Improvement of Fault Detection Method for Nonlinear Black-box Systems based on Multi-Form Quasi-ARMAX Modeling
| Authors: | Kumamaru Kousuke, Kyushu Institute of Technology, Japan
Inoue Katsuhiro, Kyushu Institute of Technology, Japan
Tsubouchi Fuyuki, Kyushu Institute of Technology, Japan
Soderstrom Torsten, Uppsala University, Sweden |
|---|
| Topic: | 6.4 Safeprocess |
|---|
| Session: | Fault Detection and Isolation for Nonlinear Systems |
|---|
| Keywords: | Fault detection, Kullback discrimination information, non-linear system,parameter estimation, quasi-ARMAX model, ship propulsion system |
|---|
Abstract
This paper is concerned with improvement of the KDI-based fault detection method so far developed by authors for nonlinear black-box systems. When modeling the system, Quasi-ARMAX model with multi-model structure is used. A fault due to unexpected change in system parameters will appear as the change of identified model. Kullback discrimination Information (KDI) can then be used as the fault detection index to evaluate the distortion in identified model. Several schemes to improve the fault detection performance are proposed, as well as the realization of a kind of fault isolation function based on a recognition approach in the model parameter space. The effectiveness of the method is verified through simulation studies on the ship propulsion system constructed for benchmark test.
