Benefits of multivariate statistical process control based on principal component analysis in solder paste printing process where 100% automatic inspection is already installed

• Main Author: Delgado, Pedro
• Publication Date: 2018
• Other Authors: Martins, Cristina, Braga, A. C., Barros, Cláudia, Delgado, Isabel, Marques, Carlos, Sampaio, Paulo
• Language: eng
• Source: Repositórios Científicos de Acesso Aberto de Portugal (RCAAP)
• Download full: http://hdl.handle.net/1822/70508
• topic: Hotelling’s T 2; Multivariate statistical process control; Normal operation conditions; Principal component analysis; Solder Paste Inspection; Squared prediction error; Variable contributions; Multivariate Statistical Process Control Principal Component Analysis; Science & Technology

The process of printing and inspecting solder paste deposits in Printed Circuit Boards (PCB) involves a very large number of variables (more than 30000 can be found in 3D inspection of high density PCBs). State of the art Surface Mount Technology (SMT) production lines rely on 100% inspection of all paste deposits for each PCB produced. Specification limits for Area, Height, Volume, Offset X and Offset Y have been defined based on detailed and consolidated studies. PCBs with paste deposits failing the defined criteria, are proposed to be rejected. The study of the variation of the rejected fraction over time, has shown that the process is not always stable and it would benefit from a statistical process control approach. Statistical process control for 30000 variables is not feasible with a univariate approach. On one side, it is not possible to pay attention to such a high number of Shewhart control charts. On the other side, the very rich information contained in the evolution of the correlation structure would be lost in the case of a univariate approach. The use of Multivariate Statistical Process Control based on Principal Component Analysis (PCA-MSPC) provides an efficient solution for this problem. The examples discussed in this paper show that PCA-MSPC in solder paste printing is able to detect and diagnose disturbances in the underlying factors which govern the variation of the process. The early identification of these disturbances can be used to trigger corrective actions before disturbances start to cause defects. The immediate confirmation of effectiveness of the corrective action is a characteristic offered by this method and can be observed in all the examples presented.