ENHANCEMENT OF STRUCTURAL HEALTH MONITORING FRAMEWORK ON BEAMS BASED ON K-NEAREST NEIGHBOR ALGORITHM

The aiming of this work is to enhancement the structural health monitoring (SHM) framework of beams structure for damage detection to treatment the drawbacks of poor detection efficiency in traditional of beams monitoring algorithms, the improvement framework on beams SHM is based on novel data classification technique through designing the k-Nearest Neighbor (k-NN) algorithm. First, the beam finite element model under impact load is analysis, and the cumulative damages are considered and introduced to beam model. The datasets of beam SHM are compiled from the sensors installed in beam structure, and then processed using kernel principal component analysis to remove the unnecessary features and reduce the scale of classification features. The k-NN algorithm parameters of beam SHM are determined by the genetic optimization algorithm (GOA) to establish the optimal SHM classification model of beam. Finally, a comparison between the present damage detection results via k-NN and traditional models via convolutional neural network (CNN) and supper vector machine (SVM) results available in literatures is established through the most significant indexes of testing to check the effectiveness and superiority of suggested method. The results show that the presented SHM model are gave higher precision, reduced the time of modeling, and improvement the total performance of damage detection model in beams. The current performance are recorded 95.3%, 91.8%, and 89.7%, for accuracy rate, recall rate, and F-score respectively.