化学工业与工程
Home  |   |  About Journal  |  Editorial Board  |  Instruction  |  Subscriptions  |  Download  |  Publication Ethics  |  Contacts Us  |  Chinese
Chemcial Industry and Engineering 2022, Vol. 39 Issue (2) :32-36    DOI: 10.13353/j.issn.1004.9533.20210399
Current Issue | Next Issue | Archive | Adv Search << | >>
A batch process monitoring method based on long short term memory network and support vector data description
JI Cheng, GU Junfa, WANG Jianhong, WANG Jingde, SUN Wei
College of Chemical Engineering, Beijing University of Chemical Technology, Beijing 100029, China

Abstract
Reference
Related Articles
Download: PDF (1819KB)   HTML ()   Export: BibTeX or EndNote (RIS)      Supporting Info
Abstract Batch process is an important operating mode in chemical process. Different from continuous processes, batch process production is not operated in a stable operating point, but a preset trajectory according to raw material ratio and operating conditions. Therefore, process data in batch process display complex characteristics including multi-stage, dynamic time-varying and nonlinear, which means traditional monitoring methods cannot be directly applied to monitor the operating status in batch process. To solve this problem, a novel process monitoring strategy is proposed in this paper. Long short-term memory network is first applied to establish a regression model by extracting the dynamic and nonlinear feature among process variables. Then residuals obtained from the regression model is considered as the monitoring object. Since the multi-stage and dynamic characteristics no longer exist in the obtained residuals, the continuous process monitoring methods can be directly applied to establish a monitoring model for residuals. Finally, the support vector data description is selected to describe the residuals of the normal data set in a hypersphere, and the distance from the test set sample point to the center of the hypersphere is used as a monitoring statistic to detect the faults. The method is applied to a simulated penicillin fermentation process, the results show that complex batch process monitoring can be transformed into a relatively simple continuous process monitoring problem to deal with, and the process faults can be detected in time.
Service
Email this article
Add to my bookshelf
Add to citation manager
Email Alert
RSS
Articles by authors
JI Cheng
GU Junfa
WANG Jianhong
WANG Jingde
SUN Wei
Keywordsdynamic regression model;   residual;   real-time monitoring;   penicillin fermentation process     
Received 2021-08-02;
About author:
Cite this article:   
JI Cheng, GU Junfa, WANG Jianhong, WANG Jingde, SUN Wei.A batch process monitoring method based on long short term memory network and support vector data description[J]  Chemcial Industry and Engineering, 2022,V39(2): 32-36
Copyright 2010 by Chemcial Industry and Engineering