Health Management for In-Service Gas Turbine Engines PHM Society Meeting San Diego, CA October 1, 2009 Thomas Mooney GE-Aviation DES-1474-1
Agenda Legacy Maintenance Implementing Health Management Choosing the Project Enabling Technologies Health Management & Prognosis Process Validation/Verification and Transition Planning Conclusions DES-1474-2
Legacy Maintenance Re-active to faults False removals Unplanned events Fixed maintenance & inspection schedules based on fleet wide statistics Limited predictive capability Legacy Systems Drive Increasing O&S Costs DES-1474-3
Implementing Engine Health Management Objectives: Provide continuous, accurate assessment of engine health Reduce or eliminate inspection tasks Improve mission planning capability Enhance prognostic capability DES-1474-4
Health Management for In-service and New Engine Programs In-Service Programs Large installed base High operating tempo High and growing O&S costs Little opportunity to upgrade New Product Development Programs Advanced control systems Competitive marketplace Technology insertion potential Health Management Technology Adaptable to Meet Customer Needs DES-1474-5
Choose the Right Project Focus on User s needs Clear impact on metrics People adapt to current methods Often unable to perceive a problem Replace margin with knowledge Individual health vs. fleet averages Pick the right team diversity Cross functional team Define success Small, early wins are important Plan transition and implementation upfront Think through all the steps 500 450 400 350 300 250 200 150 100 50 0 Focus on the User s Needs 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 Select those issues that can significantly impact operational metrics DES-1474-6
System Solution to User Needs NOT: Technical specialty IT project AI/Reasoning project Applied statistics project Sensor or control system design Marketing Apply the right technologies to achieve a system solution System Solution To User Needs DES-1474-7
Enabling Technologies Continuous on-board data recording Advanced algorithms/models Multi-purpose PC-based ground stations Open architectures High speed data transfer Data warehousing Digital architecture Diagnostic instrumentation Dynamic view of technology Avoid reasoning that something will never be possible because all previous attempts have failed DES-1474-8
Health Management and Prognosis Process SENSE DIAGNOSE PROGNOSE DECIDE Establish state awareness via engine and aircraft sensors and systems Determine the ability of a component to perform its function. Detect and isolate faults with a high degree of accuracy and a very low false alarm rate. Predict remaining useful life. Use the current state and state transition vectors to model fault progression Engine control system, aircrew, maintenance, mission planning, logistics, depot, etc. DES-1474-9
Sensing Evaluate capability with the existing sensor set Leverage sensors used for other purposes with caution Models can help Models can expand the sensor set via virtual sensors Consider retrofit issues when evaluating new/additional sensors Cost/weight/reliability Interface requirements Demand that new sensors buy their way on Measurement Cost Benefit Valve open/closed $ 70% Valve position (% open) $$ 90% Flow measurement $$$$ 98% DES-1474-10
Data Challenges and Solutions Data accuracy and bias Outliers and outlier rejection logic Noise and filtering logic Data frequency Data stability Gaps and missing data Digitization Data errors Intermittent and noisy signals Analysis of real field data Error detection and correction algorithms Improved fault detection, accommodation and isolation Reduced S/N ratio Data selection/rejection criteria DES-1474-11
Diagnostics Understanding the underlying process allows the condition to be assessed Know when diagnosis is possible Aeromechanical and thermodynamic stability Stable enough? Fix the problem vs. monitoring its health False positives and false negatives Evaluate the rate & consequences DES-1474-12
Fault Testing Real faults vs. test inserted faults Real faults are more complex Examine what really happens field experience Hindsight bias: judging past events as being clearly predictable Relying on testing and results that confirm hypothesis and downplaying contradictory findings» Need for independent outside reviews DES-1474-13
Models Models are based on an interpretation of a physical system Models are often insufficiently robust to cover real world conditions Deployed model robustness must address Field usage Installation effects Operating conditions Environmental conditions Failure conditions Model size and timing Impact on other systems 20000 18000 16000 14000 12000 10000 8000 Mission 15 10 20 30 40 50 60 Aircraft Measured Flight Data Ground Station Thermo-Aero Model Condition Assessment Model Output Despite all the data and models, experts can often determine when something is wrong and intervene based on experience it not all data and models DES-1474-14
Prognosis Prognosis models evaluate and combine usage accumulation to predict the part, subsystem and engine reaching its usable life f(t) 0. 0 0 2 5 0. 0 0 2 0. 0 0 1 5 Time to Removal (hours) for Components in Engine x at calendar time τ ETF Components 0. 0 0 1 Determines the probability of each component being the first to force engine maintenance or removal Determines the expected (average) time when engine will be removed and its uncertainty given the current distributions (forecasts) of individual removal times 0. 0 0 0 5 0 0 5 0 0 1 0 0 0 1 5 0 0 2 0 0 0 2 5 0 0 3 0 0 0 3 5 0 0 4 0 0 0 Time - t Probability and RUL is Recalculated After Each Flight DES-1474-15
Prognosis Output Fleet Management Survivorship plots Percent of parts reaching a limit as a function of operating time Local Unit/Logistics Removal Time Prediction Estimate of time remaining based on planned usage Depot/Service Shop No Build Window Probabilistic assessment of part replacements needed for desired no-build window DES-1474-16
Prognosis Prognosis tool development focused on needs of end user Need for continuous data to develop prognosis models Component-specific prognosis and total engine prognosis (engine performance + component life) Prognosis based upon actual mission data or upon mission simulations Actual Mission Approach Forecast removals based on a continuation of the current mission type -regression analysis based on current data Prognosis Model Based Actual & Simulated Forecast removals based on custom scenarios defined by the user Representative Missions Model cause of condition change Internal or External Usage is individualized Adaptation and learning DES-1474-17
Decisions Engine control system Optimize engine operation Aircrew Critical information on system Maintenance Unambiguous information Mission Planning Engine specific knowledge Logistics Lean support Service Shop Optimal build windows DES-1474-18
Data Management Data storage Memory is inexpensive, retrieval can be expensive Linking or indexing data is key Data reprocessing Is it necessary? Data Mining Learning by mining the existing databases DES-1474-19
Validation/Verification and Transition Efficacy/Accuracy Demonstrate that the health management program meets program goals Assess accuracy and uncertainty Demonstrate prognosis algorithms accuracy Integration and functionality Demonstrate functionality in the target environment Safety Evaluate safety risk compared to baseline system DES-1474-20
Summary Health management can be applied to in-service products Control escalating operating and support costs Use care when selecting the project Focus on users needs Plan transition and implementation upfront Manage the project as a health management solution Use all necessary technologies but don t lose site of the end goal Leverage enabling technologies Evaluate all proposed increases in sensors/systems Design for real data and real systems Understand the decisions that will be made from these systems Transition to target environments DES-1474-21
Thank You DES-1474-22