Research

Intensification of dynamic crossflow filtration processes in the food industry using hybrid digital twins

Crossflow filtration (CF) is a state-of-the-art technology in the food industry and intensively used e.g. in dairy processing. Tangential flow over the membrane and a high flow velocity prevent the build-up of a filtration cake to a large extent. In dynamic cross-flow filtration (DCF), rotating membrane discs allow to further increase the possible retentate concentration. Up to now, process and cleaning parameters in industrial plants are often adjusted once and fixed recipes are applied to run the processes. Thus, the potential of the processes is often not fully exploited.

Therefore, hybrid digital twins will be developed in this project to optimize two existing CF/DCF plants. These hybrid digital twins consist of a combination of different models that will approximate the real plants as close as possible. Mechanistic models, i.e. models based on fundamental physical and chemical laws, are combined with data-driven models, that use historical data from the process plants. An algorithm will use the models to optimize process parameters in real-time.

The mechanistic models will be constructed using the simulation software COMSOL Multiphysics, further programming tasks will be carried out with Matlab and Python. In addition, a laboratory-scale DCF plant with an automated data logging system will be set up at IFG for the validation of the models using different feed streams (wine, chicken broth) and for the investigation of extreme process conditions.