Structure mode shapes classification using graph convolutional networks in automotive application

Abstract

Classifying vibration mode shapes of a structure in an engineering design cycle can be a labor intensive and repetitive task. Although several methods have been proposed to automatically classify mode shapes, most existing models cannot fully represent mode shapes using both structural and modal information, limiting their application to specific structures. In this paper, we propose a graph convolutional network (GCN) model, which learns to classify mode shapes from a graph perspective. The mode shape graphs were generated using both geometric and modal information derived from the Finite Element Method (FEM). In order to validate the model’s performance, Finite Element (FE) models of Sport Utility Vehicle (SUV) types were developed as representatives of the real-world complex structures. Both quantitative and qualitative assessments are performed to emphasise the advantages of representing mode shapes as graph data. Within the developed dataset, the classification results show that GCN models achieve 100 % precision across diverse geometric configurations and varying input conditions, outperforming existing methods such as Modal Assurance Criteria (MAC) and traditional Machine Learning (ML) techniques.