Mechanical behavior of composite pipe structures under compressive force and its prediction using different machine learning algorithms

Abstract Thanks to machine learning algorithms, the performance of composites with high energy absorption capacity can be predicted with high accuracy rates with a small number of data. The aim of this study is to experimentally and numerically determine the crushing performances of glass/epoxy composite pipe structures under compressive force and to predict their compression behavior with the help of different machine learning algorithms. In the study, the crushing performances of composite pipes (peak force (PF), peak force displacement (PFD), mean crushing force (MCF), specific energy absorption (SEA), and total inner energy (TIE)) were determined for different specimen thicknesses, specimen lengths, mesh sizes, numbers of integration points, diameters ( D ), and compression directions (axial and radial). Additionally, the maximum strength values of composite pipes under force were estimated with the help of Linear Regression (LR), K-Nearest Neighbors (KNN), and Artificial Neural Networks (ANN) machine learning algorithms. The data taken from the ANN algorithm were found to be more reliable in estimating the PF and TIE values, with an accuracy rate of 92 %. When determining the MCF value, it was found that the data obtained from the LR algorithm was more reliable than other algorithms, with an accuracy rate of 80 %.