Dynamic and Statistical Operability of an Experimental Batch Process

Abstract

The operability approach has been traditionally applied to measure the ability of a continuous process to achieve desired specifications, given physical or design restrictions and considering expected disturbances at steady state. This paper introduces a novel dynamic operability analysis for batch processes based on classical operability concepts. In this analysis, all sets and statistical region delimitations are quantified using mathematical operations involving polytopes at every time step. A statistical operability analysis centered on multivariate correlations is employed for the first time to evaluate desired output sets during transition that serve as references to be followed to achieve the final process specifications. A dynamic design space for a batch process is, thus, generated through this analysis process and can be used in practice to guide process operation. A probabilistic expected disturbance set is also introduced, whereby the disturbances are described by pseudorandom variables and disturbance scenarios other than worst-case scenarios are considered, as is done in traditional operability methods. A case study corresponding to a pilot batch unit is used to illustrate the developed methods and to build a process digital twin to generate large datasets by running an automated digital experimentation strategy. As the primary data source of the analysis is built in a time-series database, the developed framework can be fully integrated into a plant information management system (PIMS) and an Industry 4.0 infrastructure.