs will be described briefly, using photographs of metallographic sections of transducers or assembly drawings identifying the various parts. The metallographic specimens are made by first vacuum impregnating the entire assembly, followed by a gradual removal of material by grinding in a surface grinder. When the desired detail is exposed, the section is lapped and polished and macro and micro photographs are taken. The sections selected here are most representative of the arrangements of probe components in the various designs. 169 Figure 4 shows a design developed by the U.K.A.E.A. in 1967 . It has the basic elements of the reluctance probe, coil and core. To assure that the core is maintained in a fixed position with respect to the housing, it has a sliding center part. This part is kept fixed forward axially by an austenitic steel pin (matching the housing in thermal expansion) . The core is carried through the face of the probe to elim- inate the effects of an austenitic steel cover. The lead terminations are brought out to the side , to be accessible during assembly. This probe was used in pairs, in push-pull, with a magnetic target ring and with a carrier frequency of 800 Hz. Push-pull operation assures linearity and temperature compensation. This design has given satis- factory service for many hours in the English bearing research programs. It has not been commercially available. At about the time that the design in Figure 4 was being developed, a very different design was made commercially. Samples of it are shown in Figures 5 and 6 and Figures 7 and 8. Both of these designs use a core composed of a bundle of ferromagnetic alloy wire. The wire passes through the center of the coil and is bent around in a U-shape to provide a return path for the flux on the out- side of the coil. Figure 5 shows a section on the probe center line. Figure 6 shows a section off the center line, where the outer part of the core is visible. Figure 7 shows a design similar to that in Figure 5, except for the use of a sturdy coil form. It is mounted in a holder that permits (later) installation of a sealing diaphragm. The designs in Figures 5 and 7 are for high temperature use, but are not hermetic. Details of the probe from Figure 7 are shown in Figure 8. The designs in Figures 6 and 8 were obtained commercially and were prototypes. Their performance was inadequate. The designs in Figures 6 and 8 were followed by the design in Figure 9. It has also a wire core. Two identical inductors are mounted in one housing. The second inductor is used to obtain temperature compensation, The function of the temperature compensating inductor is discussed in connection with the following design. The design in Figure 9 was a commercial prototype. Its performance was adequate, but the manu- facturer discontinued further development efforts. This led to the development of the inductive proximity transducer at MTI in 1969 to be * S. A. Dean, B. G. Ferguson, E. Harrison and A. Stead "A TRANSDUCER FOR MONITORING THE RADIAL CLEARANCE AT A SODIUM- LUBRICATED PUMP BEARING" United Kingdom Atomic Energy Authority The Reactor Group - 1967 - TRG Report 1472 (R) 170 used in sodium bearing research. A derivative of the probes shown in Figures 5-9 is shown in Figure 10. This probe has been made commercially for about 7 years by a succession of English companies, all under the direction of the originator of the basic probe design. It has been used successfully in substantial quantities by a large variety of users. The laminated core permits operation at a carrier frequency of as high as 50 kHz without serious degradation of the Q (Quality*) of the probe. The probe is being used in sodium and steam environments to a temperature of 1100°F (600°C). Overall, quality control and performance of the probes have been mod- erate to good over the years. No long term data has been published. An X-Ray of this probe is shown in Figure 11. It shows the form and the location of the lead terminations before they were disturbed by the lead end removal. The probe in Figure 10 operated In half-bridge configuration can be used to make a single-ended gap measurement with a dielectric medium in the gap. Temperature compensation is provided by the second inductor, which has a dummy target of the same material as that of the active (outside) target. If a conducting medium is present in the gap, and if a nonmagnetic target (and dummy target) is used, temperature com- pensation is degraded severely. No application of the latter type has been reported. A probe similar to that in Figure 10 with a single (active) inductor is shown in Figure 12. It is generally used in pairs in push-pull to obtain temperature compensation and linearity over a wide temperature range. The section in Figure 12 shows a cracked coil form. The cracking was probably caused by thermal stresses. The coil was wound wet (with application of liquid ceramic cement during winding) and became solid after firing of the ceramic cement. Subsequent axial ex- pansion of the wire caused the coil flange to crack away. The coils in the transducer in Figure 10 are not wet wound, and no cracking occurred . All designs shown thus far have a coil of a length roughly equal to their diameter placed in a magnetic core, resulting in a high inductance and a very directional and well defined external flux pattern. The design in Figures 13-14 is a radical departure from that concept. It only contains a very flat coil ("pan-cake" coil) and no core. The field of the coil reaches a distance of approximately one half coil diameter radially and axially (front and rear) . The design in Figure 13 shows a coil held in a dielectric mass. The dielectric mass is held in a metal case at a distance from the coil where the field is sufficiently weak. Around the coil no metal can be tolerated to within Ratio of inductive reactance to series resistance. 171 a distance equal to the coil radius, as otherwise the coil becomes inductively partially shorted and the probe is proportionately desen- sitized. The probe design shown in Figure 14 is widely used in many applications, It is operated with a carrier frequency of 1-3 MHz. The probe is not hermetic. It is used at low temperatures (<300°F). Similar probes are also made for high temperature service, without or with a hermetic envelope. When operated at a very much lower carrier frequency, and when held in a housing made of a resistive alloy, with only a thin face cover, the probe can be made hermetic. Under these circumstances, sufficient flux remains to permit sensing of a metal object outside the probe housing. However, the absence of a core, and in its place the presence of a conductive housing degrades the signal generating capability of the probe somewhat, and hence increases the chance of drift and noise. Also, the small section of the coil necessitates the use of very fine wire to obtain sufficient turns (and hence inductance) to off-set the inductance loss due to the absence of a core. To circumvent the problems of stresses due to differential expansion, and to locate the coil windings firmly in the transducer, the ceramic coil form is sometimes replaced with a metallic coil form. The metal form is coated with a ceramic coating by flame spraying to provide the necessary insulation to ground for the coil wire. The metal coil form causes additional losses that further reduce the probe's signal to noise ratio. Summary Development of liquid metals lubricated bearings created a need for proximity probes capable of measuring bearing orbits in sodium at high temperature. In a nearly parallel effort, such probes were developed by U.K.A.E.A. and by the AEC at MTI . Both probes became commercially available. The U.K.A.E.A. probe has led to the current design as shown in Figures 10 and 12. The MTI probe design has remained the same, except for minor changes in lead connections made in 1976 and a different choice of coil wire alloy. The probe sections shown illustrate the many design concepts that were tried to satisfy the functional requirements of the measurement. When comparing the number of components used in the various designs with core, and the necessary assembly operations and sequence of operations, the MTI design appears superior to any of the preceding or parallel designs. The present state of proximity probe technology permits the fabrication of reliable high temperature hermetic devices suitable for long term monitoring of machinery in a sodium environment. 172 t EDDY CURRENT < / < s < < s CONDUCTING SURFACE (TARGET) C ' ' ' ' ' c \ MAGNETIC FIELD COIL I = iSINUttf) E ■ tSIN(uif) PROBE CONSTRUCTION t GAP TYPICAL PROBE RESPONSE Fig. 1 - Inductive Proximity Sensor (eddy current type) 173 MAGNETIC SURFACE \\\V\\\\\\V \ \ V \N\^ \\\\\\\\^\\^\\ N \^\ S >— > \ \ N X K X X \->.\^ \ (TARGET) ~GAP SOFT MAGNETIC CORE COIL I = i sin (wt+ 9) E s e sin (wt) PROBE CONSTRUCTION A R TYPICAL GAP PROBE RESPONSE Fig. 2 - Inductive Proximity Sensor (magnetic reluctance type) 17^ i 8 3 S So S3H0NI 'NOIlVUiaiGd c o •H •u rt Vj 4-1 01 s ID FM T3 rH 0) •H pt, ^ U O ■H a 4-1 QJ QJ 3 (3 cr 00 o> 03 V-i 6 Fn o S-i w 4-J 3 O w CO r4 rH (U fxj > CO . GO •H fe 175 a ui woo -J a z ® a < *t D _) U t/> u. fli ^-\ cu Fn m 3 o 3 PM O en m nJ 6 r- CU d ^ s T3 U o O o 4-1 c/> o O TJ rH fi o CU 01 CU r^ Ph u O O 2 u rH 4-> Pu Q) cO > u r^ cu cu ■M o c> •H s G CU ■H •H H X O u <1J ° 2 Pm CD rH d CO -H CO CO CO S f a •H CU rH at rJ pq s_> ^ z 3 3 a X 4 177 T3 CU & PU nj U toO O •u o A Ph s — ' cu M O u /— \ m (1) V-i cu ■H u IS 3 00 rC •H ■U fa •rl £ PI ■rl 3 o ■H u cd Xi •H ■u 3 PI o (30 •H c 4J •H O 42 CU o w CO cu CO Pi •H 3 cu 1 o M-l M-l IW CU o pq ^o 00 •rl fa 178 u o o a) u •H /-»% JS | •u ^ •H > JS a o u •H CO 4J a) o 00 60 •H Pn Fig. 9 Section through Probe with Wire Core, with Built-in Temperature Compensating Inductor. Probe Diameter 1.0 inch (25mm) 181 o CO f~^- a M o Ptf H PM H CO O W f) a a S M o Q CJ ^ 00 -U t3 O -H d rH g CO d T3 T3 . CD rH CO 00 -H M H 3 til d pq & pq cd CD CO CD 5-i B d M to co o a co m CD CD PM CD u CD > o CO S „ d cd U 14 ftj CD 4-1 CD d co H H 182 T3 rt a) O X! £ 4J a) (^ 14-1 o CD CO >. S CO Pi T3 1 C X W 60 •H Pn 183 Q) *-N 43 s •U £ CTi 4-) rH d *■_. ' o 43 43 4-t CJ •H 13 S •rl A U~) o 1-^ rH • 01 CO M •H d bO >-i •H 0) Pn 4J CU C •rl CC •rl •U Q ct) 43 cu 4J 43 O O U 4J P4 H 01 • H !-i •rl O ■U ■H o CO d T3 tu (3 43 H O M C»0 Fm (3 •H >. 4-1 ■P ct) •H CO a (3 •H CU X & O S u o Pm u CN ^H • tot) •H Pn l8if 0) x> o u PL, ^ 4-1 ■H •H X O U Ph •-N ■U a fl o OJ •H n 4-) M CO 3 U CJ QJ 1 a >>o "3 T3 > ■u o nj c M 0) aj 3 fX cr 6 cu O PU M-l O •H a bO •H Pn 186 C7\ CTi CD Pu O H 0) > cu Q a) o >> •H s •H X o S-i 3 4-1 CO cu a, 6 cu H 42 C30 •H w ■H 187 SESSION I V RAILROAD SYSTEM DIAGNOSTICS CHAIRMAN: J . L . FRAREY SHAKER RESEARCH CORPORATION DEPARTMENT OF TRANSPORTATION SYSTEM FOR TRAIN ACCIDENT REDUCTION (DOT- STAR) James K. O'Steen Naval Surface Weapons Center White Oak Laboratory Silver Spring, MD 20910 In the course of 30 billion miles traveled each year, the two million American railroad freight cars suffer thousands of derailment accidents, 5602 in 1970. These derailments are due to numerous causes, but historically approximately half are due to failures of equipment or roadway. The cost of these derailments is not accurately known. But estimates for equipment and roadway- caused accidents total approximately 60-65 million dollars for direct railroad losses, with total losses estimated at 180-330 million dollars per year. The largest single category of equipment -caused derailments are due to broken or overheated journal bearings (hotboxes) y which accounted for 409 derailments and 45-75 million dollars in estimated total damages. Other major equipment -caused derailment sources are broken truck compo- nents and car dynamics. Similarly, the majority of roadway -caused derailments are attributable to broken rails, joints, or switches and to improper maintenance of the roadway. The Naval Surface Weapons Center (NSWC) became involved in this railroad problem as a part of the Navy's Technology Transfer Program. The Federal Railroad Administration funded NSWC to investigate possible solutions to their problem, due to NSWC's experience in the development of high production, low cost, and high reliability sensors. Considering the large amount of damage attributable to equipment and roadway-caused derailments and the relatively limited number of major causes, it is intuitive that an economically feasible solution to this problem exists. The three most viable solutions to the prevention and/ or reduction of damage from derailment accidents for the existing fleet are: preventative inspections and maintenance; wayside monitor systems; and on-train monitor systems. Each of these methods provides a cost-effective solution to part of the derailment problem. However, the most functional and cost-effective solution to the total problem would be a nation-wide system using all three methods in conjunction. The on-train monitor system is perhaps the least researched and definitely the least deployed method of derailment prevention. On-train systems offer substantial advantages in the detection of failure modes that are rapid and offer little advance warning. Monitoring of poten- tially hazardous failure modes during every second of a vehicle's life is the major advantage of on-train systems. However, providing an economically viable system for railway freight cars is a substantial challenge. Objectives for such an on-train monitor system would include: 191 o low cost o very low false alarm rate o very low maintenance o self-sufficient o monitors multiple accident causes Such an on-train monitor system is the Department of Transportation System for Train Accident Reduction, DOT-STAR. Preliminary studies for the DOT-STAR System evaluated the equipment and roadway-related derailment causes that were monitorable by on-train systems. The most economically viable equipment-related causes were journal and roller bearing hotboxes . In both cases derailments are pre- vented by an on-train monitor system. Prevention of roadway-related derailments is not economically viable with an on-train system. However, it was determined that local derailments could probably be detected by an on-train system which could in turn prevent a general derailment (train wreck) . In a portion of derailments, an initial local derailment of one wheel, one axle, or an entire truck will take place and the train will proceed some distance without further disturbance. No one will be aware of the local derailment until an obstruction or further damage to the derailed equipment leads to a general derailment, the magnitude of which is directly related to train speed. The number of such two-stage derailments is not readily determined from FRA accident data. However, review of all National Transportation Safety Board derailment accident reports between 1969 and 1976 indicate that up to 30 per cent of all derailments are of the two- stage type. Therefore, an effective on-train local derailment detector could substan- tially reduce derailment damage. The objectives of the DOT-STAR System are three-fold. First, the system is to detect incipient or actual derailments and provide for action to prevent or mitigate resulting damage. Second, the system must provide a minimum life cycle cost, require minimum maintenance, and be compatible with railroad industry operating and maintenance proced- ures. Third, the system must be capable of expanding to include additional hazards such as fire and rock-off. The DOT-STAR System is an on-train safety system that monitors and automatically activates the train's brake system upon the detection of a wheel derailment or excessive axle bearing temperatures. The braking action stops the train before either an axle bearing failure develops into a derailment or a minor derailment develops into a major wreck. Currently, the basic system consists of continuously-monitoring self -powered thermal and derailment sensors connected via a wiring net- work to an electro-explosive valve attached to the car brake line (figure 1) . The activation of any sensor produces a pulse of electrical energy that in turn fires the explosive valve, thereby venting the brake line and activating the braking system. The on-train location allows for rapid and automatic response to derailments caused by failures of either equipment or roadway. The system on each car is independent of the other cars on the train; this makes the system compatible with present railroad industry operating 192 procedures . The system is designed with low cost, long service life, maintenance free components to provide both minimum initial and life cycle costs. The sensor assembly (sensors and integral power supplies) and the electro-explosive brake venting valve are both one-shot maintenance- free devices (figures 2,3, and 4). The wiring network is rugged, shielded, and does not affect normal car or truck maintenance. Expansion of the basic system is easily provided by coupling of additional sensors (fire, rock-off, etc.) into the wiring network. The DOT-STAR thermal sensor is designed to continuously monitor the temperature of either journal or roller bearing assemblies. At or below the maximum normal bearing temperatures, the sensor remains inert. If a bearing assembly rises to a temperature above this normal maximum, the sensor fires and transmits an electrical signal that activates the train braking system. Functionally, the sensor consists