1、_ SAE Technical Standards Board Rules provide that: “This report is published by SAE to advance the state of technical and engineering sciences. The use of this report is entirely voluntary, and its applicability and suitability for any particular use, including any patent infringement arising there
2、from, is the sole responsibility of the user.” SAE reviews each technical report at least every five years at which time it may be revised, reaffirmed, stabilized, or cancelled. SAE invites your written comments and suggestions. Copyright 2013 SAE International All rights reserved. No part of this p
3、ublication may be reproduced, stored in a retrieval system or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior written permission of SAE. TO PLACE A DOCUMENT ORDER: Tel: 877-606-7323 (inside USA and Canada) Tel: +1 724-776-497
4、0 (outside USA) Fax: 724-776-0790 Email: CustomerServicesae.org SAE WEB ADDRESS: http:/www.sae.orgSAE values your input. To provide feedback on this Technical Report, please visit http:/www.sae.org/technical/standards/AIR5871AEROSPACEINFORMATION REPORT AIR5871Issued 2008-06 Reaffirmed 2013-09 Progno
5、stics for Gas Turbine Engines RATIONALE AIR5871 has been reaffirmed to comply with the SAE five-year review policy. FOREWORD Gas turbine engine prognostic technologies that are capable of predicting critical component failures or performance degradation rates are expected to improve safety and avail
6、ability, reduce life cycle costs, and optimize the timing of scheduled maintenance intervals. For many critical engine components and subsystems, prognostic approaches are currently being developed that utilize state-of-the-art modeling and analysis technologies. This document introduces and provide
7、s examples of leading prognostic modeling approaches that integrate state-of-the-art analytical and empirical models with component testing and inspection results. Specific examples related to failures of engine fan blades and engine performance degradation are described, along with a representative
8、 range of different technical approaches. The process of prognostic model calibration and verification is also discussed. TABLE OF CONTENTS 1. SCOPE AND PURPOSE 3 1.1 Purpose. 3 1.2 Field of Application 3 2. REFERENCES 3 2.1 Applicable References 3 2.2 Definition . 4 2.3 Potential Benefits of Progno
9、stics 4 2.4 Needed Terminology. 4 2.5 Acronyms 5 2.6 Prognostic Requirements 5 3. EXAMPLE PROGNOSTIC STRATEGIES 6 3.1 Analytical Prediction 6 3.2 Measured and Trend-Based Predictions of Wear and Degradation. 6 4. GENERIC PROGNOSTIC TECHNOLOGIES. 6 4.1 Experience-Based Prognostics. 7 4.2 Evolutionary
10、 Prognostics 7 4.3 Feature Progression and AI-Based Prognostics 8 4.4 State Estimator Prognostics 9 4.5 Physics-Based Prognostics 10 5. PROGNOSTIC IMPLEMENTATIONS 11 5.1 Example: Gas Turbine Engine Blade Prognostics 11 5.2 Example: Gas Turbine Performance Prognostics 15 5.3 On-going Prognostics R&D
11、Efforts . 18 6. NOTES 19 FIGURE 1 EXPERIENCE-BASED APPROACH 7 FIGURE 2 EVOLUTIONARY PROGNOSTICS 8 FIGURE 3 FEATURE/AI-BASED PROGNOSTICS 9 FIGURE 4 STATE ESTIMATOR PROGNOSTICS. 10 FIGURE 5 PHYSICS-BASED PROGNOSTICS . 11 FIGURE 6 MISSION ENVIRONMENT AND DAMAGE ACCUMULATION PROCESS . 15 FIGURE 7 EFFECT
12、S OF WASHING ON EFFICIENCY AND OVERHAUL. 16 FIGURE 8 PROGNOSTIC MODEL VISUALIZATION 17 FIGURE 9 PROGNOSTIC MODEL VISUALIZATION 18 SAE INTERNATIONAL AIR5871 Page 2 of 19_ 1. SCOPE AND PURPOSE 1.1 Purpose This document applies to prognostics of gas turbine engines and its related auxiliary and subsyst
13、ems. Its purpose is to define the meaning of prognostics with regard to gas turbine engines and related subsystems, explain its potential and limitations, and to provide guidelines for potential approaches for use in existing condition monitoring environments. It also includes some examples. 1.2 Fie
14、ld of Application This document seeks to meet the increasing interest in gas turbine engine prognostics. Specifically, the document tries to provide a timely guideline for applying prognostic technologies to enhance the capability of current monitoring and diagnostic systems. Some examples are provi
15、ded that are intended to illustrate different approaches and methodologies. 2. REFERENCES 2.1 Applicable References Bannantine, J., et al., Fundamentals of Metals Fatigue Analysis, Prentice Hall, 1980. Boyce, Meherwan, Gas Turbine Engineering Handbook, Gulf Publishing Company, 1995. Bowerman, Bruce
16、L. and OConnel, Richard T., Forecasting and Time Series, Duxbury Press, 1993. Byington, C. S. et al., “Prognostic Enhancements to Diagnostic Systems for Improved Condition-Based Maintenance”, IEEE 0-7803-7231-X/01, 2002. Engel, S. J., Gilmartin, B. J., Bongort, K., Hess, A., “Prognostics, The Real I
