Research On Detection Method Of End Face Defect In Cylindrical Lithium Battery

The 18650 cylindrical lithium battery is a battery type widely used in the market.Affected by the material of the battery itself and the irregular behavior in the production process,the battery produced is often accompanied by various defects.At present,most domestic battery manufacturers mainly use manual to test the surface defects of the battery.Manual detection mainly relies on the subjective judgment of the sorting personnel.The disadvantages of this method are that it takes a long time and the detection accuracy is poor,and the missed detection and the false detection frequently occur.Therefore,in order to improve the speed and accuracy of manual detection,it is of great significance to study an automatic detection method for surface defects of cylindrical lithium batteries.The battery material produced by the battery manufacturer is used as the test object,and the battery end face is pit or deformed,broken or wrinkled,scratched,oxidized,leaked,the bottom of the shell is welded,the solder joint is raised,and the film is covered with abnormalities.Surface defects are the detection targets,a set of perfect battery end face defect detection methods is established with high detection accuracy,fast detection speed and less manual intervention.This paper designs a cylindrical lithium battery end defect automatic detection system,which can initially realize eight common surface defects such as metal surface pit or deformation,metal surface scratch,film damage or wrinkle,film abnormality,metal surface oxidation,positive electrode leakage,negative electrode,under-weld and solder bumps,which focuses on the design of the defect imaging scheme and the design of the inspection method.Firstly,the morphological structure of the end face of the battery is introduced,and the causes and morphological features of the common defects existing on the end face are analyzed.The eight types of defects are divided into three-dimensional defects and planar defects.According to the composition of the whole detection system and the function of the imaging module,the defects are mainly present.The structural form of the part,the hardware imaging system is designed to complete the acquisition of the oblique light image and the diffuse light image.In the aspect of defect detection algorithm,the detection image is firstly evaluated for vacancy,and then the end battery image is accurately positioned by edge detection and circle fitting,and the ROI region is extracted by threshold segmentation.Threshold segmentation is performed on the pit or deformation defect based on the criterion elimination of outliers,and the defect region is extracted by morphological processing and area feature.Scratch defects are extracted by the Hessian matrix-based Steger method,and then the length and aspect ratio features are used to extract the defect lines.The film damage or the wrinkle defect is detected by the convexity change of the edge line,and the defect is identified by determining the difference set between the original line area and the convexity converted line area.For the abnormal defects of the film,multiple measurements were taken and the average value was obtained to obtain the true width of the film,and the abnormality of the film was determined in comparison with the standard width.Threshold segmentation is performed on the oxidative defect by a dynamic threshold segmentation method based on mean filtering.For the positive electrode leakage defect,the positioning is performed by shape-based template matching and affine transformation,and the defect is determined by the average gray value of each independent region obtained by the positioning.For the negative metal surface soldering and solder bump defects,firstly,the algorithm detection speed is improved by narrowing the target area to be detected,edge detection is used to extract the defect contour,and the contour length and roundness information are used as the judgment conditions to extract the defects.Finally,after comprehensive testing of the sample,the detection speed can reach 200 batteries per minute,the average miss detection rate is 9.87%,and the average false detection rate is 3.99%,which has a good detection effect.