The Research On Processing And Analysis Of Ultrasonic Flaw Signal Of Rail

High-speed rail system of China has achieved remarkable achievements inthe world for its endless operating mileage, strong integration capabilities, full-scale construction and robust system technology. With the accelerating ofexisting lines and the development of the high-speed rail, rail load is graduallyincreased and the security issues are also deteriorating. Therefore, emphasizingon defects analysis and identifying the cause of malfunction not only extend theservice life of the rail but also ensure traffic safety.As one of the most important methods, ultrasonic inspection technique(UIT)has been widely applied in rail flaw detection due to its advantages such as highsensitivity, fast detection, accurate orientation.However, the defect recognitiontechnology is not perfect, and defect data identification is still difficult to a largeextent which requires highly skilled workers. Therefore, this paper has carriedout study in view of rail defect identification and real-time acquisition, hoping topromote the development of UIT to automative and intelligent directions.Wavelet packet component energy is adopted among several eigenvalueextraction methods discussed in this paper. It also puts forward optimized LDAalgorithm and class probability density algorithm.The latter is used as featuredimension reduction method to accomplish preliminarydefect identification.This paper collects delamination, inclusion and porosity there kinds ofdefects as test object. Grouping the collected signals to built BP artifical neuralnetwork. Network performance is validated by calculating and comparingexpected output value and actual one. Data analysis point-by-point is realized bysuccessfully building the rail ultrasonic LabVIEW platform.