17、ssues Involved with Predicting Life Remaining”, IEEE 0-7803-5846-5/00, 2000. Halford, G. “Cumulative Fatigue Damage Modeling Crack Nucleation and Early Growth” First International Conference on Fatigue Damage, September 22-27, 1996. Hartman, W., and Hess, A., “A USN Strategy for Mechanical and Propu
18、lsion System Prognostics with Demonstration Results” AHS Forum 58, Quebec, Canada, June 11-13, 2002. Kurtz, Rainer, and Brun, Klaus, “Degradation in Gas Turbine Systems” Proceedings of the ASME TURBO EXPO 2000, May 8-11, 2000, Munich Germany. Ioannides, and Harris, “A New Fatigue Life Model for Roll
19、ing Bearings”, Journal of Tribology, Vol. 107, pp. 367-378, 1985. Lundberg, and Palmgren, “Dynamic Capacity of Rolling Bearings”, Acta Polytechnica Mechanical Engineering Series 1, Royal Swedish Academy of Engineering Sciences, No. 3, 7, 1947. Peltier, R. V., Swanekamp, R. C., “LM2500 Recoverable an
20、d Non-Recoverable Power Loss” ASME Cogen-Turbo Power Conference, Vienna, Austria, August 1995. Roemer, M. J. and Kacprzynski, G. J., “Advanced Diagnostics and Prognostics for Gas Turbine Engine Risk Assessment,” Paper 2000-GT-30, ASME and IGTI Turbo Expo 2000, Munich, Germany, May 2000. Roemer, M. J
21、., and Ghiocel, D. M., “A Probabilistic Approach to the Diagnosis of Gas Turbine Engine Faults” Paper 99-GT-363, ASME and IGTI Turbo Expo 1999, Indianapolis, Indiana, June 1999. SAE INTERNATIONAL AIR5871 Page 3 of 19_ Roemer, M. J. and Kacprzynski, G. J., “Development of Diagnostic and Prognostic Te
22、chnologies for Aerospace Health Management Applications,” IEEE Aerospace Conference, Big Sky, Montana, March 2001. SAE Aerospace Technical Report Style Manual, Technical Standards Division, SAE, April, 1994. Socie, D., “A Procedure for Estimating the Total Fatigue Life of Notched and Cracked Members
23、”, Engineering Fracture Mechanics Vol. 11 pp. 851-859, Pergamon Press Ltd., 1979. Sines, and Ohgi, “Fatigue Criteria Under Combined Stresses or Strains”, ASME Journal of Eng. Materials and Tech., Vol. 103, pp. 82-90, 1981. Yu, and Harris, “A New Stress-Based Fatigue Life Model for Ball Bearings”, Tr
24、ibology Transactions, Vol. 44, pp. 11-18, 2001. 2.2 Definition Prognostics is the ability to predict the future condition of a component and/or system of components. For the purposes of gas turbine engine prognostics, this definition is often further described in terms of hard failures of components
25、 or condition/degradation of performance related problems. These are further defined as follows: Failure Prognostics: Failure prognostics is focused on the prediction of damage state or failure rate of a component or system of components in an engine. Failure prognostics can either be directly or in
26、directly affected by the diagnosis of specific engine faults, depending on the level of impact the component experiences from the fault condition. Prognostic models are required to project the future condition of these components and/or system of components. Condition Prognostics: Condition prognost
27、ics is associated with the slower degradation (wear related) processes that an engine is exposed to throughout its life. It is usually associated with the diagnosis of fault(s) conditions and the capability of predicting when the symptoms of the identified fault(s) will reach an undesirable state in
28、 which system operation will be adversely affected. Prognostic models are required to project the future “path” of these identified fault(s) on total system performance or reliability. 2.3 Potential Benefits of Prognostics Improved Safety of Flight Improved Operational Availability Reduced Life Cycl
29、e Costs Optimized Maintenance/Inspection Intervals 2.4 Needed Terminology Monte-Carlo Simulation: A process for making multiple analyses of a particular deterministic problem by changing the associated parameters that effect the results/outcome of the process based on the uncertainties that exist in
30、 those parameters, usually in the form of a distribution. Diagnostics: A technique for classifying or isolating a particular “non-normal” condition associated with a system, subsystem or component down to a piece of information that is useful in understanding the off-design condition. SAE INTERNATIO
31、NAL AIR5871 Page 4 of 19_ 2.5 Acronyms AI Artificial Intelligence TBO Time Between Overhaul D&P Diagnostic and Prognostic MTBF Mean Time Between Failures EFH Engine Flight Hours FEA Finite Element Analysis LP Low Pressure (engine section) HP High Pressure (engine section) OEM Original Equipment Manu
32、facturer FOD Foreign Object Damage HCF High Cycle Fatigue LCF Low Cycle Fatigue PDF Probability Density Function PHM Prognostics and Health Management MTB Mean Time Between Inspections 2.6 Prognostic Requirements An integral part of propulsion system prognostics is fault isolation and diagnosis. Fau
