Ridge-type roughness: from turbulent channel flow to internal combustion engine

While existing engineering tools enable us to predict how homogeneous surface roughness alters drag and heat transfer of near-wall turbulent flows to a certain extent, these tools cannot be reliably applied for heterogeneous rough surfaces. Nevertheless, heterogeneous roughness is a key feature of many applications. In the present work we focus on spanwise heterogeneous roughness, which is known to introduce large-scale secondary motions that can strongly alter the near-wall turbulent flow. While these secondary motions are mostly investigated in canonical turbulent shear flows, we show that ridge-type roughness—one of the two widely investigated types of spanwise heterogeneous roughness—also induces secondary motions in the turbulent flow inside a combustion engine. This indicates that large scale secondary motions can also be found in technical flows, which neither represent classical turbulent equilibrium boundary layers nor are in a statistically steady state. In addition, as the first step towards improved drag predictions for heterogeneous rough surfaces, the Reynolds number dependency of the friction factor for ridge-type roughness is presented. Graphic abstract

Modeling the pressure strain correlation in turbulent flows using deep neural networks

The pressure strain correlation plays a critical role in the Reynolds stress transport modeling. Accurate modeling of the pressure strain correlation leads to the proper prediction of turbulence stresses and subsequently the other terms of engineering interest. However, classical pressure strain correlation models are often unreliable for complex turbulent flows. Machine learning–based models have shown promise in turbulence modeling, but their application has been largely restricted to eddy viscosity–based models. In this article, we outline a rationale for the preferential application of machine learning and turbulence data to develop models at the level of Reynolds stress modeling. As an illustration, we develop data