2.1. Mathematical and Discrete Models
Vortex heat transfer during the turbulent motion of a viscous incompressible fluid in a narrow channel with applied dimples is described by a system of Reynolds equations closed using Menter’s [
19] shear stress transfer (MSST) model equations and an energy equation. The original system of equations is solved using multiblock computing technologies implemented in the VP2/3 [
20] package. The finite volume method [
21] is the basis for solving the linearized equations of fluid dynamics, the transfer of turbulent characteristics, and energy. The SIMPLEC method was applied to solve the problem [
22]. The Rhee–Chow monotonizer [
23], modified in [
24], is used in the pressure correction block for centered structured grids with dependent variables located at the centers of computational cells. The incremental discretization scheme is applied, and in the explicit part of the transport equations, the convective terms of the equations of fluid dynamics are approximated by the Leonard [
25] scheme, and in the transport equations of turbulent characteristics, by the Van Leer scheme [
26]. To prevent non-physical oscillations in the reproduction of flows with thin shear layers, the artificial diffusion mechanism is introduced into the implicit part of the equations in combination with the use of one-sided upwind schemes for discretizing the convective terms. The conjugate gradient method is used to solve discretized equations [
27]. To accelerate the convergence of iterations in the pressure correction block, the Demidov algebraic convergence accelerator [
28] is used.
Several tasks are solved. Similarly to [
12], we consider the flow and heat transfer in a narrow channel with 15 single-row spherical and 15 oval dimples arranged in a zigzag pattern with slope angles of
. The diameter of the spherical dimple, equal to the width of the oval dimple, the maximum flow rate, and temperature
K in the inlet section are taken as characteristic parameters. The geometric dimensions of the plane-parallel channel are as follows: length
, height
, and width
. The origin of the Cartesian coordinate system
is located in the middle of the inlet section of the channel on the wall (
Figure 1a).
Furthermore, two reliefs are compared, consisting of spherical and oval dimples (with a cylindrical insert length of ), applied with a step of . Their depth () is and the rounding radius (r) is . The density of the first ensemble of spherical dimples is , and that of the second is . The integral and local characteristics of channels with dimples are normalized according to similar parameters for a channel with plane-parallel walls.
In the inlet section, following [
12], the flow parameters in the near-wall layer are determined corresponding to the chosen thickness of the boundary layer (
). Turbulence intensity in the core of the flow is set low (
). In the outlet section, the flow and heat transfer characteristics are calculated using the solution continuation conditions or soft boundary conditions. No-slip conditions are imposed on the walls. The bottom wall, heated to 373 K, and the top wall, maintained at “room” temperature, are assumed to be isothermal, while the side walls are thermally insulated. Air is considered as a coolant (
), and the Reynolds number is assumed to be
.
The calculations of flows of different scales are carried out on multi-block multilevel grids with their mutual intersection. A novel procedure for interpolating parameters in the area of joining nodes with a different grid structure is formulated, which ensures proper conservatism in solving problems. The developed factorized algorithm is generalized to the case of multiblock computational grids within the framework of the concept of decomposition of the computational domain and the generation of oblique grids with overlapping in the selected significantly different-scale subdomains. The transfer of values between intersecting grids within the framework of a multi-block grid strategy is carried out using non-conservative linear interpolation. The cell volume of the selected structured grid is divided into six pyramids with one of the faces as the base and the cell center as the vertex. Each of the pyramids, in turn, is divided into eight tetrahedra. In the process of intergrid interpolation, it is determined that the selected point, at which it is necessary to determine the set of parameters, belongs to one of the tetrahedra that form the cell, and a linear interpolation is built for it using known values at the vertices. The equivalence of the proposed method and the known method of conservative interpolation [
20] is shown. Approved computational algorithms of algebraic and elliptic type are used to construct curvilinear grids consistent with the boundaries.
To solve the problem, a computational grid is used, which consists of two blocks and contains approximately 2.7 million cells. Dimple boundaries are not distinguished, and they are considered as part of the relief.
The problem of heat transfer in a periodic channel section with two spherical and oval dimples (
Figure 1b–d) is stated as in [
12]. The asymptotic characteristics of heat transfer and hydraulic losses of dimpled channels are calculated on a stabilized section of the flow and heat transfer. We consider spherical dimples located in a corridor stack and a stack with eccentricity relative to the longitudinal middle section of the channel, as well as several ordered structures of oval dimples: single-row inclined dimples oriented at an angle of
, oval dimples arranged in a zigzag with slope angles of
, and combinations of longitudinal-transverse pairs of oval dimples (
Figure 1b,c). The parameters in the inlet and outlet sections are determined from the periodicity condition. On the walls, the characteristics are determined in the same way as in the first problem. When solving, the procedures for correcting pressure and mass-average temperature are used, following [
12,
20]. It should be noted the thematic proximity of the problems being solved with the problem of heat transfer intensification on periodic protrusions [
29]. The good agreement between the calculated and measured characteristics presented in it confirm the acceptability of the chosen semi-empirical turbulence model.
In [
6,
10,
16,
30,
31,
32], the results of a comparison of numerical predictions obtained using SST models with experimental data on separated flow and heat transfer in channels and on plates with solitary and stacked spherical and inclined oval dimples are presented. They also confirm the acceptability of the SST model [
19] for calculating convective heat transfer in intense separated and vortex flows.
2.2. Verification of the Numerical Method and Turbulence Model
The flow around a solitary oval-trench dimple on the wall of a plane-parallel channel is considered. On the experimental stand of the Institute of Mechanics of Moscow State University [
33], with a width of
m, a height of
m and a length of
m, measurements of the static pressure distributions on the surface of a dimple
m wide and a relative length of 5, a depth of
(in ratio of the width) were measured at various angles of inclination in the range of change from
to
. Dimples with a sharp edge are located in the middle of the channel at a distance of
m from the inlet section. The Reynolds number determined from the flow velocity and channel height is
. The thickness of the boundary layer at the entrance to the working section of the channel is
in ratio of the channel height, and the turbulence of the oncoming flow in the channel core is
.
The numerical analogue of the experimental channel has dimensions of . At the channel inlet, a uniform flow is set with a boundary layer thickness equal to the experimental value of . The numerical method and the selected turbulence model were tested on an oval trench hole with a width of , an elongation of , a depth of , and sharp edges rounded along a radius of . The angle of inclination of the hole is . The center of the hole is located at a distance of from the inlet section.
A block computational grid is considered, consisting of two block grids (
Figure 2a). The first Cartesian grid is built in a plane-parallel channel. It contains approximately
million cells, which thicken as they approach the walls. In the central part of the channel, the grid steps are uniform in the longitudinal and transverse directions and equal to
. Inside the channel grid, there is a curvilinear grid, coordinated with the surface of the dimple, on the heated lower wall of the channel, a block grid covering the area of the dimple. The specified area has dimensions of
, in the center of which is the center of the dimple. In the longitudinal and transverse directions, the grid is uniform with a step of
. The number of cells in the curvilinear grid adjacent to the well is approximately
million. The near-wall steps of the considered grids near the walls are
.
Figure 2b compares the numerical predictions and experimental data obtained in [
33] on the distribution of the static pressure coefficient in the section of the joint of the hemispherical segment of the dimple and its trench part, in which an extraordinary pressure drop is observed. Good agreement between the results indicates the acceptable accuracy of the calculation method, as well as the adequacy of the RANS approach and the selected modified SST turbulence model.