A cost-sensitive early fault detection method for industrial big data based on graph semi-supervised

Abstract

The present invention provides a cost-sensitive early fault detection method for industrial big data based on graph semi-supervision, and relates to the technical field of fault detection and diagnosis. This method first collects industrial process data, uses the graph semi-supervised label propagation method to update the labels of unlabeled data, and delineates suspected early faults; then performs cost-sensitive Bayesian classification of suspected early fault points to complete the analysis. Secondary update of data point labels for suspected early faults; finally, an EDC‑SVM classifier is established to conduct online fault diagnosis of industrial processes. The cost-sensitive industrial big data early fault detection method based on graph semi-supervision provided by the present invention takes the minimization of diagnosis cost as the fault diagnosis goal, classifies suspected early faults, and solves the problem of high misclassification costs in industrial fault detection. At the same time, while ensuring classification accuracy, the cost of misjudgment in fault detection is reduced and the safety of the industrial process is improved.

Description

technical field [0001] The invention relates to the technical field of fault detection and diagnosis, in particular to a graph-based semi-supervised cost-sensitive early fault detection method for industrial big data. Background technique [0002] With the rapid development of modern industry, the production equipment in modern enterprises is increasingly large-scale, continuous, high-speed and automatic. The structure and composition of equipment are very complicated, the production scale is very large, and the connection between various departments is also particularly close. The actual production process is linear, nonlinear, time-invariant, time-varying, etc. For the characteristics of different production processes, different fault monitoring methods should be selected, so as to effectively detect faults. [0003] Traditional classification algorithms usually aim at minimizing the global classification misclassification rate, and assume that the misclassification costs ...

AI Technical Summary

Problems solved by technology

Due to the different degrees of harm, the cost of misdiagnosis of equipment is not equal, and the cost of safety hazards and economic losses required to misdiagnose the fault state as a normal state is often greater than the cost of the opposite situation.

In addition, since the acquisition of fault samples is at the cost of a certain degree of equipment damage, the number of fault samples will be much less than that of normal samples. The conclusion of the diagnostic method is more inclined to the judgment of the normal state, and cannot effectively avoid the loss caused by the fault

Therefore, the goal of minimizing the misclassification rate cannot meet the actual fault diagnosis requirements.

Images