Propulsion Gas Path Health Management Task Overview Donald L. Simon NASA Glenn Research Center Propulsion Controls and s Research Workshop December 8-10, 2009 Cleveland, OH www.nasa.gov 1
National Aeronautics and Space Administration NASA Program / Project Structure Aviation Safety Program Conduct long-term, cutting-edge research that will produce tools, methods, and technologies to improve the intrinsic safety attributes of current and future aircraft. Overcome safety technology barriers that would otherwise constrain full realization of the Next Generation Air Transportation System. Integrated Vehicle Health M Management t Project P j t Develop validated tools, technologies, and techniques for automated detection, diagnosis and prognosis that enable mitigation of adverse events during flight. www.nasa.gov 2
Propulsion Gas Path Health Management Background: Based upon parameter Actuator interrelationships inherent within the Commands gas turbine cycle. Provides performance trend Control monitoring and fault diagnostics. Logic Complicated by coupling between deterioration and fault effects. Conventional Architecture: Onboard diagnostics via built-in-test (BIT) and range and rate of change checks. Ground station provides trend monitoring and additional diagnostic functionality. Emerging Onboard Model-Based Architecture: Enabled by increased avionics processing capabilities. Provides real-time continuous monitoring of engine health. Fault Detection & Isolation Logic Fleet-wide Trend Monitoring & s Measurements Onboard Model & Tracking Filter FADEC On-Board System Transmission of Snapshot reports and FADEC fault codes Ground Station Ground-Based System Gas Path Health Management Architecture www.nasa.gov 3
Propulsion Gas Path Health Management Task Technology Portfolio Enables continuous real-time monitoring of engine health Onboard Real-Time Methods Benchmark Problems & Metrics Gas Path HM Standard benchmark problem and metrics publicly available to the EHM community Enables the estimation of unmeasured engine performance parameters for diagnostics, prognostics, and controls applications Onboard Model-Based Performance Estimation Systematic Selection A holistic approach towards diagnostic system design Safety emphasis: reduce propulsion system stem malfunction plus inappropriate crew response (PSM+ICR) accidents www.nasa.gov 4
Systematic Selection Strategy (S4) for Gas Path s Modular architecture tailored to the diagnostic application of the end user Provides a systematic evaluation of candidate sensor suites relative to the diagnostic requirements Iterative Down-Select Process Candidate Suites System Model Health Related Information Suite Merit System Simulation Candidate Selection Complete No Yes Down-Select (Genetic ) Knowledge Base Final Selection Collection of Nearly Optimal Suites Statistical Evaluation System Model d l Suite Merit i h Optimal Suite Application Specific Non-application specific Systematic Selection Strategy (S4) Architecture www.nasa.gov 5
Onboard Model-Based Performance Estimation Actuator Commands Measurements Adaptive onboard engine model enables the real-time estimation of engine performance Control Logic Fault Detection & Isolation Logic On-Board Onboard Model & Tracking Filter FADEC NASA GRC developed methodology for tracking filter design has been found to provide improved estimation accuracy Model-based Performance Estimation Architecture Thrust estimation accuracy comparison (conventional vs. optimal model tuning parameters) www.nasa.gov 6
Gas Path Benchmarking Problems and Metrics NASA is developing standard benchmarking problems and evaluation metrics to enable the public comparison of candidate aircraft engine gas path diagnostic methods Propulsion Method Evaluation Strategy (ProDiMES) Problem emulates ground-based processing of snapshot measurements collected in flight Includes an example solution and evaluation metrics Based on the C-MAPSS Steady-State turbofan model Publicly available at the NASA GRC Software Repository (https://technology.grc.nasa.gov/software/) ProDiMES Graphical User Interface Transient Test Case Generator Transient Test Case Generator Problem emulates in-flight processing of streaming measurement data Includes an example solution and evaluation metrics Based on the C-MAPSS turbofan model Not currently publicly available, but a future software release is planned Desired Engine Flight Profiles Desired / Random Engine Faults User Developed Engine Engine History Creator Fault Generation and Simulation Trainer/Tester Benchmarking/ Post Processing Nominal Engine Data Nominal + Faulty Engine Data Fault Detection and Classification Results Transient Test Case Generator Architecture www.nasa.gov 7
IVHM Gas Path s Onboard Real-time Methods Data driven Fault Detection & Symbolic System Identification in Aircraft Gas Turbine Engines Penn State NRA Applies the theory of symbolic dynamic filtering (SDF) for time series data analysis Leverages changes in system dynamics (magnitude and time scale) to diagnose faults. Anomaly measure (i.e., deviation from nominal behavior) µ = d(p, p ref ) 0.70 0.60 0.50 0.40 0.30 0.20 0.10 0.00 β 10 8 α χ 6 δ 4 2 φ 0 ε γ 1 0.5 η 0 05 0 0.5 1-0.5 05 Current pattern p -1-1 -0.50 Reference pattern p ref Current Distribution 0 1 2 3 State Probability Histogram 0.70 0.60 0.50 0.40 0.30 0.20 0.10 0.00 Reference Distribution 0 1 2 3 State Probability Histogram Symbolic Dynamic Filtering φ χ γ η δ α δ χ Symbol Sequence Finite State Machine (Hidden Marko ov Model) Perron Frobenius operator (i.e., state transition matrix) Unified Nonlinear Adaptive Approach for Detection and Isolation of Engine Faults Impact Technologies Phase II SBIR A unified framework for diagnosis of sensor and component faults An approach to dealing with nonlinear engine models and faults Control inputs faults Engine degradation A bank of nonlinear adaptive fault diagnostic estimators residuals Transferable Belief Model (TBM) based residual evaluation decision Nonlinear adaptive fault diagnostic method Nonlinear Adaptive Architecture www.nasa.gov 8