Bridges Damage Assessment Techniques Improvement Through Machine Learning Algorithm

The main goal of this paper is improving bridges structures health detection results to solve the problems of large errors of detection and poor efficiency of detection in the traditional models of detection. A bridge structural health monitoring (SHM) model based on data classification technique is designed by using the k-Nearest Neighbor (k-NN) algorithm. First, the sensor network is used to collect the health status data of the bridge structure, and the kernel principal component analysis is used to process the sensor datasets to remove the redundant features and reduce the classification feature scale. Second, adopt on the k-NN algorithm, the learning of the sensor datasets, and the optimum parameters of the bridge structure health state (SHS) to establish the optimal bridge SHS detection model. The optimum parameters of classification model are determined by the genetic optimization algorithm (GOA). Finally, the bridge SHM model is analyzed by using the most significant indexes of testing for its effectiveness and superiority. The results show that the proposed model has a classification results of the bridge SHS with higher precision, lower modeling time. The overall performance of proposed classification model is significantly better than exists methods, and the current method performance are recorded 97.4%, 93.7%, and 91.3%, for accuracy rate, recall rate, and F-score respectively. This indicates that the current SHM model provides an effective tool for bridge structure health research.