Monitoring nonlinear system using Bond Graph and PCA method

Abstract

The Fault Detection and Isolation (FDI) method in general contains detecting, locating and also identifying the considered faults take place in the dynamical system. In this paper, the modeling FDI is treated with the Bond graph (BG) model at first of all and then Principal Component Analysis (PCA). However, use the BG method, allow us at the first place to detect all faults which are given from the Diagnostic Bond Graph (DBG) and design the reason of a system dysfunction. In the classical methods using BG, the localization procedure and the fault isolation main are essentially based on the Fault Signature Matrix (FSM) produced by the Analytical Redundancy Relations (ARRs). Furthermore, we present, in this paper, a statistical method to detect and then isolate any anomaly in the system. The main purposes reflected, firstly, the residue generation and fault isolation in a new method using BG model and secondly the PCA method. In this paper, we try to detect and isolate the faults which are presented in the operating mode. Simulations are computed on a six tanks system to validate the approach in MATLAB/Simulink and also to indicate the satisfactory results of the proposed FDI procedure.