Medium-thick Plate Stress Analysis Feature Extraction And Defect Prediction Based On Genetic Algorithm

With the continuous improvement of the enterprise information,the computer system of the steel enterprise has accumulated a large amount of production process data after a long time.How to analyze these data is a serious issue which the iron and steel enterprises are facing,since it can not only create higher profits,but also improve the competitiveness of enterprises in the same industry.The internal stress produced by medium-thick plate during production is a key factor which affect the product quality of the plate.Based on the large-scale industrial data accumulated in the process of plate production,this paper studies the data-driven product defect prediction model through using machine learning and data analysis theory.The research has two major aspects:1)The research based on statistical learning stress analysis extraction of plate and prediction of defect classification.We use the optimal subset selection method,heat map and cluster analysis method to select the data which have significant influence on the stress defect of the plate,and then establish the logistic classification prediction model.After that,we use cross-validation method to train and verify the model.Besides,the research use R language as an analysis tool,to extract variable and set up prediction model.2)The research based on the mixing of hybrid genetic algorithm and machine learning to extract feature of plate’s stress analysis and predict defect classification.The 0-1 coding mechanism of genetic algorithm is used to realize the data feature selection.What’s more,logistic classification prediction model and cross-validation method were used to decode the individual coding in genetic algorithm to calculate the fitness value of individual.Meanwhile,the paper uses evolutionary mechanism of genetic algorithm which contains the aid of c#language to realize the optimal selection of data features.Ultimately,the experimental results show that the proposed two feature extraction and prediction models can accurately predict the stress and strain classification in the plate production process.Finally,the research compares the two methods and studying the advantages and disadvantages,so as to solve the problem of medium-thick plate defects prediction better.