Sensor Fault Diagnosis using Auto-Associative Neural Networks Based algorithm

Auto-associative neural network (AANN) has been recently used in sensor fault diagnostics. In this paper we introduce a new AANN based algorithm named Improved AANN (I-AANN) for sensor single-fault diagnosis. The algorithm is two-aimed approach which estimates the correct value of the faulty sensor as isolates the source of the fault simultaneously. Performance of the algorithm is compared with the so called Enhanced AANN (E-AANN) via computational time and fault reconstruction accuracy and is shown that I-AANN has higher performance, it can isolate the source of fault very quickly and also accurately. A Dimerization process model is used as a case study to test and compare performance of algorithms. Results demonstrate that I-AANN has superior performance.