Artificial neural networks and phenomenological degradation models for fatigue damage tracking and life prediction in laser-induced graphene interlayered fiberglass composites

Abstract

Fiber-reinforced polymer matrix composites deteriorate mechanically due to fatigue degradation during cyclic stress. The progressive decrease in elastic stiffness over fatigue life is well-established and investigated, yet many dynamic engineering systems that use composite materials are subjected to random and unexpected loading circumstances, making it impossible to continually monitor such structural changes. LIG can detect strain and damage in fiberglass composites under quasi-static and dynamic loads. ANNs and traditional phenomenological models may assess damage development and fatigue life utilizing LIG interlayered fiberglass composites’ piezoresistivity. Passive experiments monitor LIG interlayered fiberglass composite elastic stiffness and electrical resistance during tension–tension fatigue stress. Electrical resistance-based damage metrics follow similar trends to elastic stiffness-based parameters and may accurately depict damage development in LIG interlayered fiberglass composites over fatigue life. In specimen-to-specimen and cycle-to-cycle schemes, trained ANNs and phenomenological degradation and accumulation models predict fiberglass composite fatigue life and damage state. In a specimen-to-specimen scheme, a two-layer Bayesian regularized ANN with 40 neurons per layer beats phenomenological degradation models by at least 60%, with R² values more than 0.98 and RMSE values less than 10⁻³ . A two-layer Bayesian regularized ANN with 25 neurons per layer exhibits R² values more than 0.99 and RMSE values less than 2×10⁻⁴ when more than 30% of the original data is used in a cycle-to-cycle method. Piezoresistive LIG interlayers and ANNs can correctly and constantly predict fatigue life in multifunctional composite structures.