Structural Damage Detection Using Mutual Information and Improved Reptile Search Algorithm for Fused Smooth Signals Affected by Coloured Noise

Abstract

Structural health monitoring (SHM) faces a significant challenge in accurately detecting damage due to noise in acquired signals in composite plates, which can adversely affect reliability. Specific noise reduction techniques tailored to SHM signals are developed to tackle this issue. Gaussian smoothing proves effective in reducing noise and enhancing signal features, thereby facilitating the identification of damage-related information. Optimization algorithms play a crucial role in damage detection, especially when integrated with smoothing and fusion techniques, as they provide optimal solutions to SHM challenges. A model-updating-based optimization algorithm is proposed for detecting damage in structures using condensed frequency response functions (CFRFs), even in the presence of various types of noise and measurement errors. The CFRF signals are first smoothed using an optimized Gaussian smoothing technique as part of the proposed method. Then, the proposed methodology integrates diverse smoothed signals using a raw data fusion approach, including those from different excitations, frequency ranges, and sensor placements. Fused smoothed signals are then fed into a new objective function, incorporating mutual information (MI) and Gaussian smoothing to mitigate correlated coloured noise. The proposed objective function also introduces a hyperparameter tuning of Gaussian smoothing to enhance its performance. Optimization via the improved reptile search algorithm (IRSA) updates the objective function, optimizing damage and smoothing parameters. The hybrid method detects damage in numerical composite laminated plates with different layers and boundary conditions, demonstrating its effectiveness as an SHM technique. Comparative evaluations of other state-of-the-art methods show that the proposed method outperforms its counterparts, making it a promising damage detection approach to address the noise challenge in the SHM field.