DESIGN FACTORS AND THEIR EFFECT ON PCB ASSEMBLY YIELD - STATISTICAL AND NEURAL-NETWORK PREDICTIVE MODELS

This study relates circuit board design features to assembly yields. D ata used were collected over a period of one year from two circuit boa rd assembly shops at AT&T. Design parameters that may affect the assem bly yield were identified using knowledge of the assembly process. The se parameters were then quantified for a set of board designs and rela ted to the actual assembly yield by the statistical regression models and the artificial neural network (ANN) models. These models are able to predict the assembly yield with a root mean square (RMS) error of l ess than 5%. They can be used to predict the assembly yield for new bo ard designs on the same line. Alternatively, they can be used to compa re the performance of different lines by comparing the expected yield r a given design with the actual yield.