The numerical models of the pump-turbine and plant concrete in this paper include the pump–turbine flow model, the preloading spiral case structural model and pump-turbine and plant concrete structural model. The models will perform in sequential order to simulate the construction and operation process in a PSSC.
3.1. Numerical Simulation Flowchart
The numerical simulation of the fluid and structure models should be performed under a certain work step. To take the preloading clearance and contact nonlinearity into consideration, the preloading spiral case model is first established to simulate the initial preloading clearance between the PSSC and concrete. Nodes on the interfaces between these two structures in the FSI model are set to be separated due to the clearance, which will change when under hydraulic excitation. The hydraulic pressure load is calculated by steady and transient CFD simulation of the unit flow passage model. After Fast Fourier Transformation (FFT), the time-frequency distribution of the load can be used to analyze the hydraulic pressure fluctuation characteristics. The hydraulic pressure load per timestep is applied on the locations associated with the structural mesh of the unit and the plant concrete model based on the FSI method. By introducing the contact algorithm, the structural dynamic response and evolution mechanism of preloading clearance can be simulated for the analysis of the hydrodynamic characteristics and contact status of the PSSC and concrete.
To ensure the grid nodes of PSSC correlate with that of the concrete, the FEM model of the entire structure, including all the components of the two structural models, should be meshed together first. Other components and concrete in the preloading spiral case model are suppressed. Then, its solution part is applied to the unit and plant concrete model, in which the previous spiral case and mandoor are replaced by the deformed one. There will be an initial preloading clearance between the PSSC and concrete in the unit and plant concrete model.
The entire flowchart of the simulation process is shown in
Figure 1 and more details of the model establishment are described in the following sections.
3.2. Pump-Turbine Flow Model
The prototype pump-turbine model is analyzed, and its 3D modeling of the flow passage is composed of a spiral case, stay vanes, guide vanes, a runner, a draft tube and two pressure balance pipes. In addition, the clearances, such as the headcover and bottom ring, are also considered. A penstock in front of the spiral case is set to reduce the numerical simulation error of the spiral case pressure fluctuation influenced by the inlet boundary conditions. The mesh of the CFD modeling of the flow passage is shown in
Figure 2 and
Figure 3. The basic parameters of the unit are shown in
Table 1.
where
is the safety factor, ranging from 1.25 to 3.00 and taken as 1.25 when three or more grids are applied, and the subscript
b represents different grid schemes, where a larger value of
b corresponds to a finer grid.
is the grid refinement ratio, and the expression is as follows:
where
is the number of grid elements,
is the computational dimension.
is the relative error of the numerical calculation results using two sets of grids with adjacent quantities:
where
is the normalized efficiency for the numerical discrete solution of the selected convergence parameter.
The fluid domain is meshed by the tetrahedron–hexahedron hybrid elements based on ANSYS MESH 2021. The spiral case, runner, draft tube, pressure balance pipe and penstock are modeled using tetrahedral elements, while the vane space, headcover and bottom ring are modeled using hexahedral elements. There is grid refinement near the nose vane of the spiral case where swirling flow may occur. Considering the flow in the near-wall region, a boundary layer mesh is set with five layers and a first layer thickness of 1.0 mm under a 1.2 growth rate at the near wall of vanes and blades, ensuring that the y+ meets the requirements of the SST k-
model. Four sets of meshes with different element sizes are constructed to verify the mesh independence and the calculation accuracy by comparing the results of normalized efficiency under the rated condition in
Figure 4. The grid convergence index (GCI) is calculated to quantitatively verify the reliability of the mesh, which is related to the approximation error and the true error [
33,
35,
36]. The expressions of GCI and the relative errors
are shown above. As is shown in
Table 2, the GCI of 7.25 × 10
6 mesh setting is 0.0345%, which is much less than 5%, indicating that the mesh setting has met the accuracy requirement. As a result, the mesh size of the spiral case, stay vanes, guide vanes, runner, draft tube, pressure balance pipe, headcover, bottom ring and penstock can be adjusted with 1,453,828, 543,760, 431,608, 2,791,382, 1,091,405, 260,108, 134,800, 158,200, and 388,136, respectively. A total of 7,253,227 elements are finally adopted in the pump-turbine flow passage model.
To exactly describe the pressure variation in the fluid domain, the locations of a dozen monitoring points are shown in
Figure 5, wherein SC1–SC4 are located at the spiral case, VS1–VS4 are located in the inter-vane space of stay vanes, and NV1–NV4 are located in the vaneless space.
ANSYS CFX 2021 is used to carry out the 3D CFD simulation under the turbine’s rated conditions. Since pump-turbines generally operate close to their rated conditions with stable flow characteristics, the research on the turbine’s rated conditions can exclude disturbances caused by hydraulic instability when analyzing the impact of preloading pressure and thus have better practical application benefits. The turbulence model is set as shear stress transport (SST) with a convergence criterion of 10
−4. According to the rated rotational speed of the runner (375 rpm), the total time is set to the duration required for ten revolutions (1.6 s), and the time step is 8.889 × 10
−4 s corresponding to the time needed for a 2° rotation. The total pressure type inlet as well as the static pressure outlet are applied to the model boundary condition. The rotor–stator interfaces between the runner and guide vanes and runner and draft tube are set to the Transient Rotor Stator [
37]. The wall boundary condition is set as a non-slip wall. For a rotation modeling, the Transient Rotor Stator is used to account for transient interaction effects at a sliding (frame change) interface. It simulates the transient relative motion between the components on each side of the General Grid Interface (GGI) connection, updating the interface position at each timestep as the grid changes. This model is widely used to predict the transient flow interaction between the stator and rotor passage.
