Research Of Train Wheel Surface Detection Based On S-Transform Profilometry

Railway safety has been the subject of attention from the State and society. Railway wheel surface defects detection is an important aspect of railway safety. The fast and reliable method to detect wear, scratches, cracks and other defects is important. Currently methods used for the detection of wheel surface defects include artificial method, cutting-line 3D reconstruction method, magnetic particle flaw detection method and ultrasonic test method, and so on. Manual detection method is time-consuming, cutting-line 3D reconstruction method can hardly get the entire circumference of defect information, it is mostly used for testing the entire wheel tread wear and is not suitable for detection of defects, magnetic particle method is simple but time-consuming, ultrasonic testing method has high sensitivity for metal detection, but it relies on couplant, and has complex testing procedures. For fast and reliable wheel tread defect detection, this paper introduces the s-transform profilometry into flaw detection of wheel tread. This method has the advantage of high resolution, fast, non-destructive, the wheel surface defects information can be reconstructed with just one image. S-transform profilometry theory and application on wheel surface defects detection is studied in this paper. The main work of the paper is as follows:Paper first introduce the principle and implementation of optical three dimensional measurement, focused on the fringe pattern analysis and phase demodulation. The three advantages of S-transform is described. Connection between the S-transform, the Fourier transform and Wavelet transform is studied. Using the relationship between S-transform and Fourier transform, the image can be transformed to Fourier spectrum, and pretreated to reduce the noise. Then the Fourier spectrum can be S-transformed and the effect of phase demodulation can be improved.Researched the principle of S-transform filtering and phase demodulation. Derived the Ridge expression in the S-transform coefficients. Analyzed four typical method of frequency domain filtering solution based on the ridge, and got the two dimensional filter window based on S-transform ridges, and pointed out the problems of filtering scheme. Simulated the reconstructed results through the four filters, analyzed the errors, and got that the reconstruction error of flat-topped Gaussian window is the smallest. By adjusting the height of the simulation model, analyzed the four filter reconstruction error under Windows at different height rate, and got that under different rate of height, flat-topped Gaussian window reconstruction error is the smallest, and the greater high rate, the more reconstruction error.The two dimensional filter window’s problem based on S-transform ridges is resolved by polynomial fitting on the ridge, avoiding filters’not continuous, and got that the reconstruction error with flat-topped Gaussian window after polynomial fitting on the ridge is the smallest. By introducing a second order phase factor, the accuracy of phase demodulation is improved. In the position of large height rate, the reconstruction error is smaller introducing the second order phase factor. Simulation and experimental validation show that the reconstruction result is more accurate after the improvements.Wheel tread defect detection platform is built and its basic parameters and selection of equipment is introduced. Analyzed the defect detection process. Wheel tread defect detection experiments were conducted, compared the effect of image preprocessing before and after. Analyzed the reconstruction errors before and after the polynomial fitting on the ridge and introducing a second order phase factor. Results show that the improved s-transform profilometry for wheel surface local defects inspection has high reliability, it has guiding significance on application of rail wheel surface defects detection.