A Modified Lasso Model for Yield Analysis Considering the Interaction Effect in a Multistage Manufacturing Line

摘要

Yield analysis aims to rapidly determine the yield loss factor and maximize yield. In a multistage line involving sequential process steps with various tools in each step, such as semiconductor fabrication lines, the interaction between the factors of adjacent process steps can significantly affect the yield. Although various yield analysis methods have been proposed, none can easily consider this interaction. If the process step distance of factors comprising the interaction is long, the interaction effect will be difficult to interpret. This study proposes a yield analysis least absolute shrinkage and selection operator (YA-Lasso) to identify the influencing interaction factors based on the process step distance. The YA-Lasso efficiently selects interaction factor candidates by combining yield loss factors and calculates the importance of factors identified using the score function. We analyze the importance of the yield loss factors selected via the proposed method and comparison methods using simulation data and the data collected by a multilayer ceramic capacitor manufacturer. The results show that compared with the other methods, the proposed method identifies the significant factors within a shorter time.