Machine Learning-Based Investigation of Buckling in Thermoplastic Composite Pipes with Delamination Defects under Bending Loads

Thermoplastic composite pipes (TCP) are increasingly used in offshore engineering due to their high specific strength, corrosion resistance, and fatigue performance. However, delamination defects, which may arise during manufacturing, transportation, and service due to process-induced imperfections and bending loads, compromise through-thickness load transfer. These defects can lead to local wrinkling, buckling, and potential structural failure, undermining both integrity and operational safety. This study develops a three-dimensional finite element (FE) model to characterize the mechanical response of TCP under pure bending with the circumferential delamination. The FE model results are validated against experimental results. The elastoplastic behavior of the thermoplastic materials is incorporated, and an embedded constraint is used to simulate fiber-matrix coupling within the reinforcement layer. Parametric analyses quantify the effects of delamination ply number and width on critical buckling capacity. Based on the FE-generated dataset, several machine-learning models are developed to predict the buckling of delaminated TCP under bending. The results show that these models accurately predict the critical response, offering valuable insights into bending-induced buckling and supporting structural design, process control, and defect-tolerance assessment.