Hybrid modeling supported development of an industrial small-molecule flow chemistry process. (February 2023)

• Record Type: Journal Article
• Authors: Polak, Jakub; von Stosch, Moritz; Sokolov, Michael; Piccioni, Lorenzo; Streit, Alexander; Schenkel, Berthold; Guelat, Bertrand
• Is Part Of: Computers & chemical engineering. Volume 170(2023)
• Publication Date: 2023-02
• Subjects: Multivariate Data Analysis -- Hybrid modeling -- Plug Flow Reactor -- Flow Chemistry -- Continuous manufacturing; Chemical engineering -- Data processing -- Periodicals; 660.0285
• DOI: 10.1016/j.compchemeng.2022.108127
• Languages: English
• ISSNs: 0098-1354

Abstract

Highlights: Hybrid model captures impact of raw material variations and process parameter changes on flow chemistry process behavior. Model validity constrained process optimization yields process conditions for which the model predictions are accurate. Proposed approach inherits great potential for Quality by Design guided process development. Abstract: Continuous manufacturing of small molecules finds increasing application in developing novel processes due to its significant advantages. The development of these processes is increasingly being supported by models, in a quest to reduce the number of experiments and increase insight. With the Quality by Design paradigm in mind, answers to questions anent the impact of changes in process parameters and material attributes on the process performance and product quality are sought to be answered. In this contribution, a hybrid model is used to describe the impact of raw material variations and process parameter changes on the product quantity and quality of flow chemistry process. The process model is then used for process optimization aiming at maximizing the product concentration. During the optimization, a constraint that accounts for the risk of model misprediction is considered to explore only process conditions for which the model can describe the process behavior. The obtained optimal conditions are in agreement with expectations and underline the potential for hybrid model-based process development.