Abnormality detection for the equipment online monitoring with data depth

A novel method for detecting the abnormal machine in the online monitoring system is proposed. Because of the parameters drifting, the traditional method can not effectively find the abnormal point according their performance figure. First, we suggest taking the symmetric transformation for the data about their ideal point, and then take the combination of the symmetric images and their original data as the new sample set. Second, we compute the outlyingness of the depth of the current status with respect to the combined set using data depth, then assess the status according to the value of the depth. Furthermore, we also discuss the method for the functional data. Experimental result shows the effective of the method.