Prediction on quasi-static compression deformation modes of circular tubes based on machine learning

Abstract

This paper proposes a methodology that combines finite element simulation and machine learning to predict the deformation pattern and the number of circumferential lobes of circular tubes. It calibrates the deformation modes with finite element simulation to obtain rich data and then classifies them through various machine learning models. In addition, it conducts refined classification and prediction on the number of circumferential lobes. By the performance of both the training and testing sets of the machine learning model, the random forest model delivers the best performance in predicting deformation modes. The classification accuracy, precision, and recall on the test set were 0.990, 0.937, and 0.987, respectively. The decision tree model demonstrates the best performance in predicting several circumferential lobes. The classification accuracy, precision, and recall on the test set were 0.978, 0.971, and 0.985, respectively. The machine learning model constructed in this study ensures precise classification and prediction of deformation modes for thin-walled circular tubes under given working conditions, making up for the insufficient experience in predicting the number of circumferential lobes. It has a guiding significance for the energy absorption evaluation of thin-walled structures and also guides the design and optimization of thin-walled circular tube dimensions.