Study On Quality Prediction Method Of Heat Treatment Process Of Copper Strip

In the industrial field,the production quality control of alloy products is an important issue.How to establish a quality prediction model through process data,and feedback the quality information of products online to the processing process for control adjustment timely,has important practical significance for the economic benefits of enterprises.In the production process of the copper strip,the heat treatment is a process that has a great influence on the microstructure of the final product.As the key equipment of the heat treatment process,the air cushion furnace is not easy to apply a traditional regression prediction method to obtain a good prediction result due to the nonlinear non-Gaussian characteristics of the process data and the different sampling numbers of the batch data.Aiming at the quality prediction problem of copper strip in heat treatment process,this thesis studies the applicability based on some traditional methods,and proposes a new method based on mixed data feature extraction and improved local weighted partial least squares to achieve quality prediction.Details are as follows:(1)For the complex industrial characteristics of the mentioned batch process data samples,the industrial data is preprocessed with the proposed mixed data feature extraction method.After the initial analysis of the data set,the raw statistical features of each batch data and the statistics of the slow feature conversion are extracted and combined into a mixed feature vector to represent a batch process data of the copper strip.The high-order statistic is used to solve the nonlinear non-Gaussian property of the data,and the statistical analysis is used to solve the problem of data unequal length.The feature extraction results of the data are provided to the subsequent regression modeling method.(2)For the quality prediction of process data,the fuzzy clustering method is used to analyze and filter the modeling samples,and then the improved local weighted partial least squares method is used to establish the prediction model and output quality prediction result.Compared to the ordinary partial least squares method,this method can give a more accurate prediction result due to the local approximation.For the further optimization of prediction accuracy,the ReliefF weighting algorithm is used to modify the definition of sample similarity,so that the results of local weighting are more reasonable.Using the heat treatment process data of the copper strip in the actual production as the input,the quality index test data of the product as the output,the prediction experiment is completed on the proposed algorithm.The analysis of the experimental results shows that the proposed algorithm for the quality prediction of copper strip has certain effectiveness.