Improved Differential Gray Wolf Algorithm Optimized Support Vector Regression Strip Thickness Prediction Method

Strip thickness prediction has important contribution to solve the problems of accurate strip thickness control and saving raw materials. Support vector regression (SVR) is presented to apply for strip thickness prediction, and the key problem to be solved is determining parameters of SVR. A strip thickness prediction method is proposed based on SVR optimized by improved differential gray wolf algorithm (denoted as HGWO-SVR). Firstly, the feature of the strip data is extracted by mutual information calculation method. Next, the differential evolution algorithm is introduced to enrich the diversity of the gray wolf population in gray wolf optimizer (GWO) and avoid falling into the local optimum, and coefficient vector improved gray wolf optimizer (HGWO) is used to balance the ability of global search and local search. Then, HGWO is used to select optimal kernel coefficient sigma and penalty factor C in SVR model. Finally, establish HGWO-SVR model and input the characteristics of strip data into the model to predict the strip thickness. The results state clearly that HGWO-SVR has better predictive performance than GWO-SVR and SVR.