Multiple Hypotheses Testing-Based Operating Optimality Assessment and Nonoptimal Cause Identification for Multiphase Uneven-Length Batch Processes

详细信息    查看全文

• 作者：Yan Liu ; Fuli Wang ; Yuqing Chang ; Ruicheng Ma ; Shumei Zhang
• 刊名：Industrial & Engineering Chemistry Research
• 出版年：2016
• 出版时间：June 1, 2016
• 年：2016
• 卷：55
• 期：21
• 页码：6133-6144
• 全文大小：570K
• 年卷期：0
• ISSN：1520-5045

文摘

As a novel research issue, the operating performance assessment for industrial processes has received growing interest in recent years. In this study, a multiple hypotheses testing-based operating optimality assessment and nonoptimal cause identification method is proposed to handle the online assessment of the batch processes with multiphase and uneven-length characteristics. A new Gaussian mixture model (GMM)-based offline phase division procedure is proposed. Compared to the existing methods, it can not only ensure that each phase contains the consecutive sampling instants but can also retain the characteristics of phase-uneven-length for retaining more-accurate phase information. The extra post-processing is not needed in phase division, and the assessment models of each phase are set up along with it. In online application, both the local and global online assessments are performed for a new batch based on the multiple hypotheses testing technology by controlling the false discovery rate. For the nonoptimal batch, the cause variables can be determined based on the variable contributions to the assessment index. Finally, the effectiveness of the proposed method is verified through a fed-batch penicillin fermentation process.