Abstract Summary (Max 250 words)
Mixing time prediction in stirred tank reactors (STRs) remains computationally demanding when using conventional CFD approaches, limiting their practical application for design optimization and sensitivity studies. This work presents the first experimental validation of recurrence CFD (rCFD) methodology for mixing analysis in a 30L STR equipped with a dual-impeller configuration (pitched blade turbine and Rushton turbine). The rCFD approach exploits the pseudo-periodic nature of impeller-driven flows to extrapolate passive tracer transport using a pre-computed flow field database, bypassing repeated solution of the Navier-Stokes equations during the extrapolation phase. Validation was performed by comparing rCFD predictions against lattice Boltzmann method (M-STAR), finite volume method (Fluent), and experimental conductivity probe measurements at a fixed injection location. The 95% mixing time criterion was evaluated at two probe volumes and across the overall domain. Computational benchmarking revealed that rCFD requires approximately 6 seconds of clock time per second of simulated process time, compared to 240 seconds for M-STAR and 10,200 seconds for Fluent representing a speedup of nearly four orders of magnitude over conventional FVM. This computational efficiency enabled a parametric study encompassing 90 injection configurations (10 angular positions at 36° spacing, 3 radial distances, and 3 heights), a scope infeasible with traditional CFD. Results identify optimal injection locations for minimizing mixing time and reveal the sensitivity of mixing performance to injection position. The findings demonstrate rCFD as a viable tool for rapid screening and optimization of mixing processes in STR applications.
