Incipient Sensor Fault Detection by Directly Monitoring Sliding Window Based Singular Values*

Timely and accurate detection of incipient faults is critical to guarantee the normal operation of industrial processes. Nowadays, complex systems are usually equipped with a large number of sensors, which may be vulnerable to faults due to harsh environments. Statistical process monitoring is commonly used for fault detection purpose. Nevertheless, traditional fault detection methods are not sensitive enough to incipient faults, leading to the occurrence of many missed alarms. In this paper, the incipient fault detection task is achieved by monitoring the changes of sample singular values within a sliding window. Two incipient sensor fault types are considered, i.e. the sensor constant bias fault and sensor precision degradation fault. In addition, the rationale behind this strategy is also theoretically analyzed. Finally, a numerical example and the continuous stirred tank reactor process demonstrate the effectiveness of the proposed method.