---

With the development of the technology and increase of human dependency on structures, healthy structures play an important role in people lives and communications. Hence, structural health monitoring has been attracted strongly in recent decades. Improvement of measuring instruments made signal processing as a powerful tool in structural heath monitoring. Wavelet transform invention causes a great evolution in signal processing. Wavelet transform decomposes a signal into several groups based on scaled and translated basic functions. In this study, a novel methodology based on wavelet transform using complex Morlet wavelet has been introduced for system identification. This process includes a multivariable constrained optimization problem for selecting suitable complex Morlet wavelet. Using selected wavelet, modal parameters and flexibility matrix of structure can be estimated properly. Because of small modal participation of higher mode using finite number of modes leads to flexibility matrix with acceptable accuracy. Since damages cause change in structural properties, a damage index based on flexibility matrix has been applied and its performance has been investigated in some structures.

---

Bridge Health Monitoring (BHM) plays a vital role in ensuring the safety, reliability, and long-term performance of bridge infrastructure. This study proposes an ARMA–Wavelet–Artificial Neural Network (AWAN) framework for predicting unmeasured bridge deck acceleration responses from limited sensor measurements. The proposed methodology integrates Auto-Regressive Moving Average (ARMA) modeling for temporal feature extraction, Continuous Wavelet Transform (CWT) for signal denoising, and a feed-forward Artificial Neural Network (ANN) for nonlinear response prediction. The combined framework exploits both spatial and short-term temporal correlations to achieve accurate response reconstruction while maintaining computational efficiency. The proposed framework was validated using three bridge models, including a simply supported beam, a two-span steel grid benchmark, and a scaled single-plane cable-stayed bridge. Prediction performance was evaluated using different statistical metrics under multiple loading scenarios. The results demonstrated excellent agreement between the predicted and measured acceleration responses, with higher prediction accuracy generally achieved at mid-span locations than near the supports, reflecting differences in local structural dynamics. In addition, the framework maintained stable performance under moderate temperature variation, demonstrating its robustness for practical bridge health monitoring applications. The proposed AWAN framework provides an efficient and reliable approach for reconstructing unmeasured structural responses while reducing sensor requirements. Its combination of prediction accuracy, computational efficiency, and robustness makes it a promising tool for data-driven bridge health monitoring and response reconstruction.