This study presents a comparison of the performance of machine learning (ML) techniques, specifically multi-dimensional gene expression programming (MGEP), tensor basis neural network (TBNN), and also proposes a novel universally interpretable machine learning architecture to model the turbulent scalar flux (UIML-s) to enhance turbulence models for fluid flows at different Prandtl numbers in channels with complex shapes of walls in the channel cross section. In particular, peripheral subchannels of rod bundles are of primary interest. However, the accuracy of mean velocity and scalar distributions predicted by commonly used turbulence models still poses a challenge compared to data extracted from high-fidelity eddy-resolving numerical simulations, particularly for engineering applications involving complex geometry flows. In the present study, by utilizing an explicit algebraic expression for the nonlinear Reynolds-stress term obtained through both the evolutionary MGEP optimization and TBNN, the secondary flow structure has been adequately predicted in the cross-wise mean velocity distributions in the square duct and the rectangular channel with three longitudinal rods. This structure is also observed in the data from the concurrent runs performed by direct numerical simulation (DNS) but is completely absent in the results produced by a baseline Reynolds-averaged Navier–Stokes (RANS) closure, which employs the linear eddy viscosity model for the Reynolds stress tensor. Comparison of MGEP and TBNN has shown their nearly equal performance in a square duct flow; however, MGEP works better for the more complex geometry channel with three rods. Furthermore, based on the velocity field produced by the RANS-MGEP model, the ML modification of the gradient diffusion hypothesis, integrated into the aforementioned novel RANS-ML model called as UIML-s, significantly improves the mean scalar distributions in a flow with three bumps serving as a prototype for the peripheral subchannel of rod bundle. The normalized root mean squared error decreases from 13.5% to 7.6%, bringing the predicted distributions closer to the DNS data, particularly in the near-wall region. Another approach, MGEP-s, also yields the acceptable results, which are nearly identical to those from UIML-s. These findings highlight the potential of using data-driven calibration of turbulence models with nonlinear closures to enhance the predictability for RANS simulations of fluid flows, heat, and mass transfer in channels with complex geometry.
The paper presents the results of numerical simulation of a submerged jet flowing from a circular hole in a wall with introduction of various types of perturbations. Active methods of flow control, including imposition of axial and helical harmonic oscillations on the inlet profile of the jet velocity, vibration of the jet nozzle, as well as their combinations, are considered. It is found that the external forcing leads to the effects of jet splitting at Reynolds number Re ≥ 1000 in wide ranges of perturbation frequencies and amplitudes, as it is shown in the studies of other authors. The mechanisms of interaction of vortex structures during jet bifurcation are investigated, and the angle of flow expansion in the bifurcation plane, which demonstrates an increase with a growing Re, is estimated. Calculations at 500 ≤ Re <1000 show that in order to obtain and enhance the above effects, it is necessary to optimize the forcing parameters, in particular, the type, frequency, and amplitude of perturbation. It is concluded that mechanical vibrations of the nozzle appear to be a more efficient way to control the flow than helical excitation of the inlet velocity profile.
To describe the evolution of an interface between two immiscible media, an equation for a volume fraction function is derived, with the interface curvature effect being described by a “continuum model” of a surface tension force. A numerical study of the Rayleigh-Taylor instability problem is performed for different density ratios ρ1/ρ2 on the interface, including the real cases corresponding to available experimental data. At the initial stage, the instability development is independent of ρ1/ρ2 and consistent with the Taylor linear theory, then (for ρ1/ρ2 < 5) a spiral-like Kelvin-Helmholtz instability structure is observed. For ρ1/ρ2 < 2, the instability development pattern remains symmetric until large times when (same as for large ρ1/ρ2) an asymmetry appears. The surface tension and the viscosity result in the suppression of the Rayleigh-Taylor instability disturbances and secondary small-scale irregularities of the interface.
1 June 2023 in Journal of Applied Mechanics and Technical Physics
DNS and RANS computation results for flows in two-dimensional channels with bumps are processed to generate input and output data for a machine learning method aimed to enhance the Reynolds stress anisotropy model and, thus, improve the RANS approach accuracy. The tensor basis random forest method is chosen as a machine learning tool. The prediction of the new model for the Reynolds stress anisotropy tensor is in better agreement with DNS data for two channel flow geometries than those obtained by the conventional linear eddy viscosity model.
The surface tension effect is introduced according to the continuum model for the surface tension force. A smooth variation of the mollified function of the volume fraction (the color function) across the interface between immiscible fluids occurs due to the convolution of the original color function with the smooth kernel function. The polynomial eighth-degree kernel formulated for plane two-dimensional flows bounded by solid walls or symmetry planes is tested for the Rayleigh — Taylor instability problem.