Comparison of Experiments to Computational Fluid Dynamics Models for Mixing Using Dual Opposing Jets in Tanks With and Without Internal Obstructions

This paper documents testing methods, statistical data analysis, and a comparison of experimental results to computational fluid dynamics (CFD) models for blending of fluids, which were blended using a single pump designed with dual opposing nozzles in an 8-foot-diameter tank. Overall, this research presents new findings in the field of mixing research. Specifically, blending processes were clearly shown to have random, chaotic effects, where possible causal factors, such as turbulence, pump fluctuations, and eddies, required future evaluation. CFD models were shown to provide reasonable estimates for the average blending times, but large variations—or scatter—occurred for blending times during similar tests. Using this experimental blending time data, the chaotic nature of blending was demonstrated and the variability of blending times with respect to average blending times was shown to increase with system complexity. Prior to this research, the variation in blending times caused discrepancies between CFD models and experiments. This research addressed this discrepancy and determined statistical correction factors that can be applied to CFD models and thereby quantified techniques to permit the application of CFD models to complex systems, such as blending. These blending time correction factors for CFD models are comparable to safety factors used in structural design and compensate variability that cannot be theoretically calculated. To determine these correction factors, research was performed to investigate blending using a pump with dual opposing jets, which recirculate fluids in the tank to promote blending when fluids are added to the tank. In all, 85 tests were performed both in a tank without internal obstructions and a tank with vertical obstructions similar to a tube bank in a heat exchanger. These obstructions provided scale models of vertical cooling coils below the liquid surface for a full-scale, liquid radioactive waste storage tank. Also, different jet diameters and different horizontal orientations of the jets were investigated with respect to blending. Two types of blending tests were performed. The first set of 81 tests blended small quantities of tracer fluids into solution. Data from these tests were statistically evaluated to determine blending times for the addition of tracer solution to tanks, and blending times were successfully compared to computational fluid dynamics (CFD) models. The second set of four tests blended bulk quantities of solutions of different density and viscosity. For example, in one test, a quarter tank of water was added to three quarters of a tank of a more viscous salt solution. In this case, the blending process was noted to significantly change due to stratification of fluids and blending times increased substantially. However, CFD models for stratification and the variability of blending times for different density fluids were not pursued, and further research is recommended in the area of blending bulk quantities of fluids. All in all, testing showed that CFD models can be effectively applied if statistically validated through experimental testing, but, in the absence of experimental validation, CFD models can be extremely misleading as a basis for design and operation decisions.