Low cost health monitoring of cable stayed bridges using synchrosqueezed wavelet transform and nonlinear principal component analysis

Today, health monitoring of structures has been standardized in many countries. Such systems for large and complex structures are equipped and include numerous sensors. Therefore, they are not yet practical in our country due to large final expenses. The main purpose of this paper is to introduce a low-cost health monitoring algorithm for structures based on signal processing. Accordingly, only three sensors are utilized to detect damage. Since the accuracy of signal processing method can affect the results of damage detection, in the first part of the paper, five signal processing methods are investigated. Among these procedures two methods are older and have widely used in damage detection. The three others are more recent and are fully investigated in civil engineering. In the second part, a new damage detection method is proposed. This method is a combination of signal processing methods by synchrosqeezed wavelet transform, clustering, and regression with autoassociative artificial neural networks. For this reason, data from Yonghe bridge is utilized which is recorded based on real vibration of the bridge. Results show that the proposed signal processing method is capable to effectively extract signal features. Also the damage detection method is capable to detect damage with acceptable accuracy