| Reference : A Dual-Grid Multiscale Approach to CFD-DEM Couplings for Multiphase Flow |
| Dissertations and theses : Doctoral thesis | |||
| Engineering, computing & technology : Multidisciplinary, general & others | |||
| Computational Sciences | |||
| http://hdl.handle.net/10993/36897 | |||
| A Dual-Grid Multiscale Approach to CFD-DEM Couplings for Multiphase Flow | |
| English | |
Pozzetti, Gabriele  [University of Luxembourg > Faculty of Science, Technology and Communication (FSTC) > Engineering Research Unit >] | |
| 20-Sep-2018 | |
| University of Luxembourg, Luxembourg | |
| Docteur en Sciences de l'Ingénieur | |
| 193 | |
| Peters, Bernhard | |
| Zilian, Andreas | |
| Bordas, Stéphane | |
| Mahmouidi, Amir | |
| Useldinger, Ralph | |
| Varrette, Sébastien | |
| [en] Multiscale ; CFD-DEM | |
| [en] This thesis focuses on a novel dual-grid multiscale approach to CFD-
DEM1 couplings, proposes its advantages in terms of numerical proper- ties and performance, and provides examples of engineering applications that can benefit from it. In recent years, CFD-DEM couplings are be- coming a more and more adopted solution for the numerical simulation of particle-laden flows. In particular, couplings based on the volume av- eraging technique have become a standard for numerical simulations in chemical and process engineering. Furthermore, they are rapidly spread- ing to civil, geotechnical and mechanical applications due to their ability in dealing with arbitrarily complex mixtures of continuum and granular media. Despite the several advantages that these Eulerian-Lagrangian cou- plings provide, their rigorous application to complex scenarios is currently limited by two main factors. First, the computational traceability of the solutions can become problematic due to the lack of a general theory on the subject. In particular, grid-convergence studies for the solution of the continuous phases are often not feasible due to the averaging procedure that imposes limitations on the grid structure and refinement. Second, the parallel implementation of these numerical schemes holds important disadvantages in terms of memory consumption and inter-physics com- munication load. These disadvantages are significantly limiting the ex- tension of these approaches to large-scale scenarios. This thesis collects some of the most significant works published in the last years on a novel approach that allows solving the two above- mentioned problems, and, therefore, tackling more complex and expen- sive scenarios. I refer to this approach as dual-grid multiscale approach for CFD-DEM couplings. It consists in using two different computational grids, one for the coupling between continuum and discrete entities and one for the solution of the so-obtained continuum equations. The two grids, i.e. the two problems, are in this way resolved on two different scales. The first scale or “bulk” scale is chosen to optimize the averag- ing operation. At this length-scale, the discrete entities are considered as zero-dimensional, and interact with the fluid with local exchanges of momentum, mass, and energy. The second scale or “fluid-fine” scale is identified as the one at which a unique solution for the averaged equa- tions can be provided. In practice, this is chosen as the one at which the solution of the fluid equations becomes grid-independent. An inter-scale communication is adopted by interpolating fields from the fluid-fine scale to the bulk one and vice-versa. The theoretical description of the method is first provided with par- ticular reference to the DEM-VOF coupling. Even in its simplest version, the multiscale approach is shown to generate grid-convergent solutions and significantly higher accuracy than a standard CFD-DEM coupling. This shows how the new approach is able to overcome the first main limitation described above. Then, an optimized parallel implementation of the method is pro- posed to show how this multiscale approach can provide significant ben- efits also for what concerns the execution time. Technically, this is made possible by moving the communication cost of the coupling from the inter-physics communication that characterized the standard CFD-DEM couplings to an optimized inter-scale communication routine. This en- ables the method to overcome a major bottleneck of the parallel execution of CFD-DEM couplings and therefore the second main limitation of those schemes. Finally, the dual-grid multiscale method is applied to approach in- dustrially relevant problems that were till now out-of-reach for standard CFD-DEM couplings, proving how this technique can have direct real- case application and produce immediate benefits for practitioners willing to adopt it. | |
| Multiscale_CFD-DEM | |
| http://hdl.handle.net/10993/36897 |
| File(s) associated to this reference | ||||||||||||||
|
Fulltext file(s):
| ||||||||||||||
All documents in ORBilu are protected by a user license.
