On the autonomous validation and comparison of particle models for a Newtonian laminar flow mixing model using PEPT

A methodology for the validation of a Computational Fluid Dynamics (CFD) mixing model using Positron Emission Particle Tracking (PEPT) was applied to a horizontal laminar stirred tank containing Newtonian glycerol. Particle trajectories recorded using PEPT were compared to three one-way particle-coupled CFD models (CFD-DEM, and CFD-Lagrangian, and CFD-Massless) through time-averaging and binning into a grid of 3-D voxels. A cell-by-cell comparison of particle velocities between the computational models and PEPT highlighted the accuracy of the CFD-DEM and CFD-Massless models over the Lagrangian model. Furthermore, using PEPT as a validation dataset the Pearson r-squared was used as a cost function to guide an autonomous calibration of a 5-dimensional DEM parameter space. A novel statistic, the simulation quality coefficient Qsim, was applied to assess accuracy between trajectory datasets and revealed the extent to which a model can be considered calibrated. Considering the balance between computational cost and precision, the CFD-massless model was the most effective particle model for this validation workflow. Further investigations will apply this methodology to calibrate non-Newtonian laminar CFD models.