of three parts: a spring-loaded "cocked" firing pin held in the ready position by a "C" configured release pin and a thermal battery power supply (figure 5) . The heart of the sensor, the "C" configured release ring, is made of an NSWC-developed alloy called NITINOL. This nickel-titanium alloy has a memory property such that at a given temperature NITINOL objects can be restored to their original shape even after being "permanently" deformed out of shape. When this closed "C" release ring is heated to its thermal transition temperature by an overheating bearing, this release ring springs to an open "C" configuration, releasing the firing pin which in turn activates the thermal battery power supply. The continuing development of the DOT-STAR thermal sensor was broken into two sequential efforts: journal and roller bearing applications. The journal bearing work has been completed through a full field demon- stration. This demonstration consisted of two phases. The first phase was the successful demonstration of hotbox detection and train brake application on two intentionally-caused hotboxes at a train speed of 35 mph (figures 6 and 7). The second phase, which is in progress, is designed to demonstrate a low false alarm rate. To date 32 sensors have each seen greater than 40,000 miles of actual service without a single false alarm. The roller bearing application work is presently preparing for a similar demonstration test with the expectation of similar results. The method used to attack both problem applications is identical. It consists of finite element computer simulations of bearing heat gener- ation and conduction (figures 8 through 14) . These are used to predict normal over-the-road temperatures, thermal gradients through structural members, and temperature rise-time characteristics during hotbox condi- tions. These data were used in conjunction with actual measured over- the-road and hotbox temperatures to optimize the thermal sensor design and sensor contact location. Most of the work on derailment sensors to date has involved study- ing both normal over-the-road and derailment environments. Two possible detection techniques have been determined: direct rail contact and seismic detection. In order to eventually compare the two techniques, substantial seismic over-the-road and derailment data have been collected, 193 From this data, we have found that under extreme winter frozen-ground conditions, over-the-road environments such as rough track, rough switches, etc. often exceed derailment conditions during the summer (figures 15 and 16). This is true for both acceleration and velocity. Shock spectra analysis has also revealed no significant differences between these two environments (figures 17 and 18) . Discrimination between the two environments will require filtering and integration of the velocity signature. NSWC is currently exploring this approach. 19^ PISTON ACTUATOR VALVE figure 1: D.O.T. -System for Train Accident Reduction DBMKMENr sensor WLSSWf figure 2: Sensor Assembly 195 SEMSO/t S/OJS/4/6 HOUSING sea/so/? Assembly o^journal housing AT>APt€R AXLE ROILE.R. &EARIM6 THERMAL n/ SENSOR DEQAiLfVieMT SEW&OR figure 3: Sensor Assembly Mounting Locations 196 TRAIN AIR EXPLOSIVE ELEMENT CAP (VENT TO ATMOSPHERE) PISTON figure 4: Electro-explosive Valve 1.0 INCH DIAMETER POWER SUPPLY (THERMAL BATTERY) FIRING PIN NITINOL ALLOY (250°F) RELEASE RING ALUMINUM CUP SEQUENCE OF OPERATION 1. HOT BEARING DEVELOPS 2. NITINOL SENSES 250°F 3. FIRING PIN IS RELEASED 4. STAB PRIMER ACTUATES THERMAL BATTERY 5. BATTERY SUPPLIES 6.2 VOLTS TO BRAKE ACTUATOR HOT BEARING figure 5: D.O.T.-S.T.A.R. Thermal Sensor 197 figure 6: Field Test Demonstration of D.O.T.-S.T.A.R. System. Journal Bearing Hotbox Detected and Train Stopped. Right Side. L. BFAk'tlt- , , (feo> ■4d v - A. ■30 % - <*. ~T/ms- (m/ Mures.} figure 7: Field Test Demonstration of D.O.T.-S.T.A.R. System. Journal Bearing Hotbox Detected and Train Stopped. Left Side. 198 -o rt O O, T3 1 PS t3 i CD H 4-> CD CD O > M •H O T3 nd CD CD Jh 3= P. rt 0-, U PS •H

o u 3 So (J«; 3^nj.V513ciW3X <+H CD O P •H •H P s O •H 3 tu "0 O • W> p w C Oh •H •H Cfl P ■M X 03 03 i-H 0) CD 03 oa X s i-H P 03 o c p <+n o R •H 3 4-> P o CD o ^ 2 3 T3 i-H P C 03 C o o CD u •H a ^ CD i— 1 T3 H w 03 oo CD P 3 M m 199 I . BeAK//n- ^ A. -SO ft ! figure 10: Measured Empty Car Over-the- Road Journal Bearing Temperatures . 1$ 7~/Aig- (M/Hvraj) figure 11 : Measured Full Car Over-the- Road Journal Bearing Temperatures . 200 ' z? /■///.///• 14- tmtt 1! VUU i Mil \\ figure 12: Typical Roller Bearing Finite Element Net for Heat Production and Conduction Analysis. Break for Photographic Coverage; Several PaBses Over Shims with Frequent Stops THERMOCOUPLE LOCATIONS 5 Adapter Underside Under Sensor Contact Point 6 Center of Adapter Underside 8 Adapter at Thermal Sensor Contact Point 100 120 160 TIME IN MINUTES figure 13: Measured Full Car Roller Bearing Operating Temperatures at 45 MPH, 90 Ambient Temperature . 201 /&oo\ /loo - noo - 9oo too T^ov-ue* Be^RiNt Hot Box ?ft£OlC.TED TeMPeRftToRe. SfclSS C6NT6R fy/9) Axue 4. 100 %C0 300 5tX} 600 700 8 9ao Time (sic; figure 14: Typical Predicted Roller Bearing Hotbox Temperature Rise Characteristics 202 PhILPOhD OEPAILMENT STUDY lfi D.M.I R' EMPTY DERAILMENT 2 CH 6 FOWflRO t+tEELS ZE ___ 8 199 242 i, i ^ It MA An ^B^^r NOiLitriMu [> 1 1 5" i . — i £ ■j 4 ■ m 7- 5 — ' TIME SECONDS figure 15: Typical Side Frame Accelerations During an Empty Car Derailment. DERAILMENT STUDY DMIR UINTER TRACK 112 CM •: 599 HZ TWO HmRBOURS TO MINNTAC POST 43 7 17 19 6 re i N 1 2 1 4 3 & f 8 ' "5 ' TIME SECONDS figure 16: Typical Side Frame Accelerations Traversing Rough Track. 203 PERftlLWEHT STUDY DWIR HtNTCT TBWCK CH 6 See HZ TWO HARBOURS TO MNNTAC POST 43 figure 17: Shock Spectrum Frequency Analysis of Typical Side Frame Accelerations Traversing Rough Track. K , RflllROtt) PERfflLHEHT 8TU0V • IV EMPTY DERAILMENT 2 CM 6 FOHflRD WHEELS figure 18: Shock Spectrum Frequency Analysis of Typical Side Frame Accelerations During an Empty Car Derailment. 20^ COMPARISON OF VIBRATION ANALYSIS TECHNIQUES FOR RAILROAD ROLLER BEARING DIAGNOSTICS Warren D. Waldron Shaker Research Corporation Northway 10 Executive Park Ballston Lake, New York 12019 This paper summarizes recent work at Shaker Research Corporation involving rallcar roller bearing defect diagnostics utilizing high frequency vibration techniques. The general objectives of the work are summarized in Figure 1. Although both objectives are addressed to the same component - railcar roller bearings, and both employ the same basic raw data - high fre- quency vibration, they have vastly differing requirements. A track side or wheel shop device requires ruggedness, simplicity (minimal operator training) , and a sensitivity high enough to identify defects that are easily recognizable to the naked eye. A device used for cer- tification testing, on the other hand, must be able to discern minute defects and can utilize more sophisticated techniques as it would be inherently a laboratory device. The three basic levels of sophistica- tion used in analyzing the raw accelerometer data are summarized in Figure 2. Three test vehicles or rigs were employed to evaluate the high frequency vibration approach using bearings with varying types and degrees of sur- face defects. The test rigs are shown in Figures 3 through 5. Four separate types of diagnostic system tests were performed (using the three test rigs) . The primary diagnostic system objectives and conclusions for each of the tests are summarized in Figures 6 through 9. In addition to diagnostic system evaluation, the failure progression and certification tests were designed to study bearing failure mechanisms and times to failure of various defect modes. Figures 10 through 14 summarize the results of each test in terms of bearing condition versus peak amplitude and, in the case of the certifi- cation tests (Figure 14) , the signal to noise ratio of the discrete defect frequency contained in the demodulated signal. Specific conclu- sions with regard to each test were given in Figures 6 through 9. In general it is seen that: 1. If "peak" amplitudes exceed two "g's", the bearing was unquestionably defective. 2. Below two g's, amplitude level was an unreliable indicator of bearing quality. 205 3. Using a spectrum analyzer to analyze the low frequency (0-200 Hz) demodulated signal, made it possible to rate bearings in the region below two g's. For the wheel shop tests a modified crest factor approach was also used as a bearing condition rating factor. In this case, approximately five seconds of useful magnetic tape recorded raw acceleration data for each of the 90 bearings tested was available (useful in the fact that both load and speed were constant). For each bearing a modified crest factor was created by manually dividing the peak by the average of the demodu- lated signal at the same "instant" in time. The results of this approach is shown in Figure 15. The plot shows the percentage of the total number of good bearings that would be rejected, the number of defective (yet not condemnable by AAR rules) that would be rejected, and the number of con- demnable bearings that would be accepted versus detection level (modified crest factor). For example, if the detection (rejection) level was arbitrarily set at 1.0, approximately 20% of the good bearings and 507 o of the defective bearings would have been rejected, and 20% of the con- demnable bearings would have been accepted. Thus, it is seen that the addition of a rather simple data processing technique (crest factor) greatly improves the discriminating power of the high frequency variation diagnostic approach. By performing the crest factor function electronically and improving the drive system to attain better speed consistency, the discriminating power should be greatly enhanced. General conclusions with regard to the feasibility of high frequency as a railcar roller bearing diagnostic tool are summarized in Figure 16. 206 GENERAL OBJECTIVES OF WORK EVALUATE FEASIBILITY OF DIAGNOSTIC SYSTEMS o FOR DEPLOYMENT AT A WHEEL SHOP OR AT TRACK SIDE o FOR USE DURING CERTIFICATION TESTS TO INDICATE THE "INITIAL" OCCURRENCE OF A SURFACE DEFECT. Figure 1 207 HIGH FREQUENCY (OVER 10 KHZ) TECHNIQUES UTILIZED o PEAK ACCELERATION AMPLITUDE J+H N\M- HMAa o PEAK/AVERAGE AMPLITUDE OF DEMODULATED HIGH FREQUENCY SIGNAL *: P A o FREQUENCY DOMAIN ANALYSIS OF DEMODULATED SIGNAL IN THE RANGE OF ROLLER PASS FREQUENCIES (0-200 HZ) Figure 2 208 o ROTATING OUTER RACE o MODERATE AXIAL LOAD o VARIABLE SPEED o UNGREASED BEARINGS Figure 3 209 FAILURE PROGRESSION AND CERTIFICATION TESTS o ROTATING INNER RACE o HIGH SPEED (65 MPH) o USED TO EVALUATE BEARINGS WITH MULTIPLE DEFECTS Figure 4 210 WHEEL SHOP TESTS o ROTATING OUTER RACE o MODERATE SPEED (45 MPH) o USED TO EVALUATE A "RANDOM" SAMPLE OF BEARINGS OF UNKNOWN QUALITY Figure 5 211 OBJECTIVES AND CONCLUSIONS - COMPONENT TESTS ESTABLISH FEASIBILITY o AMPLITUDE @ 10-40 KHZ GOOD INDICATOR OF GROSSLY DEFECTIVE BEARINGS o POOR DISCRIMINATION BETWEEN NEW, GOOD, AND "MARGINAL" BEARINGS EXAMINE ROTATIONAL SPEED EFFECT o AMPLITUDE DECREASES MARKEDLY BELOW 30 MPH DEFINE SYSTEM CONCEPT o H.F. (10-40 KHZ) PEAK AMPLITUDE APPROACH RECOMMENDED Figure 6 212 OBJECTIVES AND CONCLUSIONS - FAILURE PROGRESSION TESTS MONITOR DEFECTIVE BEARING DEGRADATION o DEGRADATION EVIDENT AND CONSISTENT WITH INCREASING AMPLITUDES FURTHER QUANTIFY AMPLITUDE/DEFECT SIZE RELATIONSHIP o AMPLITUDE PER SE NOT ALWAYS INDICATIVE OF DEGREE OF DAMAGE EVALUATE EFFECT OF GREASE o UNCHANNELED GREASE ATTENUATES AMPLITUDES BY 2 TO 3 o 30 TO 50 HOURS REQUIRED TO STABILIZE (VIBRATION AND TEMPERATURE) A NEWLY GREASED BEARING EXAMINE RADIAL LOAD EFFECT o LOAD HAS MINIMAL EFFECT ON VIBRATION AMPLITUDES Figure 7 213 OBJECTIVES AND CONCLUSIONS - WHEEL SHOP TESTS ESTABLISH FEASIBILITY OF TESTING WHILE INSTALLED ON AXLE o SYSTEM CLEARLY ADEQUATE FOR IDENTIFYING "NON-REPAIRABLE" BEARINGS IDENTIFY LIMITATIONS OF RECOMMENDED APPROACH o ADDED SOPHISTICATION NEEDED TO SEPARATE DEFECTIVE FROM GOOD BEARINGS o MECHANICAL APPARATUS NEEDS TO BE MORE POWERFUL AND MASSIVE EVALUATE ATTENUATION EFFECT OF WHEEL SET INERTIA o AXLE MOUNTED ACCELEROMETER ADEQUATE Figure 8 2lU OBJECTIVES AND CONCLUSIONS - CERTIFICATION TESTS EVALUATE POTENTIAL APPROACHES TO IDENTIFY "FIRST" SIGNS OF DEFECT o CREATION OF A MINUTE DEFECT WAS REPEATABLY IDENTIFIED USING SPECTRUM ANALYSIS (VERY NARROW FILTERING) OF AN ENVELOPE DETECTED HIGH FREQUENCY ACCELERATION SIGNATURE o HIGH PASS (OVER TO KHZ) OR BROAD BAND (10 to 40 KHZ) FILTERING OF THE RAW SIGNATURE PRIOR TO ENVELOPE DETECTION PROVIDED MORE CONSISTENT RESULTS THAN NARROW BAND FILTERING Figure 9 215 o rH 0) M •H Pn zo < 3 216 ^ si O o I- g z o O o z tO O n ii ii o o z x m + it o o CT> D o z u o o u o o z < z _ 1 mJ — o z ts. V 217 CM S- •r- 218 CO CD S- CD 219 3 220 RESULTS OF TESTS AT SOUTHERN RAILROAD WHEEL SHOP SHOWWG TYPICAL DEFECTS Figure 15 221 SUMMARY OF CONCLUSIONS o HIGH FREQUENCY PEAK AMPLITUDE DETECTION NEAR 2 "g" LEVEL SIMPLE AND INEXPENSIVE METHOD OF WEEDING OUT BEARINGS THAT SHOULD BE REMOVED FROM THE POPULATION o SOME FORM OF CREST FACTOR APPROACH AFTER DEMODULATION APPEARS PROMISING TO BETTER DISTINQUISH GOOD FROM MARGINAL BEARINGS IF BEARING DIAGNOSTIC SYSTEMS ARE TO BE SUCCESSFULLY INTEGRATED INTO THE R.R. SYSTEM, AAR REMOVAL AND CONDEMNING LIMIT RULES WILL HAVE TO BE RE-EVALUATED WITH REGARD TO: — - FREQUENCY AND PLACE OF INSPECTION — QUANTIFICATION OF DEFECT (EYE VERSUS AMPLITUDE) Figure 16 222 SESSION V LAND VEHICLE DIAGNOSTICS CHAIRMAN: J . A . GEORGE PARKS COLLEGE OF ST. LOOIS UNIVERSITY MAINTENANCE MANAGEMENT THROUGH DIAGNOSIS Richard G. Salter The Rand Corporation 1700 Main Street Santa Monica, California 90406 MILITARY LAND VEHICLE MAINTENANCE The high cost of military vehicle maintenance has caused an increas- ing interest in the development of improved vehicle designs and diagnos- tic aids and procedures, with the objective of reducing costs. The Defense Advanced Research Projects Agency (DARPA) , Vehicle Maintenance Program is currently addressing the question of whether advanced tech- nology can contribute the opportunity for reducing these costs. Current Army vehicle maintenance procedures include a scheduled maintenance program consisting of preoperation checks and periodic ser- vicing. Also, corrective vehicle maintenance is performed in response to overt failures. Experience in the realm of stationary equipment main- tenance suggests the need for an added component in the Army's operational scheme — periodic health screening. HEALTH SCREENING There are a number of benefits attendant to the ability to anticipate the need for maintenance rather than just reacting to the occurrence of failure. Failures tend to occur in a sequence or cascade of events, with each stage increasing the probability of the catastrophic event. The ultimate occurrence of this event can also precipitate secondary failures. If the sequence of events, or a significant decay function within that sequence can be detected in the early stages before failure has occurred, a systematically planned intervention can preclude the failure occurrence. The process, termed here health screening, can also provide the basis for reducing the scheduled maintenance program where screening can detect the need, and the lack of need, for servicing. A final benefit is the capability for a more reliable determination of the probability of a given vehicle successfully completing a future mission increment, leading in turn to a quantifiable fleet readiness prognosis. Necessary to the implementation of a comprehensive health screening program is an understanding of the significant generic failure models for each vehicle type in the fleet. These models encompass the significant wear and condition decay functions as well as the failure cascades in Section titles in this paper are keyed to the briefing charts attached. 