Optimization method for low-velocity impact identification in nanocomposite using genetic algorithm

Abstract

Fiber-reinforced plastic (FRP) is prone to invisible damage caused by low-velocity impact (LVI) during service. The structural health monitoring system is of great significance for damage monitoring and maintenance of composite materials. In this study, four fiber Bragg grating sensors were employed to collect the time domain strain signals of composite materials subjected to LVIs. Furthermore, a numerical simulation model was established to rapidly obtain impact signal dataset. The signal arrival time, peak time, and peak amplitude were selected as signal features, and the backpropagation neural network was successfully applied to determine the location and energy of LVIs. To address the issue of peak feature extraction in the strain signal processing, a genetic algorithm-based sliding window peak detection optimization method was proposed, which significantly improved the final prediction accuracy. The experimental results indicated that within a position range of 300 mm × 300 mm, the average positioning error can reach 5.1 mm; and in an energy range of 0.5–1 J, the average energy prediction error can reach 0.030 J. The proposed method achieved accurate identification of the LVI location and energy for FRP.