Damage Detection and Localization in Three‐Dimensional Woven Carbon Fiber Composites via Bayesian‐Optimized Electrical Resistance Tomography

ABSTRACT Damage detection and localization in carbon fiber composites using electrical resistance tomography (ERT) face significant challenges related to imaging algorithms and hyperparameter optimization. This study introduces a novel framework that combines hybrid regularization with Bayesian optimization to achieve precise damage localization in three‐dimensional (3D) woven composites. The Bayesian optimization process autonomously identifies the optimal hyperparameters for the hybrid regularization algorithms, enhancing the quality of the resulting damage images. We evaluated the performance of a hybrid of L1 and Iteratively Reweighted L2 (L1‐IRL2) regularization algorithms. The findings indicate that the algorithm produces superior reconstruction quality when compared to traditional methods. Furthermore, it successfully locates damages in the less sensitive central region, an area where the traditional algorithms struggled. The proposed approach demonstrated improved robustness to noise; even with a signal‐to‐noise ratio of 20 dB, it maintained a clear background image while accurately identifying damage areas. This investigation validates the effectiveness of the proposed Bayesian optimization hybrid regularization framework for damage localization in complex 3D woven composites.