Troubleshooting high-pressure issues in an industrial biorefinery process by feature-oriented modeling

Biorefinery plants manufacture chemicals by conversion of sustainable raw materials. Membrane separation processes are commonly used to separate cells from fermentation broths, and account for most of the operating costs of a biorefinery process. Membrane fouling can disrupt normal operation, increasing cleaning costs and process downtime (hence, overall manufacturing costs). Whereas investigating the causes of fouling through mechanistic models can be challenging in an industrial environment, information in process data historians can be leveraged through data-driven modeling. In this study, principal component analysis and feature-oriented modeling are combined to identify potential causes of fouling in a semi-continuous membrane separation process of an industrial biorefinery. This approach can effectively address batch duration variability issues, while exploiting process knowledge to enhance information on effects of fouling. Membrane age and operating temperature were found to be the major variables related to fouling.