Développement de stratégies numériques basées sur la DEM et la CFD-DEM améliorant la performance computationnelle de simulations d'écoulements granulaires et polyphasiques

The design and operation of industrial units require an understanding of the behavior of powders and particles, which often defy our intuitions based on liquids and gases. Inter-actions between particles involve several considerations, as do their interactions with fluid in multiphase flows. The Discrete Element Method (DEM) is among the most popular and suitable numerical methods to simulate these flows. DEM tracks particles distinctively by the law of motion and models collision interactions using force-displacement relationship albeit at significant computational cost. This method is often coupled with a fluid mechanics solver (CFD-DEM) for multiphase flow simulations. Given that simulations of industrial systems in-volve a considerable number of particles, up to several millions, the use of speedup approaches is justified. Parallelization strategies, graphics processing unit (GPU)-based methods, and algorithmic techniques were proposed to enhance DEM and CFD-DEM performance. How-ever, high-performance computing hardware is not available to everyone and can be a costly investment. Also, one of the limiting points of current algorithmic methods is that they are sometimes limited to certain specificities of the simulated case. The main objective of this project is to develop a novel method to improve computational performance of granular and multiphase flows applicable to DEM and CFD-DEM. The core hypothesis revolves around identifying quasi-static regions where particle velocities exhibit low variance among neighboring particles through granular temperature evaluation. In these areas, particle displacements resulting from collisions would not affect flow macroscopic be-havior, suggesting that computational resources should ignore contact forces and particle position computation. The mechanism was developed to reduce the computational burden of the DEM by minimizing the calculation arising from particle collisions while maintaining full calculation in critical regions. It is applicable to DEM and also to CFD-DEM solvers with the addition of an averaged advection term to support bulk motion. The adaptive disabling of particle contacts in dynamic systems creates sparse inactive areas that migrate or vanish during the simulation. This phenomenon led to the algorithm being named Adaptive Sparse Contacts (ASC).