Research On Rail Flaw Detection Based On Feature In Time-Frequency-Space

Due to the heavy traffic of the railway system,the rail is constantly subjected to rolling and friction.Under the influence of the external natural force and internal unbalance force,the interior of the rail is prone to defects,and even breaks.The safety of rail is related to the safety of railway transportation,especially in recent years,China's rapid development of high-speed rail.High speed need higher performance of rails,poor performance of rails easy to make people and material suffered great losses,so it is necessary to detect the rail whether there are defects.Ultrasonic has good transmission properties,so it is widely used in the field of rail flaw detection.Ultrasonic can penetrate the surface of the rail into the interior,it is very suitable for the detection of defects in the rail.This paper based on time-frequency-space to get the feature of the defect detection signal,using autocorrelation to get the feature of multiple signal in time domain,in frequency domain,using wavelets transform to get the signal distribution of energy,finally,using the support vector machine to identify the feature vector.The main contents of this paper are as follows:1.Using array probe to detect the internal defect of rail with linear frequency modulation ultrasound,linear frequency modulation ultrasound with a rich spectrum,can be a good indication of the defect information.The array probe can effectively receive the ultrasonic echo of defect reflection and diffraction.2.The wavelet threshold denoising method is used to denoise the ultrasonic flaw detection signal,and the feature extraction is studied in the time domain,frequency domain and space domain.the signal is decomposed by wavelet to obtain the energy occupancy in the different decomposition space as the characteristic in the frequency domain;feature vector timefrequency characteristics of multiple signal components characterization of defects.3.The support vector machine is used to identify the ultrasonic flaw detection signal,judge whether the rail is defective.The particle swarm optimization algorithm is used to calculate the most suitable support vector machine parameters in the linear frequency modulated ultrasonic flaw detection signal.4.Set up the ultrasonic flaw detection system and collect the defect detection data.After extracting the feature vectors of the data,training and testing the support vector machine.