Sensor Fault Detection with Low Computational Cost: A Proposed Neural Network-based Control Scheme

The paper describes a low computational power method for detecting sensor faults. A typical fault detection unit for multiple sensor fault detection with model-based approaches, requires a bank of estimators. The estimators can be either observer or artificial intelligence based. The proposed control scheme uses an artificial intelligence approach for the development of the fault detection unit abbreviated as 'i-FD'. In contrast with the bank-estimators approach the proposed i-FD unit is using only one estimator for multiple sensor fault detection. The efficacy of the scheme is tested on an Electro-Magnetic Suspension (EMS) system and compared with a bank of Kalman estimators in simulation environment.