Impact localization and force estimation for structural health monitoring of composite structures using spray-deposited nano-carbon strain sensors and a neural network

Abstract This study presents a smart structural health monitoring (SHM) framework that integrates stimuli-responsive nano-carbon sensing materials with system-level, data-driven interpretation. The key novelty lies in reframing the intrinsic nonlinearity and history dependence of nano-carbon sensors, typically treated as limitations, as exploitable characteristics within a coupled material–structure sensing paradigm, enabling quantitative impact monitoring without material-level stabilization. Spray-deposited nano-carbon strain sensors (NCSS), exhibiting intrinsic piezoresistive and nonlinear electrical responses under mechanical impact, are deployed as lightweight and scalable sensing layers on fiber-reinforced polymer (FRP) composite structures. Rather than suppressing sensor nonlinearity and signal variability through extensive calibration, the proposed framework leveraged these characteristics at the system level by combining distributed sensing architectures with temporal learning models. A long short-term memory (LSTM) neural network was employed to map history-dependent and anisotropic material responses to physically meaningful impact parameters, enabling simultaneous impact localization and force estimation. A total of 532 impact tests were conducted on FRP panels equipped with a U-shaped NCSS network, and their performance was benchmarked against a conventional linear regression (LR) approach. The LR model exhibited a mean radial localization error (MRE) of 43.6 mm and a mean absolute percentage error (MAPE) of 62.5% in force estimation. In contrast, the LSTM-based framework achieved a localization error of 4.52 mm and reduced the force estimation error to 9.59% MAPE by capturing the nonlinear and spatially coupled sensor responses. These results demonstrate that quantitative impact monitoring can be achieved by treating sensing as a coupled material structure system problem, rather than as a material-level signal stabilization task. Owing to its surface-deployable, low-cost, and scalable characteristics, the proposed framework provides a practical pathway for deployable smart SHM systems in aerospace, automotive, and civil composite structures.