Validation and Uncertainty Quantification of CFD Smooth Seal Models: ANSYS and Bulk-Flow

Turbomachinery faces rotordynamic issues such as imbalance, instability, and large vibrations. Such issues are costly in downtime and maintenance, while also introducing safety concerns. Careful design of components such as bearings and seals can reduce vibrations. Practitioners build computational fluid dynamics (CFD) models of these components to obtain performance predictions. The reliability of such CFD software is typically unknown, and performance studies have shown significant prediction disagreement between different models of the same scenario. Standard literature validation practice lacks rigor and formalization. Furthermore, sufficient experimental data is scarce and difficult to obtain. These constraints invoke the need for a standardized method of validating CFD models for turbomachinery components. In 2009, ASME published a standard for verification and validation of CFD models, V&V-20. This work intends to serve as (1) an assessment of the feasibility of implementing the V&V-20 procedure (2) a baseline in the refinement validation procedure for CFD models of turbomachinery component. Methods from V&V-20 are applied an ANSYS CFX model and a ROMAC model of a straight liquid seal model. One prediction from each model is evaluated in this work. In practice, this procedure would be repeated for other predictions at various input conditions. This study resulted in uncertainty quantification of flow out of the seal, or leakage: experimental uncertainty (5%), numerical uncertainty (0.1% and 0.5%), input uncertainty (16% and 7%), and model error interval (-21% to 13% and -19% to -1%). These quantifications involved a grid convergence study and perturbation study consisting of hundreds of simulations. The experiment used for validation was performed by Kaneko, et al (2003). The study found V&V-20 to be useful but tedious for practical application, especially on a complicated solver such as ANSYS CFX; it serves as a thorough and useful starting point for quantifying uncertainties and standardizing terminology.