Wavelet neural network-based method for damage detection of sustainable laminated composite beams

Recent years have witnessed a growing use of natural fiber-reinforced composites in industry, driven by increasing environmental concerns and the demand for sustainable materials. However, the heterogeneous nature and structural complexity of these composites pose challenges for accurate damage detection and monitoring, requiring the development of reliable diagnostic techniques. This paper introduces a preliminary investigation of the Wavelet Neural Network (WNN)-based damage detection framework for sustainable composites. The proposed approach is applied to flax/epoxy cross-ply laminated beams and validated through numerical and experimental investigations. To assess the performance and robustness of the method, the influence of different wavelet functions, i.e. Mexican Hat, Morlet, and Shannon, on the learning capability of WNNs is systematically evaluated. The results indicate that while the Shannon wavelet achieves the lowest mean square error (MSE) for dense numerical data, the Mexican Hat wavelet appears more reliable under sparse experimental data. Overall, the proposed approach effectively identifies defect locations without prior knowledge of structural properties or mathematical models, highlighting its potential for structural health monitoring of sustainable composites.