Knowledge-aided Model-based Fault Diagnosis for Dynamic Systems with Time-varying Parameters

点击翻译按钮获取中文摘要

In this research project, we have performed studies on the development of a Knowledgeaided Model-based Diagnosis Method for Dynamic Systems with Time-varying Parameters and obtained following research results.1.Development of a Quick Identification Method and Its Applications to Fault Diagnosis and Adaptive Control.We developed a quick system identification method named as SANQ method, which can estimate on-line time-varying parameters in continuous-time nonlinear systems by using short time record of input and output data. The method is based on the parameter adjustment of simulator describing the objective systems. It can be used for detection and isolation of faults which are caused by unexpected changes in system configuration parameters, and are used for the adaptive control of unknown time-varying systems as well. The effectiveness of the method has been confirmed through the application studies to thermal power plants and servo-driving systems.2.Research on a Knowledge-aided Model-based Diagnosis MethodIn order to decide whether the system change detected by Kullback discrimination information (KDI) is due to a fault or is under the normal operation, a neural network decision system has been established by incorporating knowledge information on the system operating modes and fault characteristics into the learning processes. The effectiveness of the knowledge-aided diagnosis method has been confirmed trough the simulation studies on fault decision of the 2nd order damped oscillator which is controlled by an adaptive way.3.Investigation on the model-based diagnosis method which is robust to modelling errorsTo realize a practical model-based diagnosis method, we investigated on a general modeling (Quasi-ARMAX Models) for nonlinear systems and proposed an robust identification method. We have then established a scheme to robust fault detection for dynamic systems with uncertainty by using a new method for evaluating modelling errors.