Enhanced damage detection for noisy input signals using improved reptile search algorithm and data analytics techniques.

The sensitivity of structural health monitoring systems to environmental and operational conditions poses a significant challenge due to their inherent susceptibility to outliers. This paper proposes an effective model-updating-based optimization algorithm that can alleviate the impact of outliers associated with field and operational fluctuations. The proposed method addresses the influence of uncertainties from sources such as white noise, colored noise, and measurement errors, which can introduce outliers in datasets. The approach comprises a hybrid procedure in which a Gaussian smoothing technique is first employed to smooth out measured data to reduce the impact of irregularities. Next, Johansen cointegration is employed for raw data fusion to further enhance the signature of shared patterns. A novel optimization algorithm based on the Reptile Search Algorithm (RSA), named Improved RSA (IRSA), is proposed to solve the objective function based on the concept of mutual information. This algorithm provides a superior solution with much improved computational speed and accuracy compared to RSA. The new hybrid method was validated by several numerical and experimental damage detection studies. Furthermore, it was compared to other state-of-the-art methods described in the literature. The results clearly demonstrate the superior performance of the newly developed method. • Applying Gaussian smoothing to handle high-frequency noise while preserving more information and maintaining functionality. • Introducing novel application of Johansen cointegration as a data fusion technique, effectively handling outliers in FRF signals. • Proposing a new objective function based on Mutual Information, Johansen cointegration, and Gaussian smoothing. • Proposing a new and robust optimization algorithm based on the concept of the Reptile Search Algorithm (RSA), named Improved RSA (IRSA). • Demonstrating the robustness of the proposed method with numerical and experimental evaluations. [ABSTRACT FROM AUTHOR]