A Bayesian variable selection method for joint diagnosis of manufacturing process and sensor faults. Issue 4 (2nd April 2016)

• Record Type: Journal Article
• Authors: Li, Shan; Chen, Yong
• Is Part Of: IIE transactions. Volume 48:Issue 4(2016)
• Page Start: 313
• Page End: 323
• Publication Date: 2016-04-02
• Subjects: Sensor fault diagnosis -- sparse faults -- Bayesian variable selection -- process fault diagnosis; Industrial engineering -- Periodicals; Génie industriel; 620
• DOI: 10.1080/0740817X.2015.1109739
• Languages: English
• ISSNs: 0740-817X

Abstract

ABSTRACT: This article presents a Bayesian variable selection–based diagnosis approach to simultaneously identify both process mean shift faults and sensor mean shift faults in manufacturing processes. The proposed method directly models the probability of fault occurrence and can easily incorporate prior knowledge on the probability of a fault occurrence. Important concepts are introduced to understand the diagnosability of the proposed method. A guideline on how to select the values of hyper-parameters is given. A conditional maximum likelihood method is proposed as an alternative method to provide robustness to the selection of some key model parameters. Systematic simulation studies are used to provide insights on the relationship between the success of the diagnosis method and related system structure characteristics. A real assembly example is used to demonstrate the effectiveness of the proposed diagnosis method.