A multi-field coupled data-driven surrogate approach for multiphysical damage diagnostic of energy harvesting composite plates

Abstract

Damage diagnosis in multiferroic composites is essential for energy harvesting systems, where cracking and property degradation significantly impact coupled frequency response and efficiency. This study proposes a multi-field coupled surrogate model for multiphysical damage diagnostics of multiferroic composite plates to address these challenges. These composites exhibit intricate magnetic, electric, and elastic interactions, making damage detection both complex and essential. Unlike traditional finite element updating methods, which are computationally intensive and iterative, the proposed 1DC-BiGRU model offers a more efficient and scalable data-driven alternative. This model integrates convolutional neural networks (CNNs) to extract critical spatial features and Bidirectional Gated Recurrent Units (BiGRUs) to capture complex feature relationships. This architecture excels in processing frequency and mode shape data, enabling robust identification of multiphysical damage patterns. Convolutional layers efficiently reduce data dimensionality while identifying interactions between features. BiGRUs handle relationships between features in a bidirectional manner, mitigating issues like vanishing gradients seen in traditional neural networks. Trained on simulated data generated from Chebyshev finite element analysis and multi-scale plate theory, the model accurately diagnoses damage locations and severities under various scenarios, including both noise-free and noisy conditions. By providing an efficient and robust framework for multiphysical damage detection, this study significantly advances structural health monitoring for multiferroic composite structures.