Artificial filter cake generation: Digital twins via stochastic 3D modeling based on μ -CT image data

In order to link the properties of a feed stream to those of a product in separation operations such as cake filtration, a comprehensive database is required which often cannot be achieved through laboratory measurements alone. For this reason, a stochastic 3D model for the generation of virtual filter cake structures is developed and calibrated to tomographic image data of experimentally built filter cake structures. In this way, digital twins of real particles can be simulated and spatially arranged to form a three-dimensional artificial filter cake. Its 3D morphology is validated with respect to geometric descriptors that were not used for model fitting, such as tortuosity, constrictivity and specific surface area of pore space. In future work, using the model developed in the present paper, a large database of systematically varied artificial filter cakes will be generated by adjusting interpretable model parameters, allowing for structure–property relationships to be statistically investigated. • Linking feed stream and product properties requires a comprehensive database. • Artificial structures are generated by means of a data-driven stochastic 3D model. • The model allows for systematic variation in particle size and shape distribution. • A diverse data basis can be generated and used for large-scale simulation studies. • Framework for quantifying the influence of particle shapes on pore space morphology.