225 which these functions play a role. Maintenance decision criteria can be derived from both the absolute level of the decay phenomenon and the rate at which decay is progressing. A number of test equipments types can be employed to determine the progression of the decay phenomenon. Standard shop test equipment (TMDE) can be employed but requires a skilled mechanic and a shop setting. A significant hookup effort and the need for experience in interpretation of results has resulted in a reticence of Army mechanics to employ this equipment, even in the fault isolation function. The Army is currently developing an improved unit known as STE-ICE. This equipment is por- table and can be used with a diagnostic connector if the vehicle is so equipped, or with a transducer harness which improves the hookup problem over the TMDE. STE-ICE provides the mechanic with testing logic but not with diagnostic interpretation. The Army's ATE-ICE equipment, now in the experimental stage, adds the capability for interpretive logic. A limiting feature of this technique, however, is that faults must be cor- rected as they are discovered or the subsequent decisions by the device may be in error. This limits the capability to perform a complete assess- ment before repair is initiated. Dynamometers allow loading of the vehicle throughout the complete dynamic range of the vehicle and in theory make possible a more complete determination of vehicle condition. Another approach to dynamic eval- uation would involve the miniaturization of the test equipment so that tests may be conducted on the road, thus eliminating the high costs of stationary energy absorption equipment. Such onboard equipments may be either connected just for the period of the test (strap-on) or may ul- timately be built into the vehicle. A final technique for screening tests is the detection and analysis of selected vehicle radiated signatures. The current state-of-the-art, however, suggests that this technique is only a long-term possibility. The radiated signal/noise characteristics from complex vehicle systems as their condition varies is poorly understood at this time. Given that vehicle signatures can be remotely detected, that condition/diagnostic signals can be discriminated and that machine intelligence can yield com- prehensive assessment of vehicle condition, the technique may be con- sidered a candidate ultimate system. Difficulties, however, beyond the problem of the capture and interpretation of signals from a complex vehicle may include the costs and logistics of the necessary equipment, particularly for a battlefield environment. FAILURE MODELS Actual vehicle failures result from a cascade or sequence of events or circumstances which ultimately lead to a catastrophic event. An ex- ample of such a sequence model may be developed around the cylinder failure associated with overtemperature operation. Basic causes can be of several types. Coolant loss may result from afterboil due to short cooldown procedures or from a system leak. Inadequate frequency of pre- op inspection would allow this loss to progress. Or, reduced coolant circulation could result from a clogged radiator on the coolant side, or 226 faulty pump operation due to a loose belt. Or, debris in the radiator on the air side could cause reduced heat rejection. Each of these cir- cumstances, or combinations of them, could cause an increasing operating temperature trend above the thermostat control point. The shape of this trend is important to its detectability . If this trend is undetected due to instrumentation resolution and/or infrequent observation, it can progress to the point where a severe overtemperature condition (boil over) results. If this condition is not observed in time, catastrophic cylinder failure results . FAILURE/DECAY FUNCTIONS Ideally, a time decay function, to be useful in a screening program, should be a graceful monotonically varying curve with a first derivative that increases in a continuing progression. If the derivative changes very slowly and then suddenly increases, precipitously, the short time available to be alerted to the change will preclude intermittent sam- pling and will require continuous monitoring of the function, with an alarm to signal the "break point." In this case only built-in test equipment can be considered competent to detect the phenomenon since off-board periodic inspection cannot practically inspect with the neces- sary frequency. If significant condition phenomena exist, however, which exhibit graceful decay functions, intermittent screen testing can be viable. IMPLEMENTATION PROBLEMS The difficulty, however, is that field failure model data for Army vehicles are woefully inadequate. The current understanding of those models is manifest in the design and qualification test criteria cur- rently in use. It is apparent from experience with current vehicles which survive their qualification tests but don't perform to these levels when deployed in the field that the field deployment failure models are poorly understood. Certain vehicles, for example, require up to five (5) engines when deployed in the field to complete the qualification- endurance time requirement. The lack of understanding of these models means that we are ignorant of the prospects for a necessary set of graceful decay functions, and therefore a means to determine the viabil- ity of screen testing as a means to reduce maintenance costs. Another inadequate resource, necessary to determining this viabil- ity, is consistent and reliable maintenance costs data for the existing fleet. Without such data it is impossible to make meaningful cost- effectiveness comparisons of candidate system alternatives with current practice. A principal difficulty, as usual, is the reliability of data 227 source documentation, VEHICLE MONITOR SYSTEM (VMS) A key component of the DARPA Vehicle Maintenance program is the VMS Demonstration project. The demonstration objectives include de~ termination of the capability of current and near- term technology to provide effective vehicle-borne instrumentation technology, and to demonstrate the effective use of such a device in the Army field en- vironment. In connection with this latter objective, we expect the pilot data developed during the demonstration to provide significant insights for the recommendation of future productive courses of action. Participants in the demonstration in addition to the DARPA Tactical Technology Office (TTO) are The Rand Corporation (program coordination, test design/supporting analysis), the U.S. Army, TARADCOM (contract administration, test planning and logistics) and RCA, Government and Commercial Systems Division (hardware/ firmware development).* The cur- rent program schedule includes engineering/vehicle tests at TECOM begin- ning in FY 1978 and the field demonstration test beginning in FY 1979. The results of the demonstration should allow improvements to vehicle design and qualification criteria and a refinement in the understanding of cost effective future Army maintenance equipments and procedures. VMS DATA DEVELOPMENT The VMS is conceived to provide data from the actual field opera- ting environment on deployed vehicles. The types of data to be devel- oped are vehicle use patterns, vehicle condition progressions, and main- tenance action patterns. Use pattern and condition data are derived from instrumented sensing of vehicle variables in a time context. Main- tenance actions are derived, primarily from inputs by maintenance per- sonnel, with attendant but unavoidable reliability questions. In addition to developing a general understanding of the distri- bution of vehicle employment patterns, failure and decay correlations will be developed. Short-term failure processes will be developed from high resolution, real time data displays** showing causative stress ex- cursions among concurrent data variations. Long-term failure correla- tions will be derived primarily from cumulative stress prediction vari- ables. Analytical techniques to be used in developing these correlations will include LOGIT analyses (employing logistic progression failure probability model) and exponential failure probability regressions. Bill Stewich of RCA will describe this development in the follow- ing paper. In general, data is compressed for storage by time stamp recording the event of threshold crossings. Each data domain is partitioned into bands of appropriate resolution. 228 As data is accumulated, quantitative failure models will be devel- oped. Generic decay functions will be described such that the form of maintenance threshold criteria can be determined and the criteria values for specific vehicle types estimated. These functions, together with the use and maintenance pattern data, will be used to describe the sig- nificant failure model cascades with steps in the sequence given proba- bility allocations and individual cascades assigned frequency functions. ANALYSIS COMPONENTS The partition of the analysis of results is required for the iso- lation of independent, or mostly independent, failure mechanisms. Physical components include the upper cylinder, engine mechanical, cooling, fuel, starting, charging, chassis electrical, drive line and braking subsystems. Operating components include the duty cycle (in- cluding the fraction of inoperative time, off- road operation, etc.), the stress pattern to which the vehicle is subjected, and the actual implementation of the scheduled maintenance actions. UPPER CYLINDER VARIABLES In the design of a data collection plan for use in long-term sta- tistical failure correlation studies, it is necessary to anticipate which variables will predict failures of the type anticipated. For example, it has long been part of the conventional wisdom that the engine piston-speed history is the primary predictor of the cylinder wear phenomena — time at speeds above 2000 feet per minute increases wear rate at increasing exponential rates, rapidly reaching a power of four (4) . I have estimated a preliminary set of independent cumulative stress variables for purposes of initial test design. These include the time histories spent at excess rpm, over temperature, low temperature/ high EGT* (loading while cold) , high rpm/high EGT (heavy load condition — inertial and pressure forces both high, but balanced), high rpm/low EGT (unbalanced inertial forces) , total cumulative hours (perhaps partitioned by rpm threshold domains) and the duty cycle portion of time the vehicle is dormant. Experience accumulated in early testing will help refine this variable list. Dependent variables will include the trend in power capacity of the engine and imbalance among cylinders in power production (and compres- sion) . In addition, the collection of failures as defined by the need for maintenance are also defined as independent variables . PRELIMINARY TEST DESIGN As a preliminary starting point in the refinement of the test de- sign and sampling plan, logit analyses are being employed to determine the failure profile, sample size and sampling rate implied by various stress predictors. For example, an occurrence 3 percent of the time * Exhaust Gas Temperature 229 (in 7.5 hrs/month operation) resulting in a 3-year mean time to failure suggest a progressive trip failure probability of 0.0014 initially (zero hours on vehicle), 0.0028 after 4.5 hours of stress (1 1/2 years of duty), and 0.0056 after 8.1 hours of stress (3 years). In order to be 80 per- cent sure that this profile, rather than a constant rate of 0.0028 pre- vails, a data sample of 30 vehicle-years of experience is necessary, if the various explanatory variables are uncorrelated. If they are corre- lated (with a factor of 0.7), the experience requirement is 100 vehicle- years. If the MTBF is doubled, the sample required is reduced by 30 per- cent, and if quadrupled, the sample required is reduced by 50 percent. These parameters would also imply a sampling rate resolution of approxi- mately one data point per vehicle trip. As the program progresses and experience and data are accumulated (both within and outside the VMS effort) on the vehicle types to be tested, analyses of this type will be used in refining the test design and ultimately the interpretation of the field demonstration results. The engineering vehicle tests will be conducted so as to benefit the test design process as well as the proof test of the VMS design. This will be accomplished by recovering data from the VMS on a schedule to insure a continuous high resolution record and to accumulate the maximum vehicle test experience on an M60 tank and an M35 2 1/2-ton truck. Our data analysis computer code for use in evaluating these data will emphasize the accessibility to high resolution data and the manip- ulation of that data into explanatory variables for long-term corre- lations. The results of the correlation studies of these variable com- binations will enable refinement and specification of the system firm- ware (PROM) including the allocation of the memory capacity among high resolution, trip data, periodic summation/histogram data, special power test data, and maintenance action input data. Some minor redesign of the sensor suit is also possible. BUILT IN DIAGNOSTICS— AN ULTIMATE CONCEPT? It is interesting to speculate, even if not highly productive as yet, on what form an ultimate system might take in the future of Army vehicle maintenance. An immediate candidate is the VSA concept but with the recognition that such systems are probably far in the future. There is also the added problem of using intermittent inspection in con- junction with precipitous failure behavior. Only continuous monitoring with warning alarm plus periodic readout can satisfy the total screening function regardless of the distribution of failure model shapes. If we examine the prospect for eventually implementing continuous monitoring, we find reasons for optimism. The form of system involved is built-in diagnostics and it will derive naturally from the experience the automotive industry is now developing as a result of imposed emission and fuel consumption regulations. These developments will assuredly produce inexpensive and survivable sensors and on-board microcomputers such that the addition of the diagnostic function is a relatively small, incremental complication. 2^0 Diagnostics will require additional computation and memory, as well as on-board status and warning displays and a data readout feature. The diagnostic logical analysis and limit testing can be accomplished by means of either the on-board system or by the data recovery device — depending on cost effectiveness tradeoffs. The benefits of such a system include two features of particular importance in the combat environment, in addition to the general bene- fits of fault warning and screen testing. One of these features is the ability to diagnose the repair requirements of a "dead" vehicle. This capability can be readily used to determine the resources necessary to regenerate the vehicle to service so that optimized maintenance resource allocation can maximize total fleet readiness. The other feature is a benefit of the programmed diagnostic logic and the ease of access to quick diagnostics at the operational level. This feature enables the company commander to rely on his motor sergeant to optimally schedule the maintenance of his vehicle fleet such that the fleet readiness is maximized. In a combat situation, this close coupling of the tactical commander with the quantitative understanding of the readiness of his vehicle fleet could greatly enhance his unit's combat effectiveness. 251 MILITARY LAND VEHICLE MAINTENANCE Scheduled Maintenance || Periodic -«--*-J Health h* - 1 Screening , ' Operator Limit Checks Vehicle Employment — i " " 1 Vehicle Fault Repair Fai ure Isolation CHART 1 HEALTH SCREENING (ANTICIPATORY MAINTENANCE) • BENEFITS: REDUCE CATASTROPHIC FAILURES REDUCE SECONDARY FAILURES REDUCE UNNECESSARY MAINTENANCE BASIS FOR VEHICLE READINESS PROGNOSIS • REQUIREMENT: • CRITERIA: GENERIC FAILURE MODELS ABSOLUTE DECAY DECAY RATE • TEST EQUIPMENT: STANDARD SHOP STE-ICE (DCA ) ATE-ICE DYNAMOMETER STRAP-ON BUILT-IN VSA CHART 2 232 FAILURE MODEL Short Cooldown After Boil Coolant Loss Infrequent PREOP Progressive Coolant Loss Cooling System Leak Clogged Radiator Coolant Side Unobserved Increasing Coolant Temperature Trend Unobserved Overtemperature Operation Boil -Over Reduced Coolant Circulation Loose Water Pump Belt Debris In Radiator Air Side Reduced Heat Rejection Catastrophic Cyl inder Failure CHART 3 FAILURE /DECAY FUNCTION SENSED VARIABLE TIME / MILEAGE CHART 4 233 IMPLEMENTATION PROBLEMS • FAILURE MODEL DATA - DESIGN / QUALIFICATION CRITERIA - AMENABILITY TO TRENDING • MAINTENANCE COST DATA - INCONSISTENCIES - SOURCE RELIABILITY CHART 5 VEHICLE MONITOR SYSTEM • OBJECTIVES VEHICLE-BORNE DIAGNOSTIC TECHNOLOGY CONCEPT DEMONSTRATION -PI LOT DATA • PARTICIPANTS ARPA - DEMONSTRATION PROGRAM RAND - PROGRAM COORDINATION Test Design/Analysis TARADCOM - CONTRACT ADMINISTRATION Test Planning/Logistics RCA - HARDWARE DEVELOPMENT • SCHEDULE ENGINEERING VEHICLE TESTS -BEGIN FY 1978 FIELD DEMONSTRATION TESTS - BEGIN FY 1979 • RESULTS VEHICLE DESIGN/QUALIFICATION CRITERIA ISSUE EQUIPMENTS/PROCEDURES CHART 6 23>+ VMS DATA DEVELOPMENT • DATA TYPES: USE PATTERNS CONDITION PROGRESSIONS MAINTENANCE REQUIREMENTS • FAILURE CORRELATIONS: SHORT TERM -- STRESS EXCURSIONS LONG TERM - - CUMULATIVE STRESS TECHNIQUES: LOG IT ANALYSIS ( LOGISTIC PROGRESSION ) EXPONENTIAL REGRESSION FAILURE MODEL: DECAY FUNCTIONS DEVELOPMENT CASCADES CHART 7 ANALYSIS COMPONENTS • VEHICLE SUBSYSTEMS - UPPER CYLINDER _ CHARGING - MECHANICAL/LUBRICATION - CHASSIS ELECTRICAL - COOLING _ drive LINE - F UEL - BRAKING - STARTING • OPERATING INFLUENCES - DUTY CYCLE - STRESS - SCHEDULED MAINTENANCE CHART 8 235 UPPER CYLINDER VARIABLES CUMULATIVE STRESS EXCESS RPM OVER TEMPERATURE LOTEMP/ HIGHEGT HI RPM/ HI EGT HI RPM/ LO EGT CUMULATIVE HOURS DORMANT FRACTION DECAY / FAILURE . POWER TREND . CYLINDER BALANCE - FAILURE REQUIRING MAINTENANCE CHART 9 PRELIMINARY TEST DESIGN EXAMPLE 3% OCCURRENCE, 7. 5 HR S / MONTH , 3YR MTBF FAILURE PROFILE SAMPLE SAE CUM. HRS. TRIP FAIL. 0. 14% PROB. • 80% SURE OF PROFILE: UNCORRELATED - 30 VEHICLE -YRS 4 5 0. 28% CORRELATED - 100 VEHICLE-YRS 8. 1 0. 