Incipient Fault Detection Based on Bond Graph Method and Different Criteria of Residuals

Incipient faults can cause gradual degradation of functionality of different components in a system. Due to slow varying nature of incipient faults, detection of these faults is difficult and complex, which this difficulty has increased with noises and uncertainty condition in the system. Bond Graph (BG) method is proposed for modeling and Fault Detection and Isolation (FDI) system design in complex and multi-domain processes. Based on BG model Analytical Redundancy Relations (ARRs) are derived and used as residuals. In this paper, by using derived ARRs and different criteria of residuals in integral form, a method for incipient fault detection is proposed. The proposed method not only has great performance in incipient fault detection, but also its performance in noisy condition is considerable. Effectiveness of the method is presented by simulation results.