Deagglomeration in turbulent flows: Particle-resolved investigation for detailed insights and advanced modeling

• Applicants: Professor Dr.-Ing. Michael Breuer; Dr.-Ing. Ali Khalifa
• Subject Area: Fluid Mechanics ; Mechanical Process Engineering
• Term: since 2026
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 573724408

This project has two main objectives: 1. To investigate the physics of fluid-induced breakup of cohesive particle agglomerates in the micron-size range within turbulent flows. 2. To develop an advanced multiscale model for describing this phenomenon grounded on the underlying physics. The goals are both based on high-resolution, particle-resolved simulations with an unprecedented level of detail. For that purpose, a large number of fully-resolved DNS-DEM simulations (Direct Numerical Simulation – Discrete Element Method) based on the immersed boundary method will be conducted to capture the detailed fluid–particle and particle–particle interactions across a wide range of breakup scenarios. That includes variations in the fluid flow (different relative velocities, rotation rates, and shear gradients), the agglomerate properties (size and shape of the agglomerate), and the primary particle characteristics (size of primary particles and strength of cohesive force). These simulations will enable a thorough physical analysis of the deagglomeration process. By examining the influence of local flow structures, fragmentation modes, and detachment directions, the study will uncover the dominant mechanisms governing breakup of particle agglomerates. This analysis will also help to identify key parameters and serve as a basis for evaluating existing breakup models. To translate the gained physical insights into predictive capability, a novel breakup model will be developed based on machine-learning techniques. In a first step different machine-learning approaches will be evaluated with regard to their suitability for this specific task, with the ability to generate human-interpretable mathematical relationships being a crucial criterion. Symbolic regression is presently seen as a promising candidate for this purpose. Trained on the high-fidelity datasets, the resulting breakup model will capture the essential features of breakup dynamics. That includes the onset of breakup, the mode of breakage, and the subsequent motion of the arising fragments. Unlike theory-based breakup models, the database will provide results for breakup cases involving various types of stresses simultaneously allowing to develop a unified model capable of describing all fluid-induced breakup scenarios. The model will be integrated into an efficient Euler–Lagrange framework for large-eddy simulations, where agglomerates are represented as effective point particles, enabling the simulation of flows with high mass loadings at reasonable computational costs. Finally, the extended framework will be applied to representative real-world systems, such as lab-scale deagglomerators, inhalers, and cyclone separators, to demonstrate its accuracy and practical relevance. This project will significantly enhance the understanding and modeling of cohesive particle-laden flows by combining detailed physical insight with the development of an interpretable, data-driven breakup model.

• DFG Programme: Research Grants