Gas Path Diagnostics Based On Sigma Point Kalman Filter For Gas Turbine

As an advanced power machinery, gas turbine has been widely used in aviation, ships, power generation, drivers and other fields. In addition, gas turbine is a complex system composed of highly advanced mechanical engineering systems and tightly coupled electronic control systems that operate in harsh environments. More and more stringent requirements in safety, reliability and economy have been strong drivers for engine health management systems to be developed. The gas path diagnostic system, which is one of most essential parts of engine health management system and foundations of achieving the condition-based maintenance, is used to reduce the cost of engine shut downs and unscheduled engine removal and repair combined with increased safety. This thesis focuses on the researching of application of nonlinear Kalman filter technology to gas path diagnostic system. Component fault diagnostic system was built up by nonlinear Kalman filter and model-based fault diagnostic method.Firstly, discrete self-tuning nonlinear model is developed based on the gas turbine component level model(CLM). Moment inertia, thermal inertia and volume inertia were considered to guarantee the modeling accuracy. Gas path components health parameters are defined to represent the health condition of components and are introduced into CLM, which made CLM to be self-tuning model that able to simulate healthy and degrading gas turbines. Besides, the statistical gas turbine model for kalman filter was established based on component level model with introduction of measurement noise, process noise, and health parameter update equation. Then gas turbine model was developed on Matlab/Simulink. Model validation with operation data on design condition and dynamic performance simulation was performed.Secondly nonlinear Kalman filter algorithm was investigated. To get higher accuracy and reduce computation, Sigma Point Kalman Filter(SPKF) was proposed by applied of simplex sampling algorithm for its high accuracy and low time consuming as well as easy implementation for nonlinear estimation.Influences on stability of diagnostic system, especially two kinds of filtering divergence phenomenon, numerical divergence and real divergence are analyzed. And the Corresponding improvements of Cholesky factorization method and noise estimator were proposed, leading to adaptive kalman filter, to deal with numerical divergence and mismatch of system. Diagnosis system consists of gas turbine represented by CLM and SPKF was developed on Matlab/Simulink.Finally, Evaluation of diagnosis performance of kalman filter on efficiency, accuracy and stability in steady operation as well as unsteady operation condition was done. SPKF designed in present thesis showed excellent performance in degradation detection and tacking in gradual drift and abrupt degradation unless for single fault or multi-fault. In unsteady condition, the filter still kept the high accuracy in degradation detection and precise tracking during abrupt fault process. All the verify case for the performance of diagnosis system, proved a good ability of designed kalman filter to do gas path fault diagnosis of gas turbine, with attractive performance.