Gaussian Process Regression and Bayesian Inference Based Operating Performance Assessment for Multiphase Batch Processes

doi:10.1021/acs.iecr.8b00234

Return

Gaussian Process Regression and Bayesian Inference Based Operating Performance Assessment for Multiphase Batch Processes

2018-05-03

PRE

DOI：10.1021/acs.iecr.8b00234

Original

Original request for help

Save

Abstract

En 中文

Batch processes have been playing a significant role in modern industrial processes. However, even if the operating conditions are normal, the process operating performance may still deteriorate away from optimal level, and this may reduce the benefits of production, so it is crucial to develop an effective operating performance assessment method for batch processes. In this study, a novel operating performance assessment method of batch processes is proposed based on both Gaussian process regression (GPR) and Bayesian inference. It is committed to solving the challenges of multiphase, process dynamics and batch-to-batch uncertainty that are contained in most of batch processes. To characterize different dynamic relationships within each individual phase, multiple localized GPR-based assessment models are built first. Furthermore, the phase attribution of each new sample is determined, and two different identification results are obtained, i.e., a certain interval and a fuzzy interval between two adjacent phases. Then different online assessment strategies are designed correspondingly. When the operating performance is nonoptimal, cause variables are identified by variable contributions. Finally, the effectiveness of the proposed method is demonstrated by the fed-batch penicillin fermentation process.

Keywords:

NONOPTIMAL CAUSE IDENTIFICATION

ONLINE MONITORING STRATEGY

FAULT-DIAGNOSIS

OPTIMALITY ASSESSMENT

PCA

MULTIMODE

MODEL

MULTIBLOCK

PREDICTION

FRAMEWORK

AI Summary

Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.