Data Mining in der Prozesstechnik als Key-Enabler intelligenter Digitaler Zwillinge für eine datengetriebene Optimierung der Prozessführung

This paper shows how similarities between process variables can be described by data mining on historical process data. As an application example Dynamic Crossflow Filtration (DCF) is considered, which is used to clarify wine or to produce protein powder from protein-con-taining suspensions. Characteristically, the filtration processes in these plants consist of indi-vidual production phases (batch processes) between which cleaning programs of the plant and the filter run. The cost of extensive cleaning must be weighed against the benefit of sub-sequent separation phases to optimize plant downtime, product quality and resource utiliza-tion. The presented data mining approach can be used to identify dependencies in coupled, sequential batch and cleaning processes.