LSTM-based fatigue monitoring of fiberglass composites using laser-induced graphene

Abstract

This study aims to advance the field o f c omposite m aterial f atigue p rognosis b y e mploying L ong Short-Term Memory (LSTM) neural networks for in-situ damage progression monitoring under random dynamic loading conditions. A unique approach is adopted, wherein Laser-Induced Graphene (LIG) interlayers are embedded into fiberglass c omposites. T hese L IG i nterlayers a re i nnovative s ensors owing t o t heir p iezoresistive properties, enabling real-time measurement of fatigue damage monitoring. The crux of this research lies in applying LSTM neural networks, specifically designed to handle time-series data, making them ideal for modeling the stochastic and unpredictable nature of fatigue loading in composite materials. Contrasting the performance of LSTM with traditional Multilayer Perceptrons (MLP), it is observed that LSTM yields superior prediction accuracy in estimating the remaining useful life (RUL) of LIG interlayered fiberglass c omposites. B y u tilizing predefined electrical resistance damage parameters, the LSTM algorithm correlates the rate of fatigue damage buildup to the impending decline in mechanical performance. This research establishes that integrating piezoresistive LIG interlayers with LSTM neural networks culminates in a robust, reliable, and closed-loop system for structural fatigue monitoring and lifecycle prediction in composite materials subjected to random dynamic loading.