Research On Defect Detection Technology Of Hub Casting Based On X-ray Image

As an important part of automobile,wheel hub plays an important role in automobile safety.Therefore,after the hub manufacturing is completed,it needs to be tested to ensure the production quality of the hub.At present,the commonly used method is X-ray irradiation detection,divided into manual detection and automatic detection.Most manufacturers use manual method to evaluate the film testing,this method is more subjective,less reliable,less automation.The existing hub casting automatic detection system has complicated operation and slow detection speed,which is easy to cause misjudgment and omission for the hub with complex geometric structure.In view of the above situation,this paper studied the common defects of wheel hub casting,studied the X-ray image collection,quality improvement,defect segmentation,defect identification and other aspects of wheel hub casting,and developed an automatic defect detection system for wheel hub casting for industrial production.The main work and innovation of this paper are as follows:(1)This paper describes the process of automatic defect detection of hub casting,analyzes the quality problems in the process of X-ray image acquisition of wheel hub casting,optimizes the process of image acquisition by using multi-frame superposition method,and analyzes the selection of the optimal frame number in the process of using multi-frame superposition method to reduce the noise.(2)In order to better convert the high data collected by the flat panel detector into 8-bit data to display the image on the ordinary display screen.In this paper,based on the idea of HDR tone mapping technology,a tone mapping algorithm based on gradient domain guide filter is proposed.The gradient domain guided filter is used to decompose the high dynamic hub X-ray image,then the base layer is compressed by the adaptive logarithmic compression algorithm,the detail layer is enhanced by s-shaped curve function,and the image is finally combined and output.This algorithm restores the image details well,and the image quality isimproved,which lays a foundation for subsequent defect segmentation and recognition.(3)Considering the problem of missed judgment and misjudgment caused by the traditional segmentation algorithm,this paper proposes a dynamic threshold segmentation algorithm based on the region of interest,and reconstructs knowledge of mathematical morphology to reconstruct the result of low threshold segmentation with the result of high threshold segmentation,so as to obtain a complete hub defect and reduce missed judgment and misjudgment.By delineating the area of interest,the influence of irrelevant background is reduced and the detection speed is accelerated.(4)In this paper,the causes of Internal defects in Wheel Hub,extraction of defect features,feature selection,defect classification and so on are analyzed.For the classification of hub defect types,the multi-classification algorithm based on support vector machine(SVM)is introduced,and a multi-classifier model is designed based on one-against-rest and improved DAG-SVM multi-classification method.Particle swarm optimization algorithm is applied to optimize the model parameters of the two-classification support vector machine used in the model.The designed multi-classifier model solves the problems of error accumulation and undivided regions in the traditional multi-classification algorithm,improves the classification accuracy and speed,and finally completes the classification,identification and classification of defects.(5)Based on the algorithms of hub defect segmentation and defect recognition,automatic hub defect detection system is developed for practical industrial production.