A large-scale statistical process control approach for the monitoring of electronic devices assemblage

In this paper, we present a new procedure for monitoring the assembly process of electronic devices. Monitoring the status of this operation is a challenge, as the number of quality features under monitoring is very large (order of thousands) and the number of samples available quite low (order of dozens). We propose an efficient approach for the on-line and at-line monitoring of such a process, by addressing two, hierarchically related, problems: (i) detection of faulty units (printed circuits boards with abnormal deposits); (ii) given a faulty unit, find a candidate set of solder deposits responsible for the anomaly. Our methodology is based on a latent variable framework using PCA for effectively extracting the normal behavior of the process. Both the variability in the PCA plane and around it (residuals) are considered. We have tested the proposed approach with real industrial data, and the results achieved illustrate its good discrimination ability.