Surface Mounting Production Line Automatic Optical Inspection (AOI) False Call Classification with Machine Learning Algorithms

Surface mount technology (SMT) has been known as the trending technology in electronic packing manufacturing. Standard SMT production lines have 3 major processes: stencil screen printing, package pick and place mounting, and forced convection reflowing. There are inspections after each of the 3 processes. The Automatic Optical Inspection (AOI) machines are used after the mounting and reflowing processes, named by the sequence in the production line, pre-reflow and post-reflow AOI. One of the important performance matrices of the AOI is the offset, which indicates the distance from the actual chip location to the desired location (pad center). If the offset is greater than the threshold, the AOI will trigger the alarm and reject the product. During the reflow process, some components were detected as “not good” at pre-reflow AOI could result in a “good” offset at post-reflow AOI because of the self-alignment performance in the reflow process. These are the false calls of AOI defined in this study. To improve the throughput, the number of false calls needs to be inspected and reduced. In this study, a classification model is proposed to inspect the false calls from the solder paste inspection (SPI) and pre-reflow AOI after comparing the performance of multiple machine learning algorithms and analyzing 30,000 experimental data points collected from the production line. The resulting accuracy rate was found to be 89.3%. As a result, it was determined that the inspection data before the reflow process could be effectively used in detecting AOI false calls.