| Popis: |
The modeling of turbulent flows continues to present a formidable challenge both in industry and academia. There are many examples of commercially available computational fluid dynamics (CFD) codes that have been formulated specifically to handle the unique challenges involved in simulating transient, or time dependent, flows with the goal of predicting the behavior of systems without the need for experimental work. The savings in time and capital cost when designing a new piece of mixing equipment can be significant when using computational methods, making this approach highly attractive. However, it is important for the user of such software to be confident that the output accurately reflects how the system will behave in reality. Without confidence in the resulting data, it is difficult to justify any particular course of action based on information gained through the computation. For this reason, no model is complete without some degree of validation to ensure that the results reflect what will actually happen. However, the sort of data necessary to fully validate such a model is difficult to obtain because any flow field we wish to study is inherently three-dimensional and time dependent. It is reasonable to expect that the data used in validating the model should also be three-dimensional and have a time component.In this work, we present a statistical validation of a direct numerical simulation (DNS) model as applied to an opposed-jet mixing vessel. The data used for validation was obtained by three-dimensional particle tracking velocimetry (PTV) and is compared with a simulation of the same system under conditions identical to those under which the experiments were performed.The results indicate that, in some cases, there is good agreement between the main velocities measured in the experiments and those calculated by our DNS code. However, the amount of inconsistency in higher order statistics combined with unusual flow behavior in the experimental data leads to the conclusion that there are problems with the experimental data itself. Nonetheless, this work represents a new approach to validation, as this type and amount of data have never before been used for validation purposes. |
