To validate model accuracy, aside from manual calculation, a series of sensitivity analyses is performed by changing one input parameter while keeping the rest of the parameters unchanged (see below). The results of the sensitivity cases are compared to the base model for difference.
Grid independent analysis
The mesh quality is important to the accuracy and convergence of the model, therefore, a grid independent analysis is carried out by checking the model with all unchanged parameters, except the meshing methodology and element size, which were varied. The mesh element quality is checked against orthogonality, aspect ratio, expansion rate and other criteria.
A separate simulation using the Shear Stress Transport (SST) turbulence model is created to compare with the base case (i.e., k-ε turbulence model) for pressure distribution, velocity field, turbulence kinetic energy, etc.
Different boundary conditions are assigned to the model one at a time to see how sensitive the flow field will be with regard to a certain boundary condition. These sensitivity cases include varying flow rates at each of the turbines, changes in pipe roughness, changes in the size and amount of suspended sediment, etc.
CFD modeling is an efficient way to analyze the hydraulic performance of the penstock and other hydraulic structures / equipment (e.g., turbines, valves, etc.). It provides reliable and detailed information regarding the flow characteristics for the designer to optimize the geometry and improve hydraulic efficiency. It also predicts the erosion on the structure surfaces due to the suspended sediment in the river, which helps the designer address the long-term maintenance issue during the design stage.
Additionally, this CFD model can be coupled with a mechanical model (i.e., another numerical 3D model to analyze the stress and strain of the structure) of the manifold in Ansys. In that case, both the hydraulic model and mechanical model will share the inputs and results (i.e., the deformed shape of the structure from the mechanical model will be the geometry input for the hydraulic model, and the pressure distribution calculated from the hydraulic model will be input force for the mechanical model). Both models will run in parallel until they reach the convergence at the same time, which will reveal a more realistic structural and hydraulic behaviour of the manifold than if done separately. This fluid-structure interactive (FSI) model will be presented and discussed in another paper.