To validate the accuracy of the simulation method, we compared the simulation results with those from the model test. The whole model of the pump-turbine has been fabricated, which includes the flow passage of the spiral case, stay ring, stay vanes, headcover, guide vanes, runner, bottom ring, and draft tube, as shown in
Figure 6. None of the flow surfaces of the turbine model are coated with any paint or varnish. The flow passage of the pump-turbine model, from the spiral case inlet to the draft tube outlet, is geometrically similar to the prototype pump-turbine, with a scale ratio λ
L = D
P/D
M of 9.9. Other basic parameters are listed in
Table 1. Except for the straight conical section of the draft tube, which is made of transparent acrylic to facilitate flow observation, all other components are made of metal.
The rated operating condition is simulated in this paper, and the parameter settings are illustrated in
Table 3. Compared with the model test results, the calculation error is less than 3%, which proves the reliability of the numerical methods used in this paper.
3.3. Preloading Spiral Case Structural Model
The preloading spiral case structural model is first developed to calculate the initial preloading clearance between the PSSC and concrete, which encompassed the PSSC (with mandoor), stay ring, test head and steel pedestal, as shown in
Figure 7a. The structural components share the same meshes with the pump-turbine and plant concrete structural model mentioned above, so the mesh setting will be introduced in the following section in detail. A total of 2,016,116 elements and 3,858,253 nodes are used in this model, ensuring adequate grid density for accuracy.
The material properties of the components are configured according to
Table 4.
During the actual construction, the spiral case will be deformed in an expansion way as the preloading pressure is applied to the inside. The concrete is placed around the PSSC later to maintain pressure. At that time, the inner surfaces of the concrete coincide with the ektexine of the PSSC. After the concrete has already hardened, the preloading pressure inside the spiral case begins to release, during which the PSSC shrinks back to its original shape and separates from the inner surfaces of the concrete. There will be an initial clearance between the PSSC and concrete after all the above steps have been completed. However, the inner surface of the concrete should adjust to the deformation of the spiral case in the simulation procedure by changing grid node coordinates or performing Boolean operations with grid models. This may lead to model penetration and stress concentration, resulting in a non-negligible decrease in accuracy.
As a result, the simulation methods proposed in this paper are used to apply negative pressure to the inside of the spiral case and guarantee the same absolute deformation value as that of the preloading pressure, as shown in
Figure 7b. The spiral case shrinks instead of being inflated, and thus a deformation equivalent to the initial preloading clearance can be achieved without modifying the concrete model. Compared to the PSSC diameter, the value of the deformation varies in the range of millimeters and has almost no impact on the subsequent analysis, whether positive or negative. In this way, the new methods can not only improve precision but also simplify the simulation process significantly.
Three different preloading schemes have been chosen for the discussion on the impact of the preloading clearance. Based on the engineering experience, it is determined that the 0.5, 0.7 and 1 multiples of the maximum static head can be regarded as representative preloading pressures, which are 3.2 MPa, 4.564 MPa and 6.52 MPa, respectively, as shown in
Table 5.
3.4. Pump-Turbine and Plant Concrete Structural Model
The 3D modeling of the pump-turbine and plant concrete structural model consists of the deformed PSSC (with mandoor), stay ring, headcover, bottom ring and plant concrete in
Figure 8, wherein the PSSC and stay ring are bolted together as a single structure. To better accommodate the complex geometries, the structural models are meshed using high-quality tetrahedral elements. The mesh is refined in typical stress concentration regions of the guide vanes and nose vane. The adaptive sizing of the mesh is set at a resolution of 2, which is used to control the mesh size, and its valid inputs are integers between 1 and 12. The model uses fast transition meshing and coarse span angle center. Considering the difference of the internal and external force transmission of the spiral case, a two-layer grid is set on the spiral case in a radium direction. Also, the contact region of the concrete is meshed by the same element size as the spiral case, ensuring the accuracy of passing data.
Four sets of mesh with different element numbers are plotted for the structural model in a mesh independent analysis, including 1.589 × 10
6, 2.996 × 10
6, 3.505 × 10
6, 4.204 × 10
6 elements.
Figure 9 presents the simulation results of normalized stress for four different sets of meshes. The element number of the first set is not so high that singularities exist and cause an abnormal high stress. Compared to the normalized stress, the mesh with 2.996 × 10
6 element numbers is the best choice for further analysis to save calculation time and ensure accuracy.
The boundary conditions of the structural model for fluid–structure coupling are shown in
Figure 10. Standard earth gravity is applied to the structures. The hydraulic pressure distributions exported from the CFD analysis are applied to the corresponding structural components at different time points. Each pressure file includes the 3D coordinates and the pressure values of each mesh node in the form of (x, y, z, pressure) on the fluid–structure coupling interface. The initial preloading clearance has been simulated in the preloading spiral case structural model, so the connection surfaces between the spiral case and the concrete are designated as the nonlinear contacts. Considering the potential separation and sliding behaviors, the friction type is introduced to the contact model. The friction coefficient is defined as 0.25, and the interface treatment is set as ‘adjust to touch’ for simulating the nonlinear contact behaviors. The remote displacement of the headcover restricts one face’s movement in the x and y direction. The lower part of the bottom ring is bonded with the concrete. Considering the constraint from the mountain, the outer surfaces of the plant concrete are assigned as fixed support.