Component registration diagnosis for printed circuit boards using process-oriented basis elements

Abstract

Modern product and process sensors are capable of making repeated measurements of similar characteristics at multiple locations on a part. Under such conditions, is possible to produce process diagnostic analyzing the multivariate process vector using a process-oriented basis (POB). Many potential production problems will have characteristic signatures that can be detected in the multivariate quality vector. Each pattern (signature) is associated with a basis element in the POB. The multivariate quality vector can be represented as a linear combination of these basis elements. Patterns with large coefficients in the representation suggest particular causes for process problems allowing manufacturing diagnostic information to be derived from the decomposition of the multivariate SPC data. The methodology is applied to the registration problem in fine pitch components for Printed Circuit Boards (PCB). Each PCB panel consists of 16 components, each of them with 8 measurements, resulting in a multivariate vector of 128 components. The suggested control procedure monitors through control charts the representation of each defined pattern to identify time periods in which such patterns are active.