Mixed-effects Gaussian process modeling approach with application in injection molding processes. (February 2018)

• Record Type: Journal Article
• Authors: Luo, Linkai; Yao, Yuan; Gao, Furong; Zhao, Chunhui
• Is Part Of: Journal of process control. Volume 62(2018)
• Page Start: 37
• Page End: 43
• Publication Date: 2018-02
• Subjects: Mixed-effects model -- Gaussian process -- Multi-process data -- Injection molding; Process control -- Periodicals; Fabrication -- Contrôle -- Périodiques; Process control; Periodicals; Electronic journals; 660.281
• DOI: 10.1016/j.jprocont.2017.12.003
• Languages: English
• ISSNs: 0959-1524

Abstract

Highlights: We develop an ME-GP modeling approach for multi-process data. ME-GP improves the prediction performance by combining common information. We quantify uncertainty in prediction results. The approach is demonstrated for real-world injection molding process. Abstract: We propose a new nonparametric approach for multi-process data analysis, in which each of the process is modeled as a combination of a fixed-effect and a random-effect Gaussian process (GP) regression model, namely, a mixed-effect Gaussian process (ME-GP) model. The ME-GP approach provides a flexible means to combine the common aspects of all processes and describe the heterogeneity among different processes. In particular, we model the mean and covariance structures of both the fixed- and random-effects simultaneously, and predict a future input using probability density distributions. We apply the ME-GP model to predict the melt-flow-length for filling of different molds in injection molding processes. It is shown that the ME-GP model obtains an improved performance against GP model only.