An online monitoring measure consistency computing algorithm by sliding window in multi-sensor system

Focusing on the method to dispose of ineffective and erroneous information from certain sensors, an online-inspection algorithm for measure consistency is proposed. Before the data fusion or the establishment of soft sensor model, the consistency measure of original data is calculated and removes those data with low quality and precision. Based on the thought of Mahalanobis distance calculation, this paper, unlike other existing algorithms, proposes an algorithm, regarding the other sensors as a virtual population, to measure the credibility of each sensor from a vertical perspective, in which the correlation of sensors' historical data, selected by sliding windows, is taken into consideration, so that an online monitoring way can be employed to attain more précised credibility sequence of sensors. The outcome of practical experimentation verifies the algorithm to be an effective one.