56% • 1.5 YR MTBF -0.7 (SAMPLE ) • .75 YR MTBF -0.5 (SAMPLE ) SAMPLING RATE -DATA POINT /TRIP ENGINEERING VEHICLE TESTS • CONTINUOUS HIGH RESOLUTION DATA • MAXIMIZE VEHICLE UTILIZATION • USE TO REFINE DEMONSTRATION TEST DESIGN — Prom — Sensors — Memory Allocation CHART 10 236 BUILT - IN DIAGNOSTICS AN ULTIMATE CONCEPT ? • EMISSIONS AND FUEL CONSUMPTION REGULATIONS - INEXPENSIVE / SURVIVABLE SENSORS - ON - BOARD MICRO COMPUTERS • DIAGNOSTICS - INCREMENTAL REQUIREMENT - ON - BOARD WARNING / DISPLAYS - READ -OUT DIAGNOSIS / REPAIR INSTRUCTION • BENEFITS - MINIMUM MAINTENANCE COSTS, MAXIMUM READINESS Anticipate Short and Long Term Failure Progressions - "DEAD" VEHICLE DIAGNOSIS - MOTOR SERGEANT READINESS MANAGEMENT CHART 11 237 VEHICLE MONITORING SYSTEM S.C. Hadden, R.E. Hanson, M-..W. Stewich RCA Government Systems Division Post Office Box 588 Burlington, Massachusetts 01803 RCA Government Systems Division is developing a Vehicle Monitoring System (VMS) under contract to the Tank-Automotive Research and Develop- ment Command (TARADCOM) and the sponsorship of the Advanced Research Projects Agency (ARPA) . The system will be a flexible research instru- ment to be installed in selected military vehicles for the purpose of collecting data supporting an ARPA-TARADCOM investigation into the high cost of operating and maintaining the land vehicle fleet. The VMS will supply comprehensive data describing the true use, condition and main- tenance exposure of its host vehicle. This data will impact future vehicle designs and maintenance procedures. A prototype development contract has been awarded to RCA following a Phase I design study during the summer of 1976 and design is now in pro- gress. The prototype will be installed in two high density vehicles: the M35A2 2-1/2 ton truck and the M113A1 armored personnel carrier.^ Future plans call for installing a number of these systems in a variety of vehicles including the M151a2 1/4 ton truck (jeep) and the M48A3/ M60A1 tank. Hardware consists of a VMS Electronics Assembly (VMSEA) and a harness assembly, both installed in the vehicle, a Maintenance Action Input (MAI) device, and Data Retrieval Equipment (DRE) . During the course of the investigation, the on-vehicle assemblies will automatically collect and store vehicle generated data and maintenance data inputted by a mechanic via his MAI whenever he services the vehicle Following nominal one month intervals, a contact team participating in the investigation will rendezvous with the vehicle, connect the DRE, extract the accumulated data and service the on-vehicle assemblies as necessary. Special vehicle tests may also be performed at this time and their results automatically compiled with the accumulated data. The data, encoded in digital format and requiring no human inputs other then those of the mechanic, will then be forwarded to a processing center for reduction and analysis. 238 Data Acquisition The data acquisition scheme is based on collecting vehicle data which can be processed to evaluate a specific set of Vehicle Indicators . The indicators, which result from the first order reduction of raw VMS data at the processing center, quantiatively describe vehicle use, condition and maintenance exposure throughout the data collection interval. They combine all the VMS generated data into a structured data base available for further analysis. Table 1 lists the indicators that have been es- tablished. The raw VMS data is acquired by monitoring key Vehicle Parameters during normal operation or special tests and by accepting coded inputs from a mechanic. The parameters which will be monitored for four high density vehicles are listed in Table 2. The selection of vehicle parameters and the design characteristics of the VMS Electronics Assembly are critically related. The assembly is not merely a recording device. With its internal microprocessor, it manages the acquisition of parameter data and performs data reduction to a level where the indicators' raw data requirements are met with a mini- mum transducer count and on-vehicle memory capacity. This is illustra- ted in Figure 1. For example, the on-vehicle processor can insure that the value of some particular parameter is recorded only when it carries some specific significance, or it can determine some critical relation- ship among several parameters and enter this single result into storage rather than the values of the individual parameters themselves. Table 1. Vehicle Indicators Usage Indicators Engine starts Failed starts Push starts Jump starts Cranking in gear Oil Temperature out of range Ambient Temperature Electrical load Fuel consumption Parking brake engagement Warm up time Engine b lipping* Total revolutions Clutch operation Engine RPM Engine overspeed Road speed Front wheel drive engagement Engine load High Load periods Terrain exposure Coasting mileage Total mileage Brake application Cool down time Vehicle acceleration/ deceleration Gear ratios Tracked vehicle turning Mechanic's term for pulsing accelerator pedal 239 T) Table 1. Vehicle Indicators (Continued) Condition Indicators Health of - Starting system Charging system Fuel system Lubrication system Breathing system Cooling system Fluid Levels - Engine Oil Brake Coolant Electrolyte Drive train slippage SI engine firing condition** Diesel power** Compression balance** Maintenance Action Input Indicators Engine Compartment Openings Action Codes Component identification codes Elapsed maintenance time ** Special Tests Data collected during special tests are obtained by monitoring the same parameters providing data during normal vehicle use. These tests con- sist of a CI engine power test, an SI engine idle misfire test and a compression balance test, all of which are significant in describing en- gine condition. They are special in that they require unusual operator vehicle interaction which may not occur during normal operation and, therefore, will be performed by the contact team at the time of data re- trieval. The CI power test requires that the engine be brought from idle to governor speed at full acceleration, followed by fuel shutoff while at governor speed; power is calculated from the acceleration and deceleration rates. The SI engine idle misfire test is performed on a warm engine running at idle. The time intervals between 101 successive point openings are precisely measured and misfires are detected from ir- regularities in these intervals. The compression balance test requires that the engine be cranked while fuel or ignition are "off"; compression balance is calculated from the starter current waveform. Hardware Overview Figure 2 illustrates the hardware configuration. The VMSEA is the core of the system. It interfaces with the vehicle harness assembly which terminates in transducers and switches installed 2k0 O 4J 4-> iH J3 s -a -a ex cu cu os cu cu c ex a a) co co c •H -O "O _ _ 00 cfl CO Cfl CO 3 O O CO 44 o : > xi 4-> Xi T3 _ 9Hi S T3 "a w o a fc « « > 4J PS CI 01 [3 -H a. a o u 3 CU CO CO -H OJ 00 C 4J 4J ■H T3 T3 *H Lj T3 M fl rt C n] o C O O 00 4-1 o cd pi OS H CO 33 HI CU 44 rH j3 •h o a 3 > 4-1 35 cu cu co T-t CU CO CO Xi CU CO C3 CU 4J tH -O T3 4J Xi T3 00 cfl cfl CO cfl o C O O CO 4J o W OS OS X CO PS T* ■H 3 w H JU CD a; X CO X 0) cfl OS CO o Xi X u ^ 4J cu pa j pa ■H a ■H nH j 3 o 3 ai cm 00 CO j CO > 4-1 CS rH c r-> 0) a) c ■H CU •H CU CU X CU J CO 44 > 4J Xl X CO -5 ^ Ji rH O CD O CO CO EX CO to CO <-H O > -> > a ex CO X X X •H 8 ■H •H CO CO CO to CQ o *3 u 3 4J cu 01 CU -H ■U w •H .IS CU ►J 3 ■H 05 3 co CO x § CU > 4-1 li, co pa CU C3 P-t cu 4J C J ■3 H JS r-f CD CX CO l-i O H O CO , . w ?1 (4 H M»l O k 43 3 CO *-- M EX fx O CJ eq G [5 o ■ X > CU CU J3 £ J -c T) 4J a -H -H •H 4J rH •H 3 3 •H CU O >-( ^^ CO 3 > fe CU 0) CO CU 0) 0) ' co 1 cu cu x ►J l-i o ex m B B CU CU rH H H -H ex ex S B CO CU CU CU QJ CU cs c U U X 3 CU CO 4i « «h ij a ■ rled cd a> CU J-l U) CD 4J X C3 x U o 3 Xl H -1 c > e_> O a) CO m > en V fr&? CU hJ cu a (U CU X ex CU m 4J CU ■u 1J ffl 3 •H rt PJ m < Ix, o M H -u a u ex a • a s & cu . . cu S H ex & h cu _, B B H ^H CU CU rH ■H H H -H ^1 O O -H ^-. 4J 4J O - CU CU C CS • l-i es cu co cfl • (0 »H "H i — I CS y_i ex oo ja o co *u CUCUCUCUCUJ-CCUCU Xt i-i Xi n u cu u ■ ■ w U H Q -v co co CO co 4-) co ex a ex ex ex 4J ex co co co co co co co CO 4J ex xi co cu (1) 4-1 1J OU c in CU cu ex ex p u 3 CU xi 3 Cfl X p < < ■CU H CJ ■J 3 CO cfl o > > o 3 x x cu cu CO CO cO CO cfl w c c £ j: a 4-> 'cScSS-g 3 rH J3 >-. T3 •n CO ex CU cu > 4J OS CU cu c -H a. a. XI •1 cu CO CO •H CO c 4J 4-1 ■H ■o •n •H u T1 M en m C CO C £ o ft/1 4J o CU OS M CO X HI > T) 44 «j CJ >„ ■H M CU rH J 3 CO £ CU CO Xi S > 4J ex co CO Xi -— CO Xi o iH o CO CO Xi -H o Xi ex Ex o O P3 CO CO CO ex ex ex CO CO co O 3 X rH 3 Xi t> CJ o CU Cfl CU >* >i Cfl CU QJ i — ( S a • cu B an cj B H s a cu cu H H H JJ dJUHIIIllllucc 44 U O m (3 < ex tx x xi x c CO cfl (fl -H "H Xi ex ex ex oo j3 •H CO CO CO ^ CU Cfl CO cu X33XXCUXX C MX CO cfl (0 cfl CU _ co c4-i ex j; x a u td nj 4J a a u CO CO QJ CO o ■o f^ X 4J B cu o Cfl cu en ^1 < cu 3 • . ao co 4J xJ •H C .J3 3 C/3 U 2 uuuuuuuuuo IMMMMMl-llHl-IMMCJCJUCJCJO Px<<<OrHCNcn~3-u-)voI^ panana HHHHHHOUU Cjc_) OSQSDSoSoSQSHHHO _| i"ww N NNNNtocotonnnnnnco action triggers, and commands received from the DKE. In the standby mode, power is applied only to those circuits nec- 2k3 Table 3. VMSEA Design Characteristics Weight 18 pounds Size 13.5" x 9.2" x 6.8" Power Consumption (10 to 32 volt input fr om vehicle) Standby Mode " 0.3 watts Active Mode - 15 watts Life on Backup Battery - at least two weeks Environment o o Temperature, Operating: -40 to Q 145 F Q Non-Operating: -70 to 185 F Humidity: 95% condensation, occasional splashings Altitude: to 15,000 Ft. operating, to 40,000 Ft. non-operating Vibration: 3G (5 - 500 Hz), 50G (10 - 1500 Hz) Design Features Microprocessor: RCA 1802 CMOS Memory: 30K Byte CMOS RAM Flexibility Programmability: Program loading via DRE Expandability: Two spare boards with power, output channels for transducer conditioning circuitry and memory expansion Applicability: Applicable to all vehicle transducer sets Ma in ta inab il i ty Internal self check with status display to contact team Isolation to component level replaceable by contract team No field calibration required Reliability Verification of data transfer at retrieval MIL-STD established reliability parts MTBF: 1030 Hours (30 months for heavy usage vehicle) 2kk essary to maintain storage and recognize action triggers which switch the VMSEA to the active mode. The backup mode is an emergency mode in which only the clock and memory circuits are powered in the event that prime power, normally drawn from the host vehicle's batteries, is interrupted. The internal battery will sustain the clock and memory for at least two' weeks . VMSEA Data Memory Major attention has been given to the architecture of the VMSEA data mem- ory It will be implemented with solid state CMOS devices to take advan- tage of their reliability and low power consumption. A 16K byte (8 bit) capacity is feasible given the density of current LSI devices and the constraint that the VMSEA be of unobtrusive size. In order to compress all of the significant vehicle data generated during a thirty day inter- val into this capacity, a scheme has been devised in which a portion of the data, particularly that surrounding a significant event, is retained xn high resolution format while all of the data is retained in a "binned" format. The high resolution format will consist of time-ordered data samples en- tered into storage at the sampling rate peculiar to each parameter. En- gine speed will be sampled at a 0.1 second rate and, in the absence of a uniform acceleration, a time-tagged entry will be made for each 200 RPM change. Should a uniform acceleration be detected, only the end points will be recorded. Engine oil level will be sampled at a much lower rate. The level will be measured following each shutdown and the first "low" indication of the day will be recorded. The sampling algorithms and high resolution format will permit a detailed reconstruction of the vehicle operating profile when necessary. However, the storage capacity required to retain all data in this format could be- come unwieldly, particularly in the case of a vehicle exposed to heavy use. The bulk of a vehicle's operating profile can be adequately recon- structed with less temporal resolution. The microprocessor will be used to reduce all data, originally sampled at high resolution rates, to the compact "binned" form. A "bin" will consist of a reserved memory segment for recording some spe- cific result computed between selectable time limits. The contents of each bin will be a single entry: a count of events, an accumulated time, etc. A typical bin for storing engine speed data will contain the total time that the speed lay between, say, 2000 and 2200 RPM during a specific week. Another example is engine oil level; a bin will contain the number of days that the oil was found to be low during the week. The actual bin structure is more complicated and is designed to extract the maximum in- formation from each parameter. Twenty-five bins will be allocated to en- gine speed data for a nominal one week interval: (a) total time between 2U5 0+ and 400 RPM, (b) total time in 200 RPM increments, 400 to 5000 RPM, (c) number of 400 RPM crossings. The entire bin structure will be soft- ware adjustable. A definite amount of storage capacity, independent of vehicle usage, can be allocated to storing binned data. Once this allocation has been made, the remainder of the memory can be allocated to storing high resolution data detailing some portion of the vehicle's operating profile. In operation, all sampled data will be both entered into high resolution storage and processed into bin storage. When the memory capacity allo- cated to high resolution storage is filled, the memory contents will be overwritten on a first in/first out basis; binned data will never be ^ overwritten, only updated. At any given time, the memory will contain all the binned data accumulated since the start of the collection period plus the most recent high resolution data. The process will continue as described to the end of the collection in- terval unless some significant event occurs which will cause overwrite to be inhibited. The VMSEA can be programmed so that any event which it can detect can be used as an inhibit trigger; fifteen triggers can be set at any one time. Should overwrite be inhibited by some event, e.g., a me- chanic's entry that he replaced the engine, predetermined amounts of high resolution data surrounding the change will be protected. Figure 3 shows the memory that has been allocated for the different forms of storage. In addition to those discussed, fixed storage areas have been allocated for maintenance action data (forty-eight mechanic entries plus dated detections of forty-eight compartment openings), trip data (a set of eight data entries for one hundred automatically recognized trips), dynamic variables of the VMSEA operating system, special test results and confidence test results. The 13. 3K bytes of data memory remaining after allocation of fixed stor- age, along with the type of vehicle and the usage which it receives, de- termines the percentage of total high resolution data generated in a nom- inal thirty day interval which can be retained. This is summarized in Table 4. The quantitative estimates of vehicle activity were developed by defining use elements *• a minimum monthly element, a cruise element, and a tactical off-road element - of specific duration and activity content. The quantity of data produced by each element was calculated and combinations of these elements were used to estimate monthly totals. 2k6 HIGH RESOLUTION DATA STORAGE (13.3 K BYTES) FIXED STORAGE AREA (2.7 K BYTES) TOP OF STACK FIXED LENGTH BLOCK (512 BOTTOM OF STACK BYTES) BIN DATA STORAGE (250 K BYTES) WEEK 1 WEEK 2 WEEK 3 WEEK 4 TRIP DATA STORAGE SPECIAL TEST RESULTS CONFIDENCE TEST RESULTS MAINTENANCE ACTION INPUTS OPERATING SYSTEM VARIABLES FIRST IN/FIRST OUT MECHANISM Figure 3. Data Memory Organization Table 4. Estimated Amount of Monthly Data Retained in High Resolution Format Vehicle Type and Monthly Use Wheeled (M151, M35) Light Moderate Heavy Tracked (M113, M48/M60) Light Moderate Heavy Hours 7.7 23.4 41.7 0.8 10.7 24.8 Data Generated In Month 9.8K Bytes 26. 3K Bytes 46. IK Bytes 2.9K Bytes 15. 9K Bytes 33. 6K Bytes Percent Retained In High Resolution Format All 50% 29% All 84% 40% Data Memory Size: 16K Bytes (g bu) Allocated for High Resolution Storage: 13. 3K Bytes 2J+7 Confidence Tests and Fault Is olation Th^ VMSEA will conduct routine confidence tests to assess its own health and that of the vehicle harness assembly. Test results will be acquired automatically during unattended operation. The bulk of the data will be sto-ed and transferred to the DRE at data retrieval for forwarding to the"processing center. There it will be used to determine long term health and calibration trends, e.g., transducer offset drifts. A portion of the results will be processed in the VMSEA and stored in a format which can be displayed to the contact team when the DRE is connec- ted This will enable the contact team to make on-the-spot corrections: replacement of internal battery, entire VMSEA, defective transducer and repair of cables. As a typical example, the DRE might display a code which indicates the presence of a fault in channel 32. The contact team will then substitute a dummy harness assembly for the real one and deter- mine if the fault persists. If it does, the VMSEA will be replaced. If it does not, channel 32 harness wires will be checked and repaired as re- quired. Finally, the channel 32 transducer (oil temperature RTD) will be replaced. Replacement of certain malfunctioning transducers may be beyond the capability of the contact team. In this case, they will call for support. Program Schedule The prototype VMS system will be completed for acceptance testing during June 1978. It will then undergo a period of testing on selected vehicles to determine what changes should be incorporated in P r °^tion unxts to optimize data acquisition. When finally operational, the VMS is expec ted to furnish data impacting the design, operation and maintenance pol- icies of military vehicles. Potential Applications The VMS is potentials applicable to all classes of ye hides - land, marine and air. The exploitation of microprocessor technology provides flexibility and adaptability through software changes rather than hard ware re-design. In addition, it minimizes data memory requirements thereby allowing the use of relatively small, but relaible and rugged solid state memories which are well suited to application m hostile and space limited vehicle environments. Acknowledgement The authors wish to acknowledge the contributions to the development of the VMS provided by Major T. Covington, ARPA, Mr. D. Sarna, TARADCOM, their staffs and their consultants from Rand and Battelle. 