Operating optimality assessment and nonoptimal cause identification for non-Gaussian multimode processes with transitions

Generally, process operating performance affects the economic benefits of enterprises, which prompts the development of process operating performance assessment. In this study, a novel operating optimality assessment and nonoptimal cause identification strategy is proposed for non-Gaussian multimode processes. Compared with the reported methods, the advantages of the proposed method are: (i) non-Gaussian distribution data is firstly considered in operating optimality assessment; (ii) besides the online assessment strategy for stable mode, that for transitional mode is developed for the first time; (iii) a novel online mode identification method is developed based on the sample posterior probabilities together with a predefined threshold, which can not only improve the accuracy of the assessment results but also reduce the risk of misclassification; (iv) the nonoptimal cause identification approach is established to identify the variables responsible for the nonoptimal performance. Finally, the effectiveness of the proposed method is demonstrated by Tennessee Eastman (TE) process.