Innovative intelligent and expert system of bridges damage identification via wavelet packet energy curvature difference method integrated with artificial intelligence algorithms

Abstract

Bridges are important infrastructure for highways. Monitoring their status is of great significance to ensure safe operations. In this work, a novel integrated technique from wavelet packet energy curvature difference (WPECD) and artificial intelligence (AI) for bridge damage identification is established. Initially, the damages are simulated in the bridge decks by changing the material stiffness reduction levels of bridge elements by three levels (5%, 10%, 15%) to study the effect of damage on the bridge response. Then the WPECD maps are plotted from vibration response before and after damage to the bridge for each stiffness reduction level. Unfortunately, given the nonlinearity of damage geometry, it is not easily feasible to use WPECD maps for damage identification accurately. Therefore, the (WPECD) maps are used for training a new architecture of recurrent neural networks with long short-term memory blocks (RNN-LSTM) for bridge damage identification by predicting the wavelet functions and wavelet decomposition layer effect of each node in the bridge. The effectiveness and reliability of the proposed approach were confirmed by numerical and experimental results. The performance of the proposed technique achieved high scores of accuracy, regression, and F-score equal to 93.58%, 90.43% and 88.17% respectively indicating the applicability of the proposed method for use on other important highway infrastructure.