Estimating and incorporating the effects of a future QE project into a semiconductor yield learning model with a fuzzy set approach

Yield prediction is a very important task to a semiconductor manufacturing factory. However, it is difficult because yield improvement is a learning process, and the uncertainty and variation inherent in the learning process are not easy to consider. In addition, the competition in the semiconductor industry is becoming more and more fierce, which significantly distorts the learning process of yield improvement. For example, some managerial actions (including increasing the release frequency and executing a quality-engineering project) might be taken to accelerate yield learning at various stages to prevent the product from losing competitiveness. The effects of such actions might be unstable, but still have to be estimated and then incorporated into the yield learning model. For this purpose, a fuzzy set approach is proposed in this study. At first, the fuzzy Delphi method is applied to aggregate the judgement results by multiple experts about the effects of a managerial action. Subsequently, a subjective correction function is designed to incorporate the aggregation result into Chen and Wang's (1999) fuzzy yield learning model. To evaluate the effectiveness of the proposed methodology, it has been applied to the data of two random-access-memory products. [Received 02 July 2008; Accepted 06 October 2008]