Premium
Regularization‐based statistical batch process modeling for final product quality prediction
Author(s) -
Yan Zhengbing,
Chiu ChihChiun,
Yao Yuan,
Doing Weiwei
Publication year - 2014
Publication title -
aiche journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.958
H-Index - 167
eISSN - 1547-5905
pISSN - 0001-1541
DOI - 10.1002/aic.14476
Subject(s) - regularization (linguistics) , process (computing) , interpretation (philosophy) , process modeling , computer science , chemical process , regression analysis , regression , mathematics , machine learning , work in process , engineering , statistics , artificial intelligence , operations management , operating system , chemical engineering , programming language
Prediction accuracy and model interpretation are two important aspects with regard to regression models. In the field of statistical modeling of chemical batch processes, most research focuses on prediction accuracy, while the importance of the latter aspect is often overlooked. In multiphase batch processes, it is possible that only a few phases are relevant to certain quality indices, while different time points belonging to the same relevant phase usually have similar contribution to the quality. The regression coefficients of batch process model should reflect such process characteristics, that is, the coefficients corresponding to the irrelevant phases should be close to zero, while the coefficients of each variable within the same phase should vary smoothly. In this study, regularization techniques are introduced to statistical modeling of chemical batch processes to achieve both accurate prediction and good interpretation. The application to an injection molding process shows the feasibility of the proposed methods. © 2014 American Institute of Chemical Engineers AIChE J , 60: 2815–2827, 2014