Method for constructing body-in-white spot welding deformation prediction model based on graph convolutional network

Abstract

A method for constructing a body-in-white (BiW) spot welding deformation prediction model based on a graph convolutional network (GCN) includes: 1) acquiring a welding feature and 3D coordinates of a spot weld to form an eigenvector and extracting designed 3D coordinates at each 3D coordinate measurement point; 2) encoding, by an encoder, eigenvectors and designed 3D coordinate vectors into hidden space vectors of spot welds and hidden space vectors of the coordinate measurement points, respectively, and constructing a graph topology G through a k-nearest neighbors algorithm; 3) decomposing a Laplacian eigenvector of the constructed graph topology G to acquire frequency domain components, and linearly transforming eigenvalues corresponding to the frequency domain components to construct a multi-layer GCN; 4) inputting the thermodynamic and kinetic information of each coordinate measurement point into a deep neural network and decoding a final deformation at each coordinate measurement point; and 5) optimizing the model.