ABSTRACT To address the limitations of existing methods for delamination monitoring in carbon fiber reinforced plastic (CFRP), which struggle to balance accuracy, stability, time consumption, and cost, this study proposes an improved method based on a chaotic logistic mapping multi‐layer extreme learning machine (logistic‐MLELM). This method extends the original extreme learning machine with a multi‐layer structure for enhanced learning ability and introduces a chaotic logistic mapping‐based mechanism for stable pseudo‐random input weight generation. Input data, including force, torque, bending moment signals, and process parameters, undergo time‐frequency domain feature extraction, followed by feature selection and dimensionality reduction using the maximal information coefficient (MIC) and the kernel principal component analysis (KPCA). Logistic‐MLELM is then employed for training to develop the delamination prediction model. Experimental validation indicates that the multi‐layer structure of logistic‐MLELM improves accuracy by at least 9.09%, while the enhanced input weight generation method stabilizes the predictive performance within the high‐accuracy range. Comparative experiments with other machine learning methods reveal that the proposed method achieves higher accuracy, stability, and a faster runtime.
Yao et al. (Thu,) studied this question.