Machine learning empowered failure criterion of fiber-reinforced polymer composite

Abstract

Accurate failure prediction is crucial for the reliable design and optimization of Fiber-Reinforced Polymer Composites (FRPCs), as the complex interactions among fiber, matrix materials and interface pose significant challenges to traditional failure criteria. To address these challenges, this study proposes a novel Data-Augmented Sparse Identification (DASI) framework based on 2D plane-stress states that integrates autoencoders, sparse identification, and intelligent safety factor methodologies. This framework leverages test data from 212 specimens to effectively identify and quantify the critical factors controlling the failure of FRPCs, enhancing prediction accuracy and robustness beyond the capabilities of conventional approaches. The inclusion of an intelligent safety factor, which offers a dynamic constraint to the DASI failure criterion, helps enhance safety margins while optimizing material utilization. The validation of the DASI failure criterion through numerical simulations of perforated and notched FRPC laminates demonstrates its superior ability to predict the failure behavior of FRPCs, confirming its potential for practical engineering applications.