Lattice-Boltzmann-Based Algorithms for Predicting Impeller Performance

A novel approach for performing large-eddy simulations (LES) of industrial-sized mixing processes is presented. This approach, which is an extension of the lattice-Boltzmann model (LBM), is trivially parallelizable and runs such simulations orders-of-magnitude faster than conventional CFD tools.  The mixing properties of a Rushton turbine, as predicted from the LBM model, are compared to experimental data. The vortex structure, eddy-dissipation rates, pumping number, and power number predicted from the model agree with experimental data across multiple Reynolds numbers.  Since fluid-structure interactions are described using an immersed boundary method, system set-up is trivial and no explicit meshing process is required.   As such, simulations involving more sophisticated impeller morphologies present no meaningful increase in system set-up and runtime.  Operating on modern computer architectures (e.g., GPUs, Xeon Phi, etc.), the model can simulate multiple seconds of a mixing per hour of processing time.