Recognition Method of Two Types of Insulation Joints Based on Wavelet Transform and SVM

Abstract

Mechanical insulation joints in track circuits are generally insulated by gluing, and their smoothness is generally weaker than that of electrical insulation joints, which will cause high-frequency vibration of vehicles in case of serious damage. In this paper, we propose a method to identify the two types of mechanical and electrical insulation joints effectively based on wavelet transform and support vector machine (SVM) according to the characteristic attributes of data signals. Firstly, the location of the insulation joint is realized based on the modulus maxima of the wavelet transform and the characteristics of the signal at the insulation joint. Secondly, the samples are divided according to the position point and the length of the insulation joint, and the features of the samples are extracted. Finally, the SVM algorithm is utilized to construct a classification model for distinguishing the two types of insulation joints. Experimental results demonstrate that the proposed method can effectively locate and distinguish the two types of insulation joints.