| Summary: | 碩士 === 國立成功大學 === 化學工程學系 === 85 ===
The extended Kalman filter (EKF) is one of the most popular tools for detecting and diaqnosing faults in chemical processes. Although its effectiveness has been widely recognized, implementation of EKFs is still very limited in the industries. This is primarily due to the fact that EKF estimates are often biased if multiple faults occur simultaneously. In addition, the huge computation demands also prevent practitioners from using the EKF - based techniques in practical application. Since the former shortcoming has already been resolved by Chang and Chen (1995), the prime purpose of this research is to develop systematic strategies to reduce the computation load of EKF. We have contributed significantly in the aspect. Specifically, based upon the precedence order among the parameters and state variables, several simple procedures have been proposed to identify negligible cause - and - effect relationship in the system. The EKF model, i. e. state propagation equations and / or Riccati equations for computing the gain matrix, can then be simplified accordingly. The feasibility of these proposed methods has been verified in extensive numerical simulation studies.
|
|---|
