Bayesian Network-Based Process Monitoring and Diagnosis in Complex Manufacturing Systems

• Activity Number: 190
• Type: Topic Contributed
• Date/Time: Monday, July 30, 2007 : 2:00 PM to 3:50 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Jing Li*+ and Jianjun Shi and Jionghua Jin
• Companies: The University of Michigan and The University of Michigan and The University of Michigan
• Keywords: Bayesian network ; causal model ; Statistical Quality Control ; Statistical Process Control ; diagnosis

Abstract

Process monitoring and diagnosis is a challenging problem in manufacturing systems with many variables and complex relationships. Because diagnoses require tracing backward from a fault to its root causes, the causal relationships among the variables must be identified. Bayesian networks provide a tool to model the causal relationships, enabling effective diagnoses. This paper proposes a "Bayesian Network based T2 decomposition" method, which uses the causal relationships, represented by a Bayesian Network, to guide the decomposition of the out-of-control signal in a T2 control chart for diagnostic purposes. Theoretical analysis and simulation studies demonstrate that the proposed method has a substantially enhanced diagnosability and reduced computational complexity, compared with the traditional approach that decomposes the out-of-control signal without prior causal identification.