A NLPCA hybrid approach for AUV thrusters fault detection and isolation

Abstract

The objective of this paper is to address the problem of Fault Detection and Isolation (FDI) on thrusters of an over-actuated Autonomous Underwater Vehicle (AUV) under on/off abrupt faults. The goal is pursued through Non-Linear Principal Component Analysis (NLPCA), which is the non-linear extension of the popular Principal Component Analysis (PCA). While the Fault Detection (FD) system directly exploits the model-free nature of NLPCA (data-driven approach), the Fault Isolation (FI) is achieved by properly train off-line Artificial Neural Network (ANN). The consistency and robustness of the proposed method is verified in realistic simulation.