Research On Automatic Flaw Localizition Algorithm For Mobile Wheel Set Ultrasonic Inspection System

With the rapid development of world’s high-speed railway, China railway also gets fast growth in order to conform to developing trend of national economy and meet people’s needs. In recent years, along with continuously improving of China railway speed, especially running of high-speed motor train units, more rigorous and sophisticated maintenance and monitoring of rolling stock travelling parts are needed to implement, departments of rail safety inspection must accept new challenge.Currently, some developed countries have designed and established their own advanced wheel set inspection system. In our country, a mobile wheel set ultrasonic inspection system is researched and developed independently, which is based on phased array ultrasonic technology and general ultrasonic technology. The system can realize online automatic non-destructive testing of wheel rim and spoke of all kinds of CRH, and defect various types of fatigue flaws and material flaws. However, defects are positioned by means of dragging frame manually. It’s inevitable for bringing some human operative errors and confining the efficiency of defects testing. It is difficult to adapt the requirements of rapid development of China’s high-speed railway.Aiming at designing a set of flaw localization algorithms for the in-site wheel testing data of mobile wheel set ultrasonic system. Ultimately, the time performance of data analysis in practical inspection of system would be better, and the veracity of checking and calibrating would also be improved.In the paper, on the premise of investigating and summarizing wheel testing technologies and devices at home and abroad, combining with analysis of China high-speed motor railway wheel’s structural characteristics and ultrasonic phased array technology, the features of defects and interferences are extracted, and a set of defects localization algorithms are proposed based on region segmentation and growing, automatic contrast adjusting, periodic prolongation, relativity, KMEANS cluster and HOUGH transform. By using these methods, flaws in the testing image can be positioned accurately, and the pre-setting testing effect is achieved. A field application implementation of localization algorithm is proposed here, and gives some guidelines for the work afterward.