Classification recognition of anchor rod based on PSO-SVM

Based on learning principle of support vector machine (SVM), damage detection of bolt anchorage is studied in this paper. Characteristic matrix is composed of 17 time domain or frequency domain characteristic properties including mean value, peak value of the denoised signal, etc. Principle component analysis(PCA) is used for feature extraction and several main components with higher cumulative contribution rate are chosen for SVM classifier training and testing. Particle swarm optimization(PSO) method is used to optimize the penalty parameter C and kernel function radius g of SVM. In experiments, choose 90 groups of data for three different types of anchor as the sample set, respectively. Select the first 70 samples as the SVM training set of each type anchor, and the remaining as the test set. The experimental results show that the proposed method is effective in identification of anchor, and the highest prediction accuracy can reach 93.333%.