Research on equipment fault detection based on boundary-based OCC-KELM

In response to the short installation time of active equipment, the lack of various fault samples, the difficulty of obtaining, and the low accuracy of existing algorithms for fault detection, the kernel ELM is used as the basic modeling framework, and the boundary-based threshold selection criteria are used; Combine extreme learning with one-class detection, and propose BB-OCKELM detection model. Based on the kernel ELM, and at the same time integrating the idea of one-class detection, the constraint optimization problem of KELM is realized under the constraint of lp-norm, and the output function expression of a unified one-class KELM classifier is derived; for the convenience of fault detection Implementation, based on BB-OCKELM to define the statistical inspection volume and detection threshold. The proposed method is applied to four common data sets commonly used in the field of machine learning and the fault detection application of a certain type of equipment. The experimental results show that the proposed method is suitable for comparison with SVDD, PCA, OC-SVM, and OC-KELM. For different public data sets, the proposed method can effectively balance missed and false alarms, and significantly improve the accuracy of fault detection when the time cost is equivalent.