Multi-impact diagnosis using electromechanical response of CFRPs and artificial neural network

초록

Among various structural health monitoring (SHM) techniques for carbon-fiber-reinforced polymers (CFRPs), self-sensing has gained attention due to its real-time availability, low computational demand, and minimal data acquisition requirements. However, its industrial application remains limited by poor reproducibility and inability to deliver multiple practical outputs. To address these challenges, this study investigates the detection of multiple impacts with random features using simultaneously measured electrical resistance signals. With sliding time window and artificial neural networks (ANNs), key damage parameters-impact occurrence, energy level, and location-were identified with high accuracy and reliability. A self-calibration process was introduced, which defines the effective range of input and trained network, enabling adaptive monitoring in response to structural degradation. It enhances the algorithmic robustness despite uncertainties in loading conditions and sample configurations. Errors in estimating impact energy and location were within 4.08 % and 7.282 mm, respectively, regardless of mechanical properties, stacking sequences, or damage experiences. This seamless and resilient operation from damage alert to diagnoses allows self-decision health monitoring, while its lightweight design, multi-output attainability, and robustness to randomness holds strong applicability to diverse composite structures.

키워드