Intelligent sensor fault detection based on soft computing

Abstract

Sensor fault detection is carried out based on the characteristics of the soft computing techniques; neural network and adaptive neural fuzzy inference system ANFIS. In this paper, a neural network (non-model based technique) and ANFIS has been used for detection and isolation of temperature sensor fault TMP36. The measured states are then compared with true estimated states and if their difference exceeds threshold value, the particular sensor measurement is ignored and replaced by the true estimated state. Residual generation is an essential part of model-based fault detection schemes. This paper develops and implements neural-network and ANFIS based system identification techniques for nonlinear systems with the specific goal of residual generation for fault detection purposes. The two approaches are tested on a temperature sensor model. Performance comparisons of the two neural network and ANFIS are presented.