Hidden Semi-Markov Probability Models for Monitoring Two-Dimensional Batch Operation

详细信息    查看全文

• 作者：Junghui Chen ; Yan-Cheng Jiang
• 刊名：Industrial & Engineering Chemistry Research
• 出版年：2011
• 出版时间：March 16, 2011
• 年：2011
• 卷：50
• 期：6
• 页码：3345-3355
• 全文大小：1120K
• 年卷期：v.50,no.6(March 16, 2011)
• ISSN：1520-5045

文摘

The repetitive batch operation has a two-dimensional dynamic behavior, including a finite time interval of each batch run in the time domain and infinite repetitions along the batch domain. In this Article, a novel monitoring method that combines dynamic multiway principal component analysis (DMPCA) and hidden segmental semi-Markov models (HSMM) is proposed to resolve the problem caused by the two-dimensional behavior of batch processes. DMPCA utilizes the batch-to-batch dynamic characteristics and eliminates the batch correlation among process variables. HSMM is used to construct the temporal behavior among process variables during each batch run. By constructing a two-dimensional model, the proposed method can generate simple probability monitoring charts and monitor the progress in each batch run. The proposed method has the temporal property of HSMM and the batch-to-batch dynamic characteristics of DMPCA. Its advantages are demonstrated through a simulated fed-batch penicillin cultivation process characterized by fault sources.