Research On Internal Defect Recognition Method Of Plastic Packaging Chip Based On Ultrasonic Imaging

Plastic package is a non-hermetic package,so plastic packaged chips are prone to internal defects.The internal defects of plastic encapsulated chips are displayed in the form of images through ultrasonic non-destructive testing,which is a commonly used method of chip failure analysis.Ultrasonic inspection images of internal defects of plastic-encapsulated chips have the characteristics of small defect size and a lot of interference noise,therefore,manual identification of ultrasonic inspection images of chips has disadvantages such as low efficiency and high misrecognition rate.Aiming at this kind of problem,this article focuses on the theory and key technology of ultrasonic image recognition of plastic-encapsulated chip internal defects,and provides intelligent technical support for the ultrasonic detection image recognition and analysis of plastic-encapsulated chip internal defects.The main work content is as follows:First,collect and analyze the ultrasound images inside the plastic-encapsulated chip.With reference to the actual chip production process,a variety of poor process conditions are set,and the plastic-encapsulated chips produced under these conditions are ultrasonically scanned to obtain different types of ultrasonic images of the internal defects of the plastic-encapsulated chips.Aiming at the characteristics of noise and the problem of small quantity in the chip ultrasound image,the traditional image preprocessing technology is used to reduce the noise and enlarge the image.To summarize and analyze the internal defects of plastic-encapsulated chips,firstly,a chip inspection data set is constructed,and four types of defect classification data sets are established according to the characteristics of the defect's internal position,shape and size.Second,according to the characteristics of the image,a method for detecting and positioning the chip is implemented.According to the dense distribution,small size and fixed characteristics of chips in ultrasound images,a fast chip detection algorithm is designed based on Convolutional Neural Networks(CNN).This method achieves 98%detection accuracy in the detection verification data set,8ms inference detection speed per picture,faster detection speed and accuracy than other methods,and can replace the manual detection and positioning of the chip in the ultrasound image.Third,an algorithm for identifying internal defects of plastic-encapsulated chips based on the deep learning model is designed.According to the random characteristics of the position and size of the internal defects of the chip,the attention mechanism is used to improve the network's ability to identify internal defects of the chip.In the verification experiment,the improved network is compared with the original network,and the recognition accuracy is increased by 1.04% to 99.37%,which has better recognition accuracy than other methods.Finally,the chip inspection method and defect classification algorithm are integrated,and the mapping relationship between defect types and process parameters is used as a comparison database,and a plastic-encapsulated chip ultrasonic inspection image recognition and analysis system is designed.Through actual engineering application analysis,the system can accurately detect all chips in the image and the defect type of each chip after inputting the ultrasound image,and then output the defect rate,the proportion of each defect,and the optimization direction of the plastic packaging process.Actual application value.