A Scale-Free Classification Model for Defect Diagnosis in the Pick and Place Machine

This study aims to develop a scale-free classification-based defect diagnosis model for the pick-and-place (P&P) machine in Surface Mount Technology (SMT) assembly line. SMT is a manufacturing process used to assemble Printed Circuit Boards (PCBs). The P&P process is the primary procedure that follows the application of solder paste or adhesive to the board. Generally, the industry uses an automatic optical inspection (AOI) machine to detect assembly defects just after the P&P process. However, inspection data from the AOI machine can only identify assembly defects; it cannot reveal the underlying causes of assembly failure. By conducting experiments with initial machine defects, it is possible to identify patterns associated with various root causes. Using the AOI and machine performance data, it is possible to trace the root causes of assembly defects using various machine learning methods. As the number of components used in the SMT assembly line increases, processing design of experiments (DOE), collecting sufficient data, and developing a defect diagnosis model for each type of component becomes time-consuming. The proposed model is trained on a single component type and then applied to other component types. Using the proposed model, when a new component is applied, the identification accuracies are more than 75.00% for most of the root causes without conducting DOE. It can significantly reduce the time required to process experiments, collect data, and adjust models for new types of components.