Research On Wagons Bogie Typical Fault Detection Method Based On Machine Vision

In recent years, with the rapid development of railway freight transportion industy of China, the high-speed, heavy-duty and high density have become the trend of freight train. The traditional manual inspection due to the impact of factors such as train inspection personnel have low efficiency, so this detection methods can not meet the development requirements of the current freight trains. For this, the railway company launched a safety inspection system by using machines instead of human eyes for railway wagons run fault detection, that is TFDS(Trouble of moving freight car detection system, dynamic truck operation fault detection system), the system can greatly reduce the interference from human factor and improve the detection efficiency and accuracy. However, at this stage the system is still in a train inspection man-machine combination, in order to further improve the detection efficiency and promote the TFDS to ‘machine control’ paradigm shift. The paper designed the corresponding algorithm to achieve fault detection by combining with the theory of computer vision, for three typical faults bogie wagons: Rolling shaft bolts loss fault, fault and lost the front cover of Rolling Bearing pillow spring loss fault, the corresponding algorithm design fault detection.By studying the image preprocessing technologies. Firstly, the corresponding theory of the image denoising, image enhancement and image segmentation are detailed description, and combined with the actual fault image study each algorithm principle, analyzed the scope of their respective advantages and disadvantages. Secondly, the edge detection operator is described, including Roberts, Sobel, Prewitt, Log, Laplace and Canny operator, the edge detector will be used for the rolling bearing shaft bolts loss failure. Finally, described the fault diagnosis methods, including Hough transform and template matching, these methods will be used in bogie wagons’ three typical fault detection of this paper.The assisted positioning thinking is used for Rolling shaft bolt failure, based on the relationship between the wheel shaft and the bearing shaft bolt indirectly targeted area is located, and then to interpret fault detection by Hough Transform.The fault of Rolling bearing former cover’s loss is still using assisted positioning ideas, but normalized correlation coefficient matching method is used in the interpretation of fault phase.For bolster spring missing fault, the bolster spring is located by using the template matching method based on scale-space theory and then through the template matching seek normalized correlation coefficient to the fault judgement.The paper designed and prepared the relevant system interface for fault detection provides a better interface fault interpretation by TFDS understanding and analysis. Finally, the three algorithms are tested, from the results can be seen this algorithm has a better applicability and can correctly identify faults. Through the practical application shows that the three fault corresponding algorithms are stable and able to meet the actual requirements.