2U8 SYSTEMIZED DIESEL ENGINE DIAGNOSTICS Henry J. Mercik, Jr. Hamilton Standard Division United Technologies Corporation Windsor Locks, Connecticut 06096 Introduction The advent of today's electronic systems brings engine manufacturers and engine owners new concepts which allow non traditional methods for improving engine maintenance techniques. Economics of engine operation and government environmental regulations are additional motivating factors which are influencing the current interest in reducing main- tenance costs and improving engine performance. A study of past main- tenance practices reveals an inconsistency in procedures utilized for diagnosing engine problems and a pattern of standard repair procedures which are employed when a particular operational problem is reported. The repair procedures generally result in replacement of several engine components without seriously attempting to isolate the faulty component. The basic problem may or may not be solved utilizing these procedures and therefore additional maintenance may be required in a second attempt to solve the problem. These methods are inefficient and costly, a cost which is borne by the owner or a warranty program. An approach for engine maintenance which shows considerable promise for success utilizes the concept of systemizing the diagnostic procedures for identifying problem causes and directing the mechanic to a partic- ular repair procedure. A system for the maintenance of gasoline engine powered vehicles is now in production. A second generation of this system which is called AutosenseR has been developed for diagnosing diesel engine malfunctions. The basic AutosenseR unit is utilized for both the gasoline and diesel engine test systems and therefore minimizes equipment costs because of the economies resulting from fabrication in production quantities. The commercial truck market is the primary market for which the diesel test equipment has been configured. However, the test techniques which have been developed are suitable for appli- cation to marine, agriculture, construction, and off -highway diesel engine maintenance. In addition, since the system is tape controlled its uses are not confined to testing engines. Inclusion of the proper sensors and diag- nostic program on tape cassette allows expansion for testing trans- missions, vehicles and other types of equipment requiring sophisticated diagnostic procedures. 2^9 System Description The diagnostic system is a computerized tool for fault isolation of diesel engines and their subsystems. It consists of a computer console and a hand held controller. The computer console contains a data acquisition unit, teletypewriter, computer and a tape drive system. The transducer and solenoids for handling pneumatic and hydraulic signals are contained in a transducer box which is mounted to the rear panel of the console (Figure 1). The control program, test sequences, tests, test limits and specifications are contained on a small tape cassette which is placed into a tape reader. Reprogramming of the tape cassette allows for specification changes or expansion for testing additional engines and other types of equipment. ENGINE/VEHICLE SENSORS TEMPERATURE PRESSURE DATA ACQUISITION UNIT COMPUTER OPERATOR INPUT/OUTPUT DEVICE TAPE STORAGE UNIT EMISSIONS PROBE Figure 1. Diesel Diagnostic System Block Diagram Sensors and sensor location was a significant consideration when con- figuring the system. The desirability to utilize engine mounted trans- ducers is apparent, however, we selected the approach of using a trans- ducer mounted in an enclosure at the unit. The primary reasoning which influenced the decision was cost, durability and performance of trans- ducers commercially available for the possible applications. If sensors were to be specifically designed for the application, the engine mounted sensors would be preferred. Operation The HIS scheme for testing is to perform all tests on the engine without requiring a dynamometer. The non-dyno test techniques are relatively 250 new and can be run quickly in both the steady state and dynamic test modes. A comprehensive approach to diesel engine diagnostics must include the ability to test for particular recurring problems, engine subsystems and individual tests independently. To implement this approach complaint oriented and subsystem sequences have been estab- lished which, when selected as the desired sequence for testing, auto- matically lead the operator through the testing cycle (Figures 2 § 3) . Complaint Oriented Slow Crank No Crank Cranking/No Start Hard Start - Hot Hard Start - Cold Low Power Miss or Rough Running Fast or Slow Idle Smoke Oil Pressure High or Low Overheating Subsystem Sequences Fuel System Lube System Cooling System Air Intake System Turbocharger Air Charging System Electrical System Figure 2. Diesel Diagnostic Sequences Individual Tests (Partial Listing) Engine Power Power Contribution Relative Compression Blow-By Gage Accuracies High § Low Idle Battery Condition Fuel Pump Calibration Figure 3. Procedural Test The system is designed to diagnose engine problems with test sequences that are executed without external loading of the engine. By analyzing results of several tests a recommendation for a repair procedure is made to the operator. Low Power Diagnostic A common complaint of diesel engine operators is low power. An example of a diagnostic sequence for low power with its truth table and repair codes is illustrated in Figure 4. The tests selected for the sequences are listed on the left and the test results, pass or fail, are contained within the matrix. As a result of the engine performance when the diagnostic sequence is run, a pass or fail is recorded for each test. The pass -fail results for an individual test may not reveal the problem by itself. However, when included in a sequence containing the results of several tests, a determination of the problem cause can be 251 established. The sum of the test results in a sequence will refer the operator to a repair code which defines the problem cause and recommends a repair procedure. TRUTH TABLE TEST COMP P F F P POWER P AIR IN FUEL P F CH PT 2 P F < ^> CH PT 1 P ,? f ? PEAK PRESS P ^ 5> F GOV CUT OFF P l > ^ F BLOW-BY P P f iff AIR FILTER P 1 p; F TURBO P F POWER CONT P F DIAGNOSTIC CODE M S s s K s g = DIAGNOSTIC MESSAGES CODE MESSAGE 151 15? 15) 151 IS5 15' 15 ! 151 160 16 1 ENGINE NORMAL VALVE LEAKAGE CYLINDER OR PISTON DAMAGE AIR IN FUEL FUEL PUMP OUT OF CALIBRATION. RUN 928 SAME AS 155 SAME AS 155 SAME AS 155 EXCESSIVE INLET RESTRICTION SLUGGISH TURBOCHARGER INJECTOR PROBLEM Figure 4. Low Power Diagnostic Horsepower The horsepower test is a dynamic test developed for the diesel diag- nostic system (Figure 5) . Data are gathered during the free accel mode of engine operation. During a full throttle acceleration, the engine accelerates with its own inertia as a load. Torque is determined by the rate of acceleration times the engine inertia, T = la. Horsepower is obtained from the torque at any speed, HP = KTN. Actually, the torque generated during the accel is related to the specified torque, and by applying an algorithm developed for this purpose, torque which the engine is capable of producing can be determined. • FULL THROTTLE ACCELERATION • T = la • HP = K T N INGINE 1500 SPEED (RPM) 1000 Figure 5. Diagnostic Techniques - Power 252 Turbocharger Tests such as the turbo tests do not require full loading but do require a snap accel from idle to high idle and a hold at high idle for about five seconds (Figure 6). During the snap, inlet manifold pressure data are collected which when compared with previously established dynamic limits provides an assessment of turbo condition. The characteristics of turbo performance which are used in the analysis are the inlet mani- fold pressure depression, time of depression, pressure rise as a function of time and max value of inlet manifold pressure. INTAKE MANIFOLD PRESSURE (PSIG) *~"^ I __] 2500 2000 ENGINE 1500 SPEED 1000 (RPM) 500 DIAGNOSIS STUCK, SLUGGISH.OR MALPERFORMING TURBOCHARGER Figure 6. Turbocharger Health Cylinder Health A sound indicator of cylinder health is the relative compression test (Figure 7) . By monitoring the battery current drawn during engine cranking, an evaluation can be made as to the work required during compression for each cylinder. By selecting the cylinder requiring the most work for compression and comparing that with the work required during the compression stroke for each of the other cylinders, an eval- uation of relative compression can be established. This test in itself will not identify the fact that a valve, ring or piston must be replaced. However, if other tests such as blow by are run in conjunction a more definite analysis can be concluded. The system output can, therefore, differentiate between ring/cylinder problems and valve and gasket problems. 253 DIAGNOSTI C TECHNIQUE BATTERY \Aa/ \AA/ \Aa \l\l\l CURRENT (AMPS) \/\/\/ V 1/ V \j\j\I \l\l\l CRANKSHAFT ANGLE(RADIANS) I LOW RELATIVE CURRENT INDICATIVE OF LOW COMPRESSION Figure 7. Relative Compression Fuel Pump Calibration A calibration check of the Pressure-Time (PT) type fuel pump can be achieved by sampling fuel pressure at predetermined reference points during a snap acceleration of the engine. These results contribute significantly to analysis of fuel system and low power complaints (Figure 8) . 120 RAIL FUEL 80 PRESSURE (PSIG) 40 ,0 1000 1500 2000 2500 ENGINE SPEED (RPM) • FULL THROTTLE ACCELERATION • ANEROID DEFEATED • FOLLOWED BY DECELERATION FOR THROTTLE LEAKAGE TEST Figure 8. Fuel Pump Calibration Lubrication System By locating a transducer at a point downstream of the pump and regulator but before the oil filter, an assessment of the lube system can be determined. During a snap accel the pressure can be tracked and the time to reach regulation and the regulation pressure can be established 25k (Figure 9). This information, when compared with functional data from a properly operating system , can be used to determine pumping perfor- mance, regulation, lube system flow capacity, and in some systems, the point when piston cooling commences. MAIN OIL PASSAGE PRESSURE (PSIG) 60 40 20 500 1000 1500 2000 2500 ENGINE SPEED(RPM) • SLOPE INDICATES RELATIVE SYSTEM FLOWRATE • BREAKPOINT OCCURS AT PRESSURE REGULATOR OPENING • OIL TEMPERATURE ABOVE I65°F Figure 9. Lubrication System Criteria Lube system flow capacity can indicate a stoppage in flow passages. This may mean a clogged filter or passage. Relative Power Contribution The relative power contribution test is one in which the power being delivered by each cylinder is determined and compared against the power being contributed by each of the other cylinders. The cylinder found to be delivering the most power is rated 100 percent and each of the other cylinders is compared to it in percent of power contribution (Figures 10a § 10b) . DIAGNOSTIC TECHNIQUE RELATIVE POWER/CYL (PERCENT) CYLINDER NUMBER Figure 10a. Power Contribution 255 NORMAL OPERATION ONE CYLINDER DEFEATED FULL THROTTLE ACCELERATION TIME INTERVAL BETWEEN ANGLES OF ROTATION LOW POWER CYLINDER FOUND BY SMALLER THAN NORMAL CHANGE IN TIME INTERVAL Figure 10b. Diagnostic Technique - Power Contribution Cooling System The cooling system diagnostics presents some interesting problems which do not enter into diagnostics of other engine elements. This is due to the fact that the engine cooling system is not complete in itself. It must interface with a radiator or cooling system which is not pro- vided by the engine manufacturer. At this time the cooling system discussion pertains to the engine cooling system and most of the testing will be completed with the thermostat closed. The use of one pressure transducer and one temperature sensor can provide data to determine general system characteristics. The use of a second pressure transducer will allow for further analysis. Using a pressure sensor and a temperature sensor just upstream of the thermostat and running the engine at idle and high idle, thermostat operation, pump performance and shudderstat sequence operation can be determined, and whether the thermostat is functioning properly. Also, by running the engine at idle and high idle with the thermostat closed, an indication of pump performance can be determined (Figure 11) . 2400 RPM 50 25 1 t— IDLE . _ THERMOSTAT INLET PRESSURE (PSIG) 140 170 COOLANT TEMPERATURE(V) • PRESSURE DIFFERENCE BETWEEN HIGH AND LOW SPEEDS INDICATIVE OF SYSTEM FLOW RATE • PRESSURE DROP OCCURS WHEN THERMOSTAT OPENS Figure 11. Cooling System Criteria 256 Diagnostic System/Engine Interface The time required for making the necessary engine connections for diag- nostic testing is an important consideration which has a significant impact on total test duration. Total test duration is that time required to connect signal lines to the engine, conduct the test and disconnect from the engine. Electrical connections consist of attaching metal clips to various locations in the electrical system. These connections are made quickly and easily. The hydraulic and pneumatic connections are more time consuming to make unless a quick disconnect method is utilized. On the initial testing of the engine male quick disconnect fittings are installed at the desired locations on the engine. These fittings remain on the engine to minimize set up time for future diagnostic testing. The diagnostic system is equipped with mating quick set up time that can be reduced further by manifolding all hydraulic and pneumatic signals to a conveniently located panel (Figure 12) . To implement a manifolding scheme cooperation is required by the engine and vehicle manufacturers and the vehicle owner. QUICK CONNECT PANEL AND LINE HARNESS FUEL PUMP LUBE OIL INTAKE AIR COMP PRESSURE PRESSURE MANIFOLD ACTUATOR LINE OR WATER FILL VENT LINE Figure 12. Installation Interface Summary The implementation of a diagnostic product requires an awareness of what diagnostics are and the depth of diagnostic penetration desired. The 257 following is the diagnostic definition which best describes the HTS approach : Diagnostics are those methods used to isolate causes for performance and hardware malfunctions in engine systems. The diagnostic objective is to isolate faulty subsystem components involved in malperfor- mance or failure so that repair is direct and accurate . Therefore, diagnostics (fault finding) has the objective of isolating internal faults. Overall performance is not the governing criteria but is used as an indicator to determine faults. The diagnostic system then provides for the profit conscious user a tool which will enhance his ability to increase engine productivity, reduce maintenance costs and effect a savings in warranty costs for the engine manufacturer. To configure the diagnostic system and establish diagnostics the manu- facturer must comprehensively utilize mechanical, hydraulic, pneumatic, electrical and electronic technologies together with a strong systems engineering effort. 