Model-Independent Approach for Minimum Variance Performance Assessment of a Multivariate Process

• Published in: Industrial & engineering chemistry research Vol. 61; no. 5; pp. 2136 - 2151
• Main Authors: Huang, Chun-Qing, Wu, Zhong-Hao, Huang, Jun-Hao
• Format: Journal Article
• Language: English
• Published: American Chemical Society 09.02.2022
• Subjects: algorithms; chemistry; oils; process design; Process Systems Engineering; variance
• ISSN: 0888-5885; 1520-5045; 1520-5045
• DOI: 10.1021/acs.iecr.1c03218

In general, a plant model or its equivalent information, such as interactor matrix or the first several leading Markov matrices, is indispensable to the minimum variance (MV) performance assessment of a multivariate process. In this paper, a model-independent approach for performance assessment is developed to eliminate the requirement of a plant model. Without relying on any prior plant knowledge, the practical solutions are presented, in which the MV benchmark is obtained directly or is estimated in terms of the upper and lower bounds for the following different levels of available knowledge of the multivariate plant: (1) no knowledge available; (2) a known range of each entry in the pseudo first-Markov-matrix; (3) the available knowledge of the steady gain matrix for a first-order plus dead time (FOPDT) plant. The effectiveness of the proposed algorithm is verified through a classic numerical example and a practical example, namely, a “Shell” oil fractionator.