Research On Defect Detection Of Negative Redundancy Of Lithium Battery Based On X-ray Image

The electrode slice is the key component of lithium battery.Both the use performance and the safety performance of lithium battery are affected by the quality of the electrode slice.During the production of winding lithium battery,coating,winding and other processes are easy to cause the end of the electrode fold and dislocation of positive and negative electrode.Problems as reduction of battery capacity,battery internal short circuit,and even the serious explosion accidents will be caused by these defects.X-ray testing is a common nondestructive testing method in industry.It can help to obtain the internal image of the battery cell and provide an effective way for the defect detection of the battery electrode slice.In this paper,the square winding lithium battery is taken as the research object,and the X-ray image of the electrode slice is detected and analyzed.A defect detection method of negative redundancy of lithium battery is studied to detect the redundancy of the negative electrode compared with the positive electrode in each layer of the battery,and to judge whether the quality of the electrode slice is qualified through the inspection of the length and angle.In this paper,according to the mechanism and preparation process of lithium battery,the types of defects in the Overhang(negative redundancy)region and the causes of defects are analyzed.By using machine vision technology,a negative redundancy defect detection system based on X-ray image is designed.Specific research contents are summarized as follows:(1)The overall design of the system was carried out.According to the design,the hardware equipment was selected,including light source,image intensifier and camera,etc.,and the software of the system was designed.(2)The negative redundancy detection algorithm of lithium battery is realized by using image processing technology,and the defect detection in Overhang region is realized by extracting the negative redundancy.The algorithm used Butterworth low-pass filter to denoise the image in the frequency domain,and used Laplacian edge detection,morphological operation,contour extraction and other technologies to separate the positive and negative pieces.Finally,the spatial filtering and LSD were used to obtain the negative redundancy.Through experimental analysis,the accuracy of the algorithm is 98.5%,and the average detection time of the algorithm is 721 ms,which meets the requirements of real-time detection on the actual production line.(3)SVM was selected as the classifier.According to the defect characteristics of Overhang region,5 geometric features and 6 gray features were extracted.The "one-to-one" multi-class classification method was used to carry out classification experiments,and the ideal classification effect was obtained.(4)Design the interface and function of the software part of the system,and check the function of the system.The results show that it has certain practicability.