Bayesian Analysis of Multistage Process Yields

• Activity Number: 679
• Type: Contributed
• Date/Time: Thursday, August 5, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Section on Quality and Productivity
• Author(s): Patrick J. Gaffney*+ and Alan Richter
• Companies: ImClone Systems and ImClone Systems
• Keywords: Bayesian ; Yield ; WINBUGS

Abstract

Manufacturing systems can be divided into several processing steps or stages. Yield is often used to characterize the performance of each stage. For this paper, yield is defined as the ratio of output to input product mass. The computed yields of successive stages are correlated - for example, the measurement error that inflates the yield of one stage will result in a corresponding decrease in the computed yield of the next stage. The system is also subject to a loss of product - hence a drop in yield - at any stage due to operating issues; yet the yield in successive stages is independent of this loss. This paper outlines a Bayesian model for the product mass and step yields for such a system. The model is fitted using WINBUGS (and BRugs), and is used to identify a drop in yield or measurement error. An example is provided from antibody purification in the biotech industry.