An Implementation of Health Prediction in SMT Solder Joint via Machine Learning

Abstract

The assembly process of printed circuit boards (PCBs) has widely adopted surface mount technology (SMT) in past several decades due to better cost effectiveness resulted from its automated production and soldering process. Nowadays, in demand of products with miniaturized components and high density placement of components on boards, the precision of SMT have become an important and challengeable issue. To maintain good quality of solder joints of component mounted on PCBs, automated optical inspection (AOI) has been commonly utilized for component inspection. Although the AOI is able to detect the defects on solder joint of component, the false detection rate of AOI is still high. The false detection rate is especially a serious issue in the production with components of high-density and miniaturization integration since the result will affect the production yield rate and overall equipment effectiveness. Therefore, in order to reduce this kind of problem, an approach based on machine learning is proposed in this paper to predict the health of solder joint. The experimental results indicated that the proposed method is not only more efficient, but also provides a high improved rate of 88.8% in SMT process.