Fatigue life prediction of self-sensing hybrid FRP composites via electrical resistance monitoring and LSTM neural network

Abstract

Fiber-reinforced polymer (FRP) composites are widely used in aerospace, automotive, and civil infrastructure applications due to their high strength-to-weight ratio and durability. Nevertheless, predicting fatigue life remains challenging due to the material’s complex failure mechanisms. This paper presents a novel approach to real-time fatigue life prediction in self-sensing hybrid FRP composites through electrical resistance monitoring. Carbon fiber sensor tows were embedded in a hybrid composite material to enable real-time damage detection. A long short-term memory (LSTM) neural network was implemented to predict the remaining fatigue life using only the resistance data, and importantly, without the need for stress or strain inputs. Fatigue tests were conducted, with experimental results indicating a strong correlation between resistance and cycles to failure. Empirical data was utilized to train the LSTM neural network, with the model accurately predicting the remaining fatigue life across various stress levels. Moreover, the carbon fiber sensor tows provided damage early warning, indicated by a sharp increase in resistance before failure. This study highlights the potential for a simple, low-cost structural health monitoring system with the ability to significantly extend service life through accurate fatigue life prediction.