Wavelet-Entropy Approach for Detection of Bridge Damages Using Direct and Indirect Bridge Records

Abstract

Bridges as a key component of road networks require periodic monitoring to detect structural degradation for early warning. Early detection of loci and extent of structural flaws is essential to maintain safe bridge functioning. The elegant properties of continuous wavelet transform (CWT) in analyzing the signal in both time and frequency domains was the impetus to extensively employ this technique in structural health monitoring applications. However, the faint signature of structural damages in the recorded bridge responses curtails the merits of employing this technique. Furthermore, the selection process for the optimal CWT parameters that could capture signal discontinuities due to structural damages is an arbitrary process, which adds another level of uncertainty to wavelet transforms. This paper investigates compiling Shannon entropy to CWT to infer the loci and extents of structural damages in bridges. Entropy is a measure used to evaluate the randomness of the data. The more stochastic the data, the higher the entropy. In this article, Shannon entropy is used to associate a proper probability density function for the used wavelet to measure the entropy of the wavelet function at different scales. Implementing this technique facilitates selecting the optimal CWT parameters to better depict the signal; hence, identifying signal discontinuities becomes viable. The paper numerically investigates the fidelity of the proposed approach to identify bridge damages using midspan bridge response as well as using indirect records from a vehicle passing over the bridge. An implicit vehicle–bridge interaction (VBI) algorithm is used to mimic the vehicle–bridge interaction dynamics for different scenarios.