AI-Based Actuator/Sensor Fault Detection With Low Computational Cost for Industrial Applications

Abstract

A low computational cost method is proposed for detecting actuator/sensor faults. Typical model-based fault detection (FD) units for multiple sensor faults require a bank of estimators [i.e., conventional Kalman estimators or artificial intelligence (AI)-based ones]. The proposed FD scheme uses an AI approach for developing of a low computational power FD unit abbreviated as $i$ FD. In contrast to the bank-of-estimators approach, the proposed $i$ FD unit employs a single estimator for multiple actuator/sensor FD. The efficacy of the proposed FD scheme is illustrated through a rigorous analysis of the results for a number of sensor fault scenarios on an electromagnetic suspension system.