A support vector machine approximation model optimization method based on k-fold cross-validation method

Abstract

The present invention discloses a K-fold cross validation method based support vector machine (SVM) approximate model optimization method. The method comprises: establishing a rebound prediction digital model, using the Latin hypercube sampling springback prediction to obtain sheet movement displacement, using the z-score normalization method to preprocess the sample data, and finally using the K-fold cross validation method to optimize SVM nonlinear regression approximate model parameters. According to the method disclosed by the present invention, the SVM parameters are optimized through theK-fold cross-validation method, after the optimization is completed, the optimized results are as follows: c=1.6245, g=4, and mean square error mse=0.8066, and a 3D view and a contour view of the optimized results are obtained.

Description

technical field [0001] The invention belongs to the field of complex bending forming technology of SGCC sheet metal, and in particular relates to a support vector machine approximation model optimization method based on a K-fold cross-validation method. Background technique [0002] With the rapid popularization and renewal of automobiles in the field of life and production, the window lifter is an important movable part to adjust the automobile window, and its assembly performance and performance directly determine the stability and smoothness of the window movement. It also greatly affects people's experience. Compared with other large panels and structural parts, the structure of the window lift plate has the characteristics of complex shape, large curvature change, and high assembly size requirements. In order to obtain high-quality parts in the actual production process, it is necessary to Reduce the springback of parts as much as possible to avoid the reduction of dim...

AI Technical Summary

Problems solved by technology

Once it comes to special-shaped parts with complex shapes and many features, the analytical solution method is useless

At the same time, some studies have found that the theoretical calculation values ​​of springback under test conditions are all too large, and the average relative error is as high as 88%. Therefore, it is far from feasible to predict the springback of parts only by analytical calculation of theoretical formulas, and even Will give wrong prediction results and mislead practical applications

Images