33、lt isolation and diagnosis consists of utilizing the available evidence associated with a fault detection in order to identify the location of the fault within the propulsion system. Condition prognosis can then be used to forecast the remaining useful life (the operating time between detection and
34、an unacceptable level of degradation). If the identified fault affects the life of a critical component, then the failure prognosis models must also reflect this diagnosis. For the diagnosis and prognosis of critical failure modes, specific requirements for confidence intervals and severity levels m
35、ust be identified by the developer and/or end user. In general, the fault detection statistics and diagnostics accuracy should be specified separately from prognostic accuracy. To specify fault detection and diagnostic accuracy, the following probabilities should be used: as a minimum: 1. The probab
36、ility of fault detection in terms of false alarm rate and real fault probability. 2. The probability of specific fault diagnosis confidence and severity. SAE INTERNATIONAL AIR5871 Page 5 of 19_ To specify prognostic accuracy requirements, the developer/end-user must first define: 1. The level of con
37、dition degradation beyond which operation is undesirable. 2. A minimum warning level of acceptable operation, given a failure mode or degraded condition. 3. A minimum probability level that remaining useful life will be equal to or greater than the minimum warning level. 3. EXAMPLE PROGNOSTIC STRATE
38、GIES 3.1 Analytical Prediction The total available useful life of an engine component is typically calculated from design parameters and an assumed operational envelope. In the simplest case, operating hours can then be tracked and projected into the future to provide a very crude forecast of remain
39、ing useful life. This is analogous to cycle counting methods such as minors rule and implementations of it such as TAC (Total Accumulated Cycles) that are often used today. More advanced methods are now being considered (discussed in a later example) that use stochastic models to represent failure m
40、ode uncertainties, projected operational parameters, and rare/random events to help improve the prediction of specific failure mode and their propagation for remaining-useful-life. Another example of a simple analytical prediction strategy is assigning a Time Between Overhaul (TBO) index to a compon
41、ent. This number can be derived from fatigue life predictions, under assumed operating loads, for various parts (gears, bearings, etc.) within a propulsion system assembly. The shortest predicted life of all the critical elements will then determine the maximum number of operating hours between requ
42、ired removal from service for overhaul, regardless of the actual condition of the specific component. A more advanced analytical modeling approach will be described in a later section. 3.2 Measured and Trend-Based Predictions of Wear and Degradation Sensors provide a continuous view of the physical
43、data that are directly related to a components performance characteristics. These data can be processed into trendable measures of the component “health” and projections of remaining useful life can be made based upon assumed operational profiles and removal limits of the trendable health measures.
44、An example of this approach is the measurement and trending of pressure head and flow for a positive displacement pump. As the performance parameters deteriorate towards an undesirable level, a remaining useful life estimate is generated by applying tools such as linear regression analysis to the tr
45、ended measurements. In more sophisticated examples, a database of physical measurements can be further processed by techniques such as feature extraction, decision making networks, and rule based expert systems. In the following section, a summary of five of the leading data-driven and model-based a
46、nalytical approaches for performing predictions on remaining useful life or wear/degradation of specific components or engine systems are discussed. 4. GENERIC PROGNOSTIC TECHNOLOGIES Prognostics simply denotes the ability to predict a future condition. Inherently probabilistic or uncertain in natur
47、e, prognostics can be applied to a system or components failure modes governed by material condition or by functional loss. Like the diagnostic algorithms, prognostic algorithms can be generic in design but specific in terms of application. This section will briefly describe five approaches to prognostics. SAE INTERNATIONAL AIR5871 Page 6 of 19_ 4.1 Experience-Based Prognostics In the case where a physical model of a subsystem or component is absent and there is an insufficient sensor network to assess condition, an experience-based prognostic mod
