Sensor fault detection in nonlinear system using threshold estimation

In this paper, an advanced study of fault diagnosis using real data signal system. This study is online and fast application for fault diagnosis sensors. The diagnosis involves two steps respectively: fault detection and fault localisation. An online fault detection approach for an experimental three tanks system is developed. This approach is based on real time signal and statistical analysis. We used the standard deviation and the mean value of several independent experimental repeated in the normal state and under the same conditions for estimating the threshold of fault detection. Then, the acquisition of signal data test in real time is used to validate this threshold estimation. Also, in this research work, a proposed technique of fault detection is implemented and validated experimentally in prototype of three tanks laboratory.