Sensor fault detection with low computational cost : a proposed neural network-based control scheme
Date Issued
September 2012
DOI
10.1109/ETFA.2012.6489628
Abstract
The paper describes a low computational power method for detecting sensor faults. A typical fault detection unit for multiple sensor fault detection with modelbased 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.
File(s)![Thumbnail Image]()
Name
ETFA2012.pdf
Size
271.29 KB
Format
Adobe PDF
Checksum (MD5)
0679538d6e7a757f96a733cccf3f4dd2