258 MECHANICAL FAILURES PREVENTION GROUP 26th Meeting Detection, Diagnosis and Prognosis List of Registrants Jerry Allen S.KoF. Industries King of Prussia, PA 19406 Richard M. Anderson Marl in Rockwell Corporation 402 Chandler Street Jamestown, NY 14701 Richard C. Arnold University of Nebraska at Omaha P. 0. Box 688 Downtown Station Omaha, NB 68101 Ronald Bannister Westinghouse Electric Corporation Lester, PA 19113 S. Bhattacharyya I IT Research Institute 10 West 35th Street Chicago, IL 60616 Eo P Blyskal Eastern Airlines International Airport Miami, FL 33148 Stephen Bonifazi Pratt $ Whitney Aircraft Box 2691 West Palm Beach, FL 33408 Jerry F. Bouska Commonwealth Edison Company 1319 So First Avenue Maywood, IL 60153 E. Roderic Bowen Foxboro/ Trans - Sonics Inc . P o Box 43S Burlington, MA 01803 Alfred P. Brackney Monsanto Company 730 Worcester Street Indian Orchard, MA 01151 H. C. Burnett National Bureau of Standards Metallurgy Division Washington, DC 20234 Colby E. Buzzell General Electric Company 50 Fordham Road Wilmington, MA 01887 Domenic Canonico Oak Ridge National Laboratory P. 0. Box X Oak Ridge, TN 37830 J. B. Catlin IRD Mechanalysis, Inc. 6150 Huntley Road Columbus, OH 43229 Phillip W. Centers AFAPL/SFL Wright -Patterson AFB, Ohio Robert N. Clark GARD, Inc. 7449 N. Natchez Avenue Niles, IL 60648 45433 259 J. Eo Clevenger Shell Oil Company 999 E. Touhy Des Plaines, IL 60018 George M c Coleman Code 18 Naval Ocean Systems Center San Diego, CA 92152 William H c Cooley Endevco 4950 Harlem Chicago, IL 60656 Raymond A. Coulombe Naval Ship Engineering Center Washington, DC 20362 William E. Deluhery Caterpillar Tractor Company Peoria, IL 61602 John Despot Hughes Aircraft Company 18541 Santa Tomasa Circle Fountain Valley, CA 92708 Neal DesRuisseaux SoKoF. Industries King of Prussia, PA 19406 Thomas Jo Dolan Engineering Consultant 510 S. Highland Avenue Champaign, IL 61820 James H. Drennan, Jr. Bell Helicopter P Oc Box 482 Ft. Worth, TX 76010 Michael Drosjack Shell Oil Company P Oc Box 3105 Houston, TX 77001 John Dunmore Spectral Dynamics Corporation Dymac Division San Diego, CA David V. Dunn Baird Atomic 125 Middlesex Turnpike Bedford, MA 01730 Raymond Ehrenbeck U.S. Department of Transportation Transportation Systems Center Cambridge, MA Kent J. Eisentraut Air Force Materials Lab Wright-Patterson AFB, OH 45433 Lloyd Ho Emery U.S. Department of Transportation National Highway Traffic Safety Admin. Washington, DC 20590 V. J. Erdeman Eastern Airlines International Airport Miami, FL 33148 F. Monte Evens Continental Oil Company 1000 South Pine Ponca City, OK 74601 Paul M. Fleming National Bureau of Standards Metallurgy Division Washington, DC 20234 Jack L. Frarey Shaker Research Corporation Northway 10 Executive Park Ballston Lake, NY 12019 Richard Gallant UcS. Naval Air Test Center Patuxent River, MD 20650 260 John A. George Parks College of St„ Louis Univ Department of Aerospace Engr Cahokia, IL 62206 Paul T. George Pratt § Whitney Aircraft Commercial Products Division 400 Main Street, EBZG3 E. Hartford, CT 06108 Nathan Glassman David Wo Taylor Naval Ship Research and Development Center Annapolis, MD 21402 Norman J. Gleicher Ocean Environmental Systems, Ltd D 1843 Palmer Avenue Larchment, NY 10538 W. L. Groves Continental Oil Company Box 1267 Ponca City, OK 74601 James Hamilton Naval Ship Engineering Center Philadelphia, PA 19112 James W. Harnach Amoco Oil Company Naperville, IL 60540 Christopher R. Harz Rand Corporation 1700 Main Street Santa Monica, CA 90406 Jerard Haury U.S. Army Systems Command St. Louis, M3 36166 Henry R. Hegner GARD, Inc. 7449 No Natchez Avenue Niles, IL 60648 Patrick Henry Illinois Coal Gasification Group 122 South Michigan Avenue Suite 2014 Chicago, IL 60603 Rudolph Hohenberg Mechanical Technology Inc. 968 Albany- Shaker Road Latham, NY" 12210 Robert R. Holden Hughes Aircraft Los Angeles, CA Leo Hoogenboom Mechanical Technology, Inc. 968 Albany-Shaker Road Latham, NY 12210 Stephen Hsu Amoco Chemicals Corporation 1612 Wadham Place Wheaton, IL 60187 Roger Hunthausen U.S. Army Eustis Directorate, AMRDL Ft. Eustis, VA 23606 Rick Hyer Fleet Maintenance Magazine 7300 N. Cicero Avenue Lincolnwood, IL 60646 Alan M. Johnson Factory Service Operations Machinery and Systems Division Carrier Air Conditioning Corp. Carrier Parkway, Bldg. TR-1 Syracuse, NY 13201 James E. Johnson Southwest Research Institute San Antonio, TX 78284 261 John H. Johnson Michigan Technological University Mechc Engr.S Engr. Mecho Dept Houghton, MI 49931 William D. Keeley Vibration Specialty Corp 100 Geiger Road Philadelphia, PA 19115 He K. Keswani Gulf Research § Development Co. P. 0. Drawer 2038 Pittsburgh, PA 15230 Raymond B Kight Naval Air Rework Facility 5 Huntington Drive Pensacola, FL 32508 Ec E. Klaus The Pennsylvania State University Department of Chemical Engineering University Park, PA 16802 Alfred C. Koch Ocean Technology 2835 N. Naomi Burbank, CA 91504 Robert S. Kulp Pratt § Whitney Aircraft West Palm Beach, FL 33408 George Lang Nicolet Scientific Corporation North Vale, NJ 07647 Richard Lee UoSo Government (Army) JOAP-TSC, NARF Code 360 Pensacola, FL 32508 Rudy Lenich Caterpillar Tractor Company 100 N E. Adams Street Peoria, IL 61629 Robert Leon Franklin Institute Research Lab 20th Place Philadelphia, PA 19103 L. Leonard Franklin Institute Research Lab 20th Place Philadelphia, PA 19103 Samuel Jo Leonardi Mobil Research § Development Corp. Paulsboro, NJ 08066 Paul Lewis Mechanical Technology, Inc. 968 Albany- Shaker Road Latham, NY 12110 Theodore Liber I IT Research Institute 10 West 35th Street Chicago, IL 60616 Wieslaw Lichodziejeswki CARD, Inco 7449 N. Natchez Avenue Niles, IL 60648 Basley K. Lindboom Marl in Rockwell Corporation Division of TRW 402 Chandler Street Jamestown, NY 14701 Jon Lippe SINTEF Dept. of Machine Dynamics N-7034 Trondheim-NTH Norway James R. Long AVCO Corporation 4807 Bradford Drive Huntsville, AL 35805 262 Bill Macaulay Stone Piatt Electrical Canada Ltd. 165 Steelcase Road Markham, Ontario Canada L3R 1G1 Robert E„ Manning Assoc. Research Consultant UoS, Steel Research Monro eville, PA 15146 Richard Manteuffel Borg-Warner Research 1200 S. Wolf Road Des Plaines, IL 60018 George E. Maroney Fluid Power Research Center Stillwater, OK 74074 Clifton S McDonald Fruit Growers Express Company 4928 Birch Lane Alexandria, VA 22312 William R„ McWhirter, Jr David W. Taylor Naval Ship Research and Development Center Code 2731, Automation $ Control Div. Annapolis, MD 21402 Alice Mentley Sperry Vickers 1401 Crooks Road Troy, MI 48084 Henry J. Mercik, Jr Hamilton Standard Division of United Technology Windsor Locks, CT 06096 Leslie D. Meyer General Electric 1036 University Road Schenectady, NY 12308 Greg Michel Bruel § Kjaer Canada Ltd. 71 Bramalla Road, Suite 71D Bramalla, Ontario Canada L6T 2W9 Louis A. Mileto Engineering Services Section Systems Engineering Test Directorate NATC Patuxent River, MD 20670 Richard S. Miller Office of Naval Research Arlington, VA 22217 Raymond F. Misialek Naval Ship Engineering Center Philadelphia, PA 19112 John S. Mitchell Endevco Rancho Vie jo Road San Juan Capistrano, CA 92675 H. A. Musallam U. S. Army USAREUR Materiel Laboratories APO, New York, NY 09028 Bill Needelman Pall Corporation 30 Sea Cliff Avenue Glen Cove, NY 11542 James W, Neel Standard Oil Company of Indiana Box 400 Naperville, IL 60540 Frank M. Newman Southwest Research Institute P. 0. Drawer 28510 San Antonio, TX 78284 263 Roger K. Nibert Borg Warner Research Center Wolf § Algonquin Roads Des Plaines, IL 60018 Jack R. Nicholas, Jr. Naval Ship Engineering Center Washington, DC 20362 Peter O'Donnell Naval Air Engineering Center Lakehurst, NJ 08733 James K. O'Steen Naval Surface Weapons Center White Oak Silver Spring, MD 20910 Elio Passaglia National Bureau of Standards Institute for Materials Research Washington, DC 20234 William F. Peete, Jr. Naval Oceans Systems Center San Diego, CA 92152 Joseph M. Perez Caterpillar Tractor Company Technical Center, Bldg. E Peoria, IL 61629 Robert L. Perkins David W. Taylor Naval Ship Research and Development Center Code 1962 Bethesda, MD 20084 G. A. Peters A. To § S.F. Railway 1237 Navana Bars tow, CA 92311 Marshall Peterson Wear Sciences 32 Sutherland Drive Scotia, NY 12302 E. E. Pfaffenberger FMC Corporation, Bearing Div. 7601 Rockville Road Indianapolis, IN 46206 Jerry Philips David W. Taylor Naval Ship Research and Development Center Annapolis, MD 21401 Kenneth R. Piety Oak Ridge National Laboratory P. 0. Box X, 3500 Bldg. Oak Ridge, TN 37830 Anthony D Pisano Prudential Building Maintenance 247 East Ontario Street Chicago, IL 60611 J. F. PrijTim A.T. § S.F. Railway 10916 West 65th Terrace Shawnee, KS 66203 Gerald F. Rester Naval Ship Engineering Center Washington, DC 20362 Seiji Sakata Naval Air Systems Command Rep, Pacific Lualualei, Hawaii FPO San Francisco, CA 96612 Richard G. Salter Rand Corporation 1700 Main Street Santa Monica, CA 90406 Karl Scheller Technology/ Scientific Services Inc. Dayton, OH 45401 George Schmidt General Electric Company Bldg. 2-412 Schenectady, NY 12345 26k Larry C. Schroeder A.T. § S.F. Railway Depto of TR/D Motive Power Building 1001 N.E. Atchison Street Topeka, KS 66616 Robert G. Schwieger Associate Editor, Power Magazine McGraw-Hill 1221 Avenue of Americas New York, NY 10020 Victor Sedlacek Northern Natural Gas Company 215 Keo Way, Suite 310 Des Moines, IA 50309 J. P. Segurson Pacific Fruit Express Company 116 New Montgomery Street San Francisco, CA 94105 Peter B. Senholzi Naval Air Engineering Center Lakehurst, NJ 08733 E. Shamah AMOCO Chemicals Corp. P. 0. Box 400 Naperville, IL 60540 L. P. Shepp CP Rail -Motive Power Engineering Windsor Station Montreal, Quebec Canada H3C 3E4 T. Robert Shives National Bureau of Standards Metallurgy Division Washington, DC 20234 George F. Skala Environment/One Corporation 2773 Balltown Road Schenectady, NY 12309 Jean- Claude Sol EDF Etudes et Recherches 1 Av du General De Gaulle 92140 Clamart, France R. W. Spacie Harris Corp/PRD Electronics Div. 6801 Jericho Turnpike Syosset, NY 11791 William D. Sproul Borg -Warner Corporation Wolf § Algonquin Roads Des Plaines, IL 60018 Parker S. Stafford Martin Marietta Aerospace Corp c/o Jet Propulsion Laboratory 4800 Oak Grove Drive Pasadena, CA 91103 John Stark Fleet Maintenance Magazine 7300 N. Cicero Lincolnwood, IL 60646 Jesse E. Stern Code 721 NASA/Goddard Space Flight Center Greenbelt, MD 20771 M. W. Stewich M.S. 7-1 RCA P. 0. Box 588 Burlington, MA 08103 H. G. F. Stringer CP Rail -Motive Power Engineer Windsor Station Montreal, Quebec Canada H3C 3E4 David R. Sutliff Association of American Railroads Chicago, IL 265 Sami Tabry Allan Crawford Associates, Ltd. 1330 Marie Victorin Boulevard East Longueuil, Montreal Quebec, Canada J4G 1A2 L E. Tarbell Texaco, Inc. 332 S. Michigan Avenue Chicago, IL 60604 William H, Tait Dom. Foundries and Steel P 0. Box 460 Hamilton, Ontario, Canada Thomas Tanher Technical Development Company Executive Vice President Glenolden, PA 19036 R. K. Tessmann Fluid Power Research Center Stillwater, OK 74074 R. Eo Tetrev M.S. 7-1 RCA P 0. Box 588 Burlington, MA 01803 Kenneth R. Thomas Deere § Company 3300 River Drive Moline, IL 61265 Earl R. Tillman AVCO Corporation 4807 Bradford Drive Huntsville, AL 35805 Henry G. Tobin I IT Research Institute 10 West 35th Street Chicago, IL 60616 David Totten All is Chalmers Box 512 Milwaukee, WI 53201 Raymond Valor i Naval Air Propulsion Center Trenton, NJ 08628 John M. Vice U.S. Air Force SA-ALC/MMETP Kelly AFB, TX 78241 Benton E. Visser Shell Oil Company P. 0. Box 1380 Houston, TX 77001 Warren Waldron Shaker Research Corporation Northway 10 Executive Park Balls ton Lake, NY 12019 Walter Walejeski Commonwealth Edison 1319 S. First Avenue Maywood, IL 60153 William J. Whelan Rand Corporation 1700 Main Street Santa Monica, CA 90406 M. Wig ton Pacific Scientific Po 0. Box 3007 Montclair, CA 91763 Thomas P. Will Exxon Research and Engineering Co, P. 0. Box 101 Florham Park, NJ 07932 Earl Witty S K.F. Industries King of Prussia, PA 19406 266 Kathleen Wolf Rand Corporation 1700 Main Street Santa Monica, CA 90406 James L. Wright Hughes Aircraft Los Angeles, CA Andrew Yuvan HIAC Instrument Division, PSCO 100 East Irving Park Road Roselle, IL 60172 267 MFPG PUBLICATIONS Both printed and microfiche copies of the following MFPG publications (1-11) whose catalog numbers start with "AD" or "COM" may be obtained from the NTIS. National Technical Information Center 5285 Port Royal Road Springfield, VA 22161 1. Glossary of Terms AD 721 354 2. Proceedings of Meetings 1-9 (set of 5) AD 721 359 Meeting Nos. 1-5 Papers and discussion on failure analysis and control. Meeting No. 6 "Detection, Diagnosis and Prognosis", December 6, 1968. Meeting No. 7 "Failure Mechanisms as Identified with Helicopter Transmissions", March 27, 1969. Meeting No. 8 "Critical Failure Problem Areas in the Aircraft Gas Turbine Engine", June 25-26, 1969. Meeting No. 9 "Potential for Reduction of Mechanical Failure Through Design Methodology", November 5-6, 1969. 3. Proceedings of Meeting No. 10 AD 721 912 "Vibration Analysis Systems" January 21-22, 1970. 4. Proceedings of Meeting No. 11 AD 724 475 "Failure Mechanisms: Fatigue" April 7-8, 1970. 5. Proceedings of Meeting No. 12 AD 721 913 "Identification and Prevention of Mechanical Failures in Internal Combustion Engines" July 8-9, 1970. 268 6. Proceedings of Meeting No. 13 AD 724 637 "Standards as a Design Tool in Surface Specification for Mechanical Components and Structures" October 19-20, 1970. 7. Proceedings of Meeting No. 14 AD 721 355 "Advances in Decision-Making Processes in Detection, Diagnosis and Prognosis" January 25-26, 1971. 8. Proceedings of Meeting No. 15 AD 725 200 "Failure Mechanisms : Corrosion" April 14-15, 1971. 9. Proceedings of Meeting No. 16 AD 738 855 "Mechanical Failure Prevention Through Lubricating Oil Analysis" November 2-4, 1971. 10. Proceedings of Meeting No. 17 AD 750 411 "Effects of Environment Upon Mechanical Failures, Mechanisms and Detection" April 25-27, 1972. 11. Proceedings of Meeting No. 18 AD 772 082 "Detection, Diagnosis and Prognosis" November 8-10, 1972. 12. Proceedings of Meeting No. 19 COM- 74-50523 "The Role of Cavitation in Mechanical Failures" October 31 - November 2, 1973. Printed copies of the following MFPG publications (stock numbers starting with SN) may be obtained from the Government Printing Office. Superintendent of Documents U.S. Government Printing Office Washington, DC 20402 Microfiche copies of these publications (catalog numbers starting with NBS-SP) may be obtained from NTIS. 13. Proceedings of Meeting No. 20 SN003-003-01451-6 "Mechanical Failure - Definition of the Problem" May 8-10, 1974. Cost: Printed copy $6.10 269 14. Proceedings of Meeting No. 21 "Success By Design: Progress Through Failure Analysis" November 7-8, 1974. Cost: Printed copy $3.50 SN003-003-01639-0 Microfiche copy $2.25 NBS-SP-433 15. Proceedings of Meeting No. 22 "Detection, Diagnosis and Prognosis" April 23-25, 1975. Cost: Printed copy $4.25 SN003-003-01556-3 Microfiche copy $2.25 NBS-SP-436 16. Proceedings of Meeting No. 23 "The Role of Coatings in the Prevention of Mechanical Failures" October 29-31, 1975. Cost: Printed copy $2.65 SN003-003-01664-1 17. Proceedings of Meeting No. 24 "Prevention of Failures in Coal Conversion Systems" April 21-23, 1976. Cost: Printed copy $3.00 SN003-003-01760-4 18. Proceedings of Meeting No. 25 In Press "Engineering Design" November 3-5, 1976. 270 APPENDIX _ REMOTE DIAGNOSTIC TECHNIQUES USED IN VIKING LANDER OPERATIONS* Parker S. Stafford Martin Marietta Aerospace Jet Propulsion Laboratory- Pasadena, California 91103 Abstract: The problem of monitoring performance and diagnosing malfunc- tions on the Viking Landers while operating on the Surface of Mars was complicated by severe data rate and volume restrictions. This paper describes the combination of flight hardware, flight software and ground operations software used for planning, prediction and telemetry data analysis functions which allowed the Viking Flight Team to successfully operate the landers for a combined 548 Martian days to date although many malfunctions have occurred. To illustrate the use of this system, a specific anomaly is discussed in detail. Summary of the Problem The design of the Viking Landers and their subsequent operations on the Martian surface have been greatly constrained by the distances over which the radio signals must travel and the limited view periods which are available. 1 The round trip light time from Earth to Mars to Earth is approximately 40 minutes at distances of approximately 220 million miles. Each lander has a view period to Earth of about 10 hours, and although the Earth can view Mars nearly continuously through the Deep Space Network Stations at Goldstone, California; Madrid, Spain; and Canberra, Australia; these stations must be shared between the two landers, the two orbiters and other projects such as Helios, Pioneer, and Voyager. These constraints coupled with antenna size restrictions and electrical power limitations of the spacecraft, combine to produce very low rates and low total volume for data return when compared to *This paper presents the results of one phase of research carried out at the Jet Propulsion Laboratory, California Institute of Technology, under Contract No. NAS7-100, sponsored by NASA. l-This paper will address certain aspects of operating the Viking Landers on Mars from July 20, 1976 through May 19, 1977. The interplanetary cruise phase (while the landers were attached to the orbiters) and the descent to the Martian surface will not be discussed, although many of the same points apply. 273 earth based on near-earth telemetry systems. In order to maximize scientific data return, engineering data required to assess proper operation or diagnose malfunctions was kept to a minimum. Therefore, the Viking Flight Team was faced with a difficult challenge in attempt- ing to deal with the following questions: 1. How could the lander performance be monitored so that malfunctions could be detected quickly and with a high probability? 2. How can problems be anticipated and prevented? 3. How should problems be corrected once they are discovered to prevent compounding the problem by acting hastily or conversely waiting too long to respond? The lander system designers provided capabilities within the landers which allowed these questions to be answerable. The most important capabilities derived from the redundancy requirement of the Viking Project Specification which states, "No single malfunction shall cause the loss of all data return from more than one science investigation or the loss of all engineering data." As a result of this directive, the landers were designed to have par- allel or block redundancy, or alternate functions for every critical component. In addition, fuses and automatic safing modes were provided to isolate malfunctions to a specific component and terminate operations of that component until the VFT could reconfigure the system via com- mands from earth. As a result, the lander could be expected to survive until the corrective action could be taken, which could be no faster than 40 minutes and more typically several days. The remainder of the diagnostic capability required to answer the questions is provided by the lander telemetry subsystem, the operational planning and performance prediction techniques used by the VFT, and the telemetry data analysis conducted by the VFT, which are summarized in the following sections. Figures 1 and 2 are provided for reference in these discussions. Figure 1 shows the lander functional block diamgram, including cruise and entry subsystems. Figure 2 is a sketch of the lander after it is deployed on the surface of Mars. Telemetry/Communications Description The communication links to a lander are depicted in Figure 3. Earth can communicate to the VL through the direct link (S-band) at 4 bits/second. The lander can return data over the direct link at 250, 500, or 1000 bits per second on the "high-rate" subcarrier and at 8-1/3 bits/second 27^ over the low- rate engineering subcarrier. The primary downlink method is the relay link (UHF) through which the lander can transmit data to either of the two orbiters (when they are in view) at 16000 bits/ second. The orbiter stores the data on its tape recorder for trans- mission to earth typically a few hours later. The data is gathered on the lander by the Data Acquisition and Proces- sing Unit (DAPU) . The DAPU either provides the data directly to the UHF or S-Band systems for real-time transmission, or stores it into the Data Storage Memory (DSM) or on the tape recorder for later transmission. All of these modes are controlled by sequences stored in and executed by the Guidance Control and Sequencing Computer (not shown) which issues commands to all other VL components to acquire, transfer and transmit data. The types of data that are collected are shown in Figure 2. The System Test Equipment interface was not used after launch nor the Viking orbiter interface after VO/VL separation. Each of the science instru- ments supply data in frames of various lengths to the DAPU. These data contain measurements pertinent to the instrument hardware status (current, voltage, temperature, mode discretes, etc.) as well as the desired scientific information. No on-board diagnostic processing of these data is done. The GCSC provides readouts of selected portions or all of its memory (up to 18432 twenty-four bit words) when scheduled in its stored sequences. This memory readout contains important infor- mation for status monitoring and diagnostic analysis such as: a count of the number of commands received from earth, the values of all sequencing times and subsystem commands which have been or will be issued until modified by the next earth uplink, all instructions of the flight program, and a table called the Rl table, which is a record of the last 2000 commands and the times that have been issued through GCSC register Rl to all science instruments and the DAPU. The Rl table typically spans the last 3 to 6 days of operation. The engineering data collected by the DAPU for landed operations (exclu- sive of the science and GCSC data) is formatted into two formats (termed Format 4 and Format 5) . Format 5 consists of frames of 768 bits containing 79 different measurements. Format 5 is designed for contin- uous data acquisition during direct links although it is occasionally used at other times and recorded in the DSM or directly to the tape recorder. Format 4 consists of frames of 768 bits and contains 74 measurements. It is collected one frame at a time and is scheduled by the GCSC to provide spot sampling throughout the day between downlinks. The engineering formats (1,2, 2A, 2P, and 3) used prior to landing are continuous formats and were constructed similarly to Format 5 but contained other measurements pertinent to those phases. All engineering formats are time -tagged by the GCSC software clock, (except the cruise Format 1 which used earth received time) . The normal method of data collection used during the landed mission is to store all data except imaging and Format 5 to the DSM. The DSM holds 275 196608 bits and when it fills, the DAPU issues an interrupt to the GCSC causing the GCSC to command the DAPU to dump the DSM contents to the tape recorder. In this way runup and rundown of the tape recorder is minimized. Imaging and Format 5 may be written to the DSM at 8-1/3 bits/second for special purposes. Normally, imaging and Format 5 are put out directly to the UHF or S-Band links or written directly to the tape recorder. GCSC data is always transmitted real-time over the links. Table 1 illustrates the data content of a typical relay link and a typical direct link. During its primary mission a lander would have a relay link every Martian solar day (1 SOL = 24 hours 39 minutes) and a direct link every other SOL. During the extended mission (after November 1976) the typical data return is two relay links and two direct links per lander per week. The Planning/Prediction Process A fundamental prerequisite to proper performance monitoring or diag- nostic analysis of an operating system is to understand thoroughly what the nominal performance will be. Particularly in a system such as the Viking Lander where the visibility into operating characteristics is limited by the small data volume and intermittant data samples, the prediction must be as complete and accurate as possible. Furthermore, if the operating environment is unknown or unpredictable with time, the prediction system must be constructed so that the new information from telemetry analysis may be factored in. The Viking Lander Mission Operations software system was designed according to these concepts. Figure 5 shows the uplink preparation software which is run in order to implement and verify a segment (typ- ically 5 to 10 SOLs) of the lander mission. The events leading up to and the overall strategy employing this software system are discussed in reference 1 and will not be repeated here. However, the simulation and prediction functions used for performance monitoring and diagnostic analysis will be discussed. LSEQ (Lander Sequence of Events) generates a sequence of events at a one-second granularity from user level inputs (1)* (i.e., time and type of pictures to take, duration of relay link, meteorology sample number and duration) . It is a functional simulation of the VL data generation Reference 1: Lee, B. G. and Porter, Dr. J. D. ; Design and Implementation of Viking Mission, AIAA Paper No. 77-270, January 10, 1977. * The numbers in this discussion refer to the numbered points in Figure 3. 276 subsystems and the on-board GCSC software. It provides a sequence of events listing and discrete events file (2, 4, 8) for use by other programs, a detailed prediction** of all data to be returned (10) on each scheduled downlink, and saves a map of the DSM and tape recorder contents . The LSEQ sequence of events is also passed to the COALES (Combined Orb iter Lander Sequencing) program to generate the operational timeline for mission controllers and DSN personnel (9) . LCOM (Lander Command) generates from the LSEQ files a Hamming Coded message file which will be transmitted to the appropriate lander after the verification process is complete. LCOMSM (Lander Command Simulation) is a complete bit-for- bit model of the GCSC and contains the flight program which executes the planned mission period in the ground computer system at JPL. This program produces a listing of all GCSC input/output to the millisecond level which it compares to the LSEQ predicted events (3) , thus verifying that the flight software after receipt of the command file will execute the planned sequence. In addition, it creates a file of what the GCSC memory readout data will contain during the direct and relay links. Power management on the lander is a significant problem because the Radio Isotope Thermal Generators only provide about 80 watts of power to recharge the 4 lander batteries, although the system loads may reach peaks of 400 watts. The LPWR1- LTEMP -LPWR2 program set allows this problem to be analyzed prior to committing the vehicle to sequences which may totally discharge the batteries or thermally stress components. LPWR1 (4) converts the LSEQ sequence of events into a predicted system load (watts vs time) which LTEMP uses to predict lander temperatures at about 400 nodes within its thermal model. Since some components have thermostatically controlled heaters or temperature sensitive efficiency, LTEMP provides the predicted thermal history back to LPWR (6) for the final energy balance prediction. From these programs, the values of all lander temperature measurements and the system voltage/current measure- ments may be predicted. Various mode discretes (tape recorder on, etc.) are also related to these predicts. Two other programs, LDLINK and RLINK, are used to predict direct and relay link communication performance. From the products of these programs, it is possible to predict every command which will be generated on board the VL if the sequence is executed nominally, the amount and type of data vs time is known, the RF signal strength of each is known, and a complete power and thermal **The data type and volume is predicted but the value of each individual measurement is not. 277 history is available. The factors which cause deviations from these predictions are caused by planning errors, changes in the Martian environment, or hardware degradation. It is the task of the downlink analysis people to find these factors. Downlink Analysis Process The downlink data from one of the spacecraft are received at a DSN station and transferred via high speed data lines to the Jet Propulsion Laboratory where it is processed and displayed. Figure 6 illustrates this system. The TLMP program operates on the data real-time and provides decommutated data to the Ground Display video tubes. Real time operations personnel monitor this data for quick- look status as it is received. Important measurements are bounded by alarm limits which will flash a warning if they stray outside of limits so that the monitor person can notify other analysts of problems quickly. For routine analysis, typically the next day, the data is run through DECSET to product prints and plots of all engineering data. DECSET strips out all science data by type and provides files to each science team for analysis using their software. DECSET compares the GCSC memory readout to the predicted memory readout from LCOMSM and generates a list of non-compares which are later analyzed by Flight Software experts for anomalous conditions. The Rl table previously described is provided in this printout and may be compared against the LCOMSM discrete event listing in problem diagnosis. The system data record from TLMP is also run through a program called Mark IV which provides a frame by frame listing of all data received. This listing is useful in checking tape recorder performance and to insure consistency between the planning- LSEQ data prediction process and what the vehicle actually returns to earth . Although it is not shown in Figures 5 or 6, both LPWR and LTEMP are capable of reading files of downlinked power and temperature data to use in comparison with the predicted values. In doing this, the analysts may refine their lander models of these functions to either upgrade accuracy or account for changing Martian environments. This is especially important in thermal predictions since they are affected by the optical depth of the sky, the Mars-Sun distance, time of year, winds, and dust coating on lander surfaces. Laboratory Support Two important laboratories were used during Mission operations for prediction, analysis, and diagnostic purposes,. The first was located at Martin Marietta in Denver and was connected to JPL by high-speed data lines. The facility contained a complete real-time simulation using guidance and control hardware and flight software which was used to verify the landers descent trajectories prior to separation. For 278 cruise and landed operations the Proof Test Capsule (PTC) was used to checkout sequences and analyze unusual hardware modes of operation. The PTC is a complete functional lander using qualified flight hardware, but connected to the ground checkout system. This laboratory was useful in verification of flight software changes and diagnosis of the battery charger failure that occurred during interplanetary cruise. The second laboratory is located at JPL and contains a full-size mockup of the lander with actual flight-type cameras, surface sampler and X-ray, GCMS and Biology soil inlet systems. This lab is still in use for design and verification of all surface sampler sequences. Since the lander views Mars non- linearly through its cameras, it is not a straight forward process to acquire a soil sample at a specific location. The sequence of commands (extend, elevate, azimuth, rotate, vibrate, etc.) that the surface sampler needs can be worked out using this mock- up and the Martian surface as modeled from lander pictures. Then if the real VL surface sampler has a "no-go" either due to hardware malfunction or unexpected environmental factors, a detailed sequence baseline has been established for failure analysis. This laboratory was essential during operations and allowed quick recovery from several surface sampler problems. Analysis of an Anomaly On January 19, 1977, the Lander Performance Analysis Group DECSET analysts, H. Fuquay and J. Murphy, noticed that the VL-2 SOL 131 relay link (which had been processed the previous night) did not contain all the meteorology data that had been predicted. In reviewing the Mark IV listing of all data frames received it was quickly determined that the meteorology instrument had stopped providing data on about SOL 130, 13:42 local lander time. Management was quickly notified and an anomaly meeting convened. At that meeting it was apparent that all the prediction products had not yet been compared to the downlink received so the meeting was adjourned after an action item list was generated aimed at localizing the problem, with a status report meeting scheduled for 9 a.m. the next day, January 20. In the meantime, the SOL 132 relay link DECSET run had been completed and found to contain no new meteorology data. Figure 7 shows a timeline of VL-2 downlinks during this period. However, in reviewing the LSEQ-LCOMSM predictions versus the data received from the Mark IV listing and the Rl table from the GCSC, R. Bender, the LPAG Deputy Chief, found the cause of the problem in a few hours of analysis and discussion with subsystem experts. Figure 8 shows the predicted versus actual timeline. The actual downline confirmed the predictions exactly until 14:03:49 when the DSM filled and was dumped to tape. This dump occurred about 30 minutes earlier than pre- dicted. It was followed by a Format 4 to DSM which is always done 279 automatically by the flight software to mark the start of data in the DSM since all the science data is not time-tagged. Then the low rate format 5 to DSM continued and all further sequencing was normal except no meteorology data was collected, indicating that either the instru- ment had failed or its data collection electronics was struck. Using the Mark IV listing which provides a bit -by-bit accounting of data received, R. Bender was able to construct the timeline of Figure 9. This figure shows that when the Format 5 to DSM at 8-1/3 mode was started, the DSM contained 189288 bits. At 190464 bits the DSM/DAPU generates an interrupt to tell the GCSC "I am nearly full, dump me to tape". However, from discussions with the LPAG telemetry subsystem analyst, C. Seese, it was found that the DAPU would not generate the interrupt until the mode in process completed (at 14:03:48 on the time- lines) . In the meantime , the GCSC had received an interrupt from the meteorology instrument at 13:57:33 saying "my buffer is full, dump me to the DSM". However, the GCSC says to itself, "the DAPU is busy putting Format 5 in the DSM, I will hold the MET request until Format 5 is done". At 14:03:48 the GCSC then terminates the Format 5 to DSM mode by issuing the command to dump MET to DSM. Milliseconds later, the GCSC received the request from the DAPU to dump the DSM to tape which the GCSC issued at 14:03:49, allowing the standard 1 second for the MET data transfer to complete. However, the MET data frame is long enough to overflow the DSM past its upper limit, 196608 bits, at which point the DSM automatically puts the DAPU to standby, terminating the data shift pulses to the MET instrument before the 5156 bit frame is totally transferred. Since the MET instrument resets its "data ready" flag after the transfer is complete, it is hung-up waiting for a shift pulse which never comes. The GCSC never receives another MET data ready to command a MET dump to DSM, so MET is forgotten. The hardware design characteristics were confirmed through discussions with MET instrument engineers and DAPU/DSM engineers and the flight software operation reviewed with the Lander Command and Sequencing analysts, S. Lowrie and J. Kerekes, so R. Bender's diagnosis was complete. The next step was to design the recovery plan. In this condition, the MET engineers and LPAG power analysts confirmed that only 3 to 6 watts extra power was being consumed. This level was to small to be of concern to loss of battery capacity and it was confirmed by comparison of LPWR predicts to downlink measurements on the SOL 132 relay link that an additional power load of this size was present. The temper- ature increase was too small to be detectible by LTEMP/ downlink comparisons. There was no critical concern to lander safety, however, the meteorology science team wanted to reinitiate data collection as soon as possible. By this time it was Thursday morning, January 20, and too late to command the lander via the SOL 134 direct link. It was decided to build another separate command fild for the SOL 137 uplink already scheduled for Saturday, January 22. The solution was to simply 280 issue via real-time command, the GCSC to DAPU command to dump MET to DSM. Low rate format 5 to DSM data collections which did not have a DSM to tape dump commanded immediately before them were eliminated to prevent recurrence of the problem. This anomaly was an interesting test of the VL data monitoring and analysis system because it was very subtle and used most of the predic- tion tools. Normally, a problem such as this would have been caught by the real-time analyst monitoring the SOL 131 relay link. Unfortunately, due to a problem at the DSN station, a portion of the relay link data was lost between the station and JPL and had to be recalled which usually takes 1 to 2 days. Since the analyst did not receive all the planned data, his analysis of data received versus predicted (total frame counts) was incomplete, but agreed closely enough that he did not suspect a lander problem. The SOL 131 direct link contained a DSM dump which would have shown no MET data, but since it is the first IK data received on the ground in a direct link, it is usually not acquired for the first 5 minutes, and in this instance, worse than that. There- fore the problem was obscured again. It was caught by the DECSET analyst very early in his process of evaluating the SOL 131 relay link results. If he had missed it, it would have been caught by either the meteorology team or the LPAG tape management analyst in later reviews. The second interesting aspect of this problem is that it is not a hard- ware failure (the hardware has operated perfectly since the command fix) but a system design deficiency which excaped detection through extensive pre-launch testing and never occurred in mission operations for 175 SOLs of VL-1 and 130 SOLs of VL-2 operations. It only occurs when the DSM is used for continuous data collection (normally imaging or format 5) and would not be a problem if the DAPU issued the GCSC interrupt immediately rather than at the completion of the mode. Since it is inherent in the hardware design, the Lander Command and Sequencing team has had to work around it by taking the following actions: 1. LSEQ has been upgraded to model all data. The reason the DSM dump occurred early was that the Format 5 to DSM during ranging in the direct link was not modeled by the SOL 130 version of LSEQ. 2. Whenever low rate continuous data is put into the DSM, it is dumped just prior to initiating this mode and the data volume checked to be less than 190,000 bits. 3. If (2) cannot be done, the science instrument buffers are dumped after completion of the continuous mode to remove any unexpected hangups. 4. The real-time operations personnel have been instructed to check downlink time tags of science data to insure no instrument has 281 stopped. (This cannot be done for XRFS, GCMS or Biology, since they do not time tag their data) . Conclusions and Recommendations This paper has attempted to illustrate how diagnostic failure analysis is applied to the Viking Landers operating on the surface of Mars over 220 million miles from earth using a very constrained telemetry return. At the date of this writing, Lander 1 has operated for 296 SOLs and Lander 2 for 252 SOLs (1 SOL = Martian day of 24 hours 39 minutes) and achieved all major science objectives planned before landing. Between them, the landers have experienced 23 major problems all of which were detected quickly by the Viking Flight Team and all of which were corrected in time spans of 2 to 7 days, resulting usually in minor data loss or delay of plans. The system used to detect these problems and continue mission operations is a complex ground and flight hardware and software system summarized herein which requires expert and dedicated people to operate. The Viking Flight Team was made up of this type of people, most of whom had been hand-picked from the design development and test teams who put these systems together. Most projects having a similar remote diag- nostic task will not have this luxury, either due to cost, or because normally the system designers are not involved with the operation of the system after it is delivered. Therefore, it is of interest to discuss the strengths and weaknesses of the Viking system. First, in order to do remote performance monitoring and diagnostic analysis an accurate prediction system is required. This was achieved on Viking through use of ground computer simulations which ultimately performed very well. However, the development of this system was more expensive than need be because of the following factors : 1. The flight software for the lander computers was developed as an independent program until launch (when it came under control of the Flight Team) making the command software (LSEQ, LCOM/LCOMSM) difficult to design. 2. The data system was not standardized, and was difficult to obtain modeling information for. All science data had different formats, different frame lengths, and several did not time tag their data. This complicated both the prediction and downlink analysis software and caused many uncertainties in analysis. 3. Some of the science teams had not designed their software require- ments to include all the modes of operation they intended to use so the ground software could not model these modes, necessitating changes "on the fly" during operations. 282 Second, in order to optimize data return for science, engineering data return was kept at the minimum required for nominal performance moni- toring. This could have been alleviated by making more use of dual channel transmission and recording, one low rate channel for engineering and high rate channel for science. The Viking orbiters made much better use of this approach than the landers. The Viking lander data system, on the other hand, had a near optimal design for maximization of data return over constrained downlink opportunities. After LSEQ modeling was shown accurate enough, tape recorder content could be predicted to less than 15 seconds out of 40 minutes record time at 16000 bits/second. Therefore, the downlink could be designed to contain a minimum of "margin" of filler data. This was accomplished by buffering each instrument so that it did not have to be serviced immediately when it had collected data, by using the DSM as temporary storage and the tape recorder as long-time storage. Data could be saved on the tape for several downlinks and played back later; however, the engineering data for status monitoring would be delayed with it. This collection technique allowed many experiments to run and collect data simultaneously with a minimum of interference and conflict resolution. It is a very useful technique for diagnosis to record on-board all commands (and their times) issued by the on-board computer. This tech- nique was partially developed on the VL but could have been enhanced by including all commands and doing an automatic compare of predict vs actual results. Finally, it is desirable to automate the prediction versus actual data comparison as much as possible to reduce operation at manpower and increase the probability of detection of malfunctions early. The Viking Lander process was semi -automated but required expert review of all results. Standardization of the data system and the ground processing software files and timing could have allowed simpler compare processes with more automatic discrepancy detection. 283 TABLE 1 TYPICAL RELAY AND DIRECT LINK CONTENTS A. SOL 281 Relay Link (VL-1 to VO-1) Frames New Data Format New Old Wrap Percent of Total 4 444 58 43 2 5 630 19 21 2 SSCA 41 .02 BIOL 21 2 1 .2 MET 660 113 22 16 XRFS 1134 118 50 2 IMAGE 1 3549 55 IMAGE 2 1200 18 GCSC 296 4 Mission Period Recorded SOL 276/13:30:00 to SOL 281/11:50:00 Efficiency = New Bits/ (Total Bits) = 96% Total Data Returned 22611333. bits Link Duration 25 minutes 52 seconds B. SOL 283 Direct Link Format New Old Wrap 4 37 46 MET 32 64 5 (Hi) 37 46 GCSC 253 5 (Lo) 15 Comment DSM Dump Usually Lost DSM Dump Usually Lost DSM Dump Usually Lost 93 Seconds Between Samples Mission Period: Real Time SOL 283/17:20:00 to SOL 283/17:44:00 Total Data Returned 1189376. bits 28^ ABBREVIATIONS bps - bits per second COALES - Combined Lander Sequencing Program DAPU - Data Acquisition and Processing Unit DECSET - Downlink Decommutation Decalibration Set program D/L - downlink to earth from VL DSM - Data Storage Memory DSN - Deep Space Network FOVLIP - First -order Viking Lander Imaging Program GCMS - Gas Chromatograph Mass Spectrometer GCSC - Guidance, Control, and Sequencing Computer HGA - High Gain Antenna IRU - Inert ial Reference Unit JPL - Jet Propulsion Laboratory LCOM - Lander Command Program LCOMSM - Lander Command Simulation Program LDLINK - Lander Direct Link Program LGA - Low Gain Antenna LPWR - Lander Power program LSEQ - Lander Sequence of Events program LTEMP - Lander Temperature program MET - Meteorology Instrument MV - millivolts PCM - Pulse Code Modulation PTC - Proof Test Capsule Rl - Register 1 of the GCSC RLINK - Relay Link program RPA - Retarding Potential Analyzer RTG - Radio Isotope Thermal Generator SOL - Martian mean solar day (24 hours, 39 minutes, 35.25 seconds) TLMP - Telemetry program (Lander real-time) UAMS - Upper Atmosphere Mass Spectrometer UHF - Ultra-High Frequency U/L - Uplink to VL from earth V - volts VFT - Viking Flight Team VL - Viking Lander VO - Viking Orbiter W - watts XRFS - X-Ray Fluorescence Spectrometer 285 PYROTECHNICS SUBSYSTEM Initiators POWER SUBSYSTEM ~1 VO Power VO Control Bio- Shield Power Assembly (BPA) VO Power RTG I ,1 RTG* ICon Power Conditioning and Distribution Assembly (PCDA) trol Power Load Bank Charge Control VLC Power Power * >■ Batteries (4) I J STRUCTURES & MECHANISMS i SUBSYSTEM Bioshield Aeroshell & Base Cover Aerodecelerator Heat Shield Mechanisms Lander Structure GUIDANCE CONTROL SUBSYSTEM n i Pyrotechnic Control Radars Control L Tt7, ERMAL CONTROL SUBSYSTEM Thermal Control Guidance Control & Sequencing Computer (GCSC) Control 4 » Control Valve Drive Amplifier (VDA) Control Inertial Reference Unit (IRU) Control Commands r Control Control Control SCIENCE SUBSYSTEM n Surface Sampler Landed Science Experiment Entry Science Experiments Control Control ; PROPULSION SUBSYSTEM RCS/Deorbit Propulsion Terminal Propulsion & Attitude Control VO Hardline Commands o ~l S Band Communication UHF Communication COMMUNICATION SUBSYSTEM VLC Engineering Data Data Data L. J ;4 ._i Data Acquisition & Processing Unit (DAPU) Tape Recorder Data Storage Memory TELEMETRY SUBSYSTEM L_™ I Figure 1 VIKING LANDER CAPSULE FUNCTIONAL BLOCK DIAGRAM Data to Earth Earth Commands Data to VO Data Hard lined to VO 286 U toO •H 287 %J < «/l as CO •H P>4 288 CO I 03 CD E * C If cow O (3 2 CO O Tape Rcdr a f^ > > a> E o E '5 CO E CO CO < D CO o (3 5) to E > at O O o o CD *-• u to H- k. 3 CO f s n o CO o "• CO oc o X CD 289 < o o CO «/> U O rj O UJ LU a z < o z > CO ir\ a) u 3 too •H § LU Z _ o — CO == S — <-> -D 290 >- O LU U o 5 O ^D £ 0) LU 3 O s toO •H to MM O to OS UJ S on LU CO >- o o 8 Ml 00 291 oo K K cm" CM i m >- O o CVJ CD CVJ U vO o 3C O Ocsj CM i > oo 3 .-« o CO > □ O < CQ co O. > CS "cvj i i OO co> O < CQ rr% a. :> OO uo> >- < O < i— i »- —i _j co5 ^£ D <* CVJ LU O ce> CVJ O F<~ O -J co> *1 O =) Q CO •H CO Q >- r- < l-H Q -A ■A CO CM CVJ >- < o 5 Q CO 292 < > z o u Q UJ SI oo oo FMT5 TODSM AT 8 1/3 ir\ oo •— • i_ O oo 5=25 E1b - 111 z ^^ ^^B O in UJ „ X OO |_ O £ s s £ I— oo H- C OS = 2 I— < O h £ ^ 2 1/1 CO e 5 1 J ? GO Z 1- O 5 c _l K ss CM s ^ 00 i r 1 1- 5 < id > Q Sh UJ . . —j £ s oe ^ — ' o £ on <£_ o Ift ►— ert ^ _ S0*£*1 21*1*1 2I*0*T &f ; €0*T o* : K : CI Z0 : Ifr : 0I OS*F0I oo gi o 1— U_ 1— ««. 21*1*1 21 : W*l M : fiO*l 6fr ! 0)*I o* : fl)*l »«*£! o o o UJ OS Gu OO < < oo 00 0) M ■H K'ZTOl OO CO Q O s oo 6M0H)1 C0 : S0K)1 _J UJ a. s 00 LTV t So o 8^1 3 cjS =3 U_ __ en U- -^ LU O "T as " t= o t: £ a_ S3 | &- h- o ^ , «*£ uJ _1£ ^ o ^ 3d >- Q I— iTi 5^ •H 15 0Q oo oo ©8 c\j ^x 8 en o 29^ M % BS-KWA (REV. 7^73) U.S. DEPT. OF COMM. BIBLIOGRAPHIC DATA SHEET 1. PUBLICATION OR REPORT NO. NBS SP 494 4. TITLE AND SUBTITLE 2. Gov't Accession No. Detection, Diagnosis and Prognosis --Proceedings of the 26th Meeting of the Mechanical Failures Prevention Group, held at the IIT Research Institute, Chicago, 111., May 17-19, 'l977 3. Recipient's Accession No. 5. Publication Date September 1977 6. Performing Organization Code 7. XKXMSKM EDITORS T. Robert Shives and William A. Willard 9. PERFORMING ORGANIZATION NAME AND ADDRESS NATIONAL BUREAU OF STANDARDS DEPARTMENT OF COMMERCE WASHINGTON, D.C 20234 12. Sponsoring Organization Name and Complete Addre . sponsoring Urbanization Name and Complete Address (Street City, State ZIP) Inst. for Mat Is. Res., NBS, Washington, DC 20234; ONR, Arlington, VA 22217; NADC, Warminster, PA 18974; Frankford Arsenal, Philadelphia, PA 19137; FAA, Washington, DC 20591; NASA, Greenbelt, MD 20771; ERDA, Washington, DC 20545 15. SUPPLEMENTARY NOTES 8. Performing Organ. Report No. 10. Project/Task/Work Unit No. 11. Contract/Grant No. 13. Type of Report & Period Covered 14. Sponsoring Agency Code Library of Congress Catalog Card Number: 77-14098 16. ABSTRACT (A 200-word or less factual summary of most significant information. If document includes a significant bibliography or literature survey, mention it here.) These proceedings consist of a group of twenty-four submitted papers from the 26th meeting of the Mechanical Failures Prevention Group which was held at the IIT Research Institute in Chicago, Illinois on May 17-19, 1977. The central theme of the proceedings is detection, diagnosis and prognosis as related to mechanical failure. Papers are presented that discuss oil analysis, signature analysis techniques, new detection, diagnosis and prognosis techniques and equipment, railroad system diagnostics, and land vehicle diagnostics. 17. KEY WORDS (six to twelve entries; alphabetical order; capitalize only the first letter of the first key word unless a proper name; separated by semicolons) Failure detection; failure diagnosis; failure prevention; ferrography; land vehicle diagnostics; oil analysis; railroad system diagnostics; signature analysis; wear 18. AVAILABILITY |Q[] Unlimited I I For Official Distribution. Do Not Release to NTIS (X] Order From Sup. of Doc, U.S. Government Printing. Office Washington, D.C. 20402, SD Cat. No. C13 .10:494 LJ Order From National Technical Information Service (NTIS) Springfield, Virginia 22151 19. SECURITY CLASS (THIS REPORT) UNCLASSIFIED 20. SECURITY CLASS (THIS PAGE) UNCLASSIFIED 21. NO. OF PAGES 296 22. Price $4.50 USCOMM.DC 29042-P74 £11. S. GOVERNMENT PRINTING OFFICEi 1977-240-848/324 There's anew look O" .1 ' / • • • the monthly //y^^^^s, J / magazine of the Nation- ' jb^ ^^^IHaBu^/ al Bureau of Standards. /^ ^^^^ A ^*'" ^ ea * urec ' are special ar- L ^k V / tides of general interest on " lS **- ^j y^ current topics such as consum- N^^w// er product safety and building »^ technology. In addition, new sec- tions are designed to . . . PROVIDE SCIENTISTS with illustrated discussions of recent technical developments and work in progress . . . INFORM INDUSTRIAL MANAGERS of technology transfer activities in Federal and private labs. . . DESCRIBE TO MAN- UFACTURERS advances in the field of voluntary and mandatory standards. The new DIMENSIONS/NBS also carries complete listings of upcoming conferences to be held at NBS and reports on all the latest NBS publications, with information on how to order. Finally, each issue carries a page of News Briefs, aimed at keeping scientist and consum- alike up to date on major developments at the Nation's physi- cal sciences and measurement laboratory. (please detach here) SUBSCRIPTION ORDER FORM Enter my Subscription To DIMENSlONf "IBS at $12.50. Add $3.15 for foreign mailing. No additional postage is required for mailing within the United States or its possessions. Domestic remittances should be made either by postal money order, express money order, or check. Foreign remittances should be made either by international money order, draft on an American bank, or by UNESCO coupons. Send Subscription to: □ Remittance Enclosed (Make checks payable to Superintendent of Documents) O Charge to my Deposit Account No. NAME-FIRST, LAST I I I I I I I I M I M I I I I I I I I I I I I I I I I COMPANY NAME OR ADDITIONAL ADDRESS LINE STREET ADDRESS M I I I I I I I I Ill I I I I I I I I I I ll MAIL ORDER FORM TO: Superintendent of Documents Government Printing Office Washington, D.C. 20402 PLEASE PRINT NBS TECHNICAL PUBLICATIONS PERIODICALS JOURNAL OF RESEARCH reports National Bureau of Standards research and development in physics, mathematics, and chemistry. It is published in two sections, available separately: • Physics and Chemistry (Section A) ^V Papers of interest primarily to scien* ..Vorking in these fields. This section covers a br .<$* .ige of physi- cal and chemical research, wit> ^-^ ^r emphasis on standards of physical measu*- y* , fundamental con- stants, and properties of m r ^* '.^sued six times a year. Annual subscription: D' \ 0>> ^, $17.00; Foreign, $21.25. • Mathematical Sck ac&v Section B) Studies and conr- 4 » .is designed mainly for the math- ematician and .