Steel defect tracing method based on Bayesian network

Abstract

A steel defect tracing method based on a Bayesian network belongs to the technical field of industrial design, the method takes the Bayesian network as a basic model to perform accurate tracing on steel defects, and the method mainly comprises the following steps: 1) performing anomaly detection on sample parameters in a data set to obtain anomaly labels of features or parameters; 2) giving an initial weight matrix of the Bayesian network; 3) adjusting the weight matrix according to the abnormal condition of the sample; and 4) obtaining a Bayesian network structure according to the adjusted weight matrix so as to determine the importance of each parameter in the steel product defect generation process. Aiming at the difficulties of many defects, complex steel production measured data characteristics, difficult analysis of steel defect causes and the like in the steel production process, the method can fully mine the relation between the data characteristics, can conveniently obtain the corresponding relation between sample defects and parameter anomalies under the condition of not depending on priori knowledge, is popular and easy to understand in the whole algorithm process, and is suitable for large-scale popularization and application. And the method has relatively high interpretability.

Description

technical field [0001] The invention relates to the technical field of industrial design, and provides a method for tracing the source of steel defects based on a Bayesian network. Background technique [0002] In the iron and steel production process, the occurrence of defective products brings a lot of pressure to economic cost, labor cost and time cost. Therefore, in addition to finding out the defective product itself, it is also particularly important to trace the cause of the defective product. The steel production process is long and complex, and the occurrence of defective products is often the combined effect of many factors. In addition to judging based on the production experience of workers, auxiliary technical means are also required to further reduce the occurrence of defects. The method we take is to generate a directed graph to represent the dependencies between defects and the factors that cause them by using a Bayesian network set correlation algorithm. At...

AI Technical Summary

Problems solved by technology

At the present stage, Bayesian methods include common plain Bayesian networks, Bayesian networks based on fault trees, and distributed Bayesian networks. Li Lijuan and others stated in the article "Bayesian Network-Based Mechanical Processing Defect Traceability Method", the above-mentioned methods are difficult to deal with the network structure learning problem in the absence of prior knowledge, and there are many limitations in use and poor interpretability

Images