Research On PCBA Solder Defect Inspection System Based On Machine Vision

With the continuous development of electronic technology,Printed Circuit Board(PCB)has become increasingly miniaturized and high density,and the market demand has increased year by year.PCBA(Printed Circuit Board and Assembly)refers to the entire process of PCB blank board passing through the upper part or plug-in,and also refers to the product after the PCB process is completed.In order to improve production efficiency,many companies have introduced streamlined welding machines to replace traditional welding work.However,many abnormal solders,such as multiple tins,short circuits,and tin tips,will still appear in the factory production environment.This paper focuses on the bad solders that may occur during the DIMM SKT soldering process in PCBA,and researches and designs the hardware and software systems and corresponding image algorithms for PCBA solder detection based on machine vision.First,the paper introduces the current development status of automatic optical inspection,solder inspection and deep learning.Then,through the analysis of the solder to be detected,a suitable image acquisition device is designed,a solder image acquisition algorithm is designed,a suitable convolutional neural network model is selected and improved,and the software operation interface of the system is developed Finally,test and analyze the performance of the entire system.In terms of solder image acquisition,combined with the characteristics of the system design mechanism,a fast and effective image stitching method was developed,and the automatic acquisition of solder images was achieved by using MARK hole positioning and affine transformation In terms of solder defect detection algorithms,various classic convolutional neural network models are studied,and VGGNet with high accuracy is selected for improvement,which mainly includes adding a residual network module and replacing the fully connected layer with global average pooling.The models are compared and the experimental results are given.Finally,the entire system was tested on the production site of the enterprise to analyze the system performance.The results show that the detection rate of the system for PCBA solder defect detection has reached 98.9%,which meets the production needs of enterprises.