Multimode Process Monitoring Based on Mode Identification

作者：Tan Shuai^*; Wang Fuli; Peng Jun; Chang Yuqing; Wang Shu

来源：Industrial & Engineering Chemistry Research, 2012, 51(1): 374-388.

DOI：10.1021/ie102048f

Many industrial processes have multiple operation modes due to different manufacturing strategies or varying feedstock. Fault detection for a multimode process is a complex problem, as monitoring for both stable and transitional modes should be taken into consideration. In this paper, a novel method based on the similarity of data characteristics is proposed to realize mode identification for modeling data. Different models are developed to capture the major tendencies of process variables. Especially, the transitional regions between neighboring stable modes, which have their particular dynamic characteristics, are modeled, respectively. Online monitoring procedures are formulated on the basis of mode identification. It is more efficient than a model matching strategy using traversing method. At last, the efficacy of the proposed method is illustrated by applying it to a continuous annealing line process and the Tennessee Eastman process. Both results of real application and simulation clearly demonstrate the effectiveness and feasibility of the proposed method.

• 出版日期2012-1-11
• 单位东北大学