Sensor fault detection and localization using decorrelation methods

Abstract This paper presents a computer-based fault supervision method for the detection, localization and identification of ‘small’ sensor failures like biases or drifts, independent of parameter variations in the process to be controlled. The algorithm is based upon analytic redundancy methods and uses a mathematical model of the process to compute estimates of the measured quantities. These are compared with the measurements, so that the resulting residuals contain the interesting faults of the sensors, changes of process parameters and parts caused by an inaccurate model. On the assumption that the latter have similar or correlated effects on all or many of the residuals, the model faults are suppressed by an adaptive decorrelation method. Statistical tests eliminate bad data and give an alarm if a faulty instrument is detected. A recursive least-squares algorithm estimates the size of the sensor fault and specifies it as a constant bias, time-variable drift or scale-factor deviation. Some results of applications are given in this paper for the supervision of temperature measurements of a steam generator used in power stations.