Buckling load prediction and peridynamic-based damage detection in composite structures

This study develops a predictive model to estimate the critical buckling load of pultruded composite columns using a finite element–group method of data handling (GMDH) neural network, considering the effect of fiber orientation. Four types of composite columns with different cross-sectional shapes are analyzed. A total number of 243 tests are designed using the Design of Experiments (DOE) method to generate an optimized dataset for neural network training. The predicted buckling loads serve as the starting point for damage analysis, where the peridynamic method is employed to investigate crack initiation, branching, and overall damage evolution in the columns. The proposed approach provides a predictive relationship between fiber orientation and critical buckling load based on the GMDH model. The novel aspect of the proposed model is that it can evaluate the effect of different fiber orientations on the buckling load without the need for additional numerical simulations or expensive experimental testing. The proposed approach also compares the traditional Hashin failure criterion with the peridynamic results, indicating that peridynamics captures damage patterns and crack propagation more accurately compared to experimental results. The results indicate that closed-section composite columns can increase the critical buckling load by approximately 40%. Furthermore, validation against experimental data shows an average error of approximately 9.8% in the predicted critical buckling loads, demonstrating the accuracy of the proposed GMDH neural network model. This integrated framework demonstrates how predictive modeling of buckling combined with peridynamic damage analysis provides an efficient and reliable framework for designing and assessing the structural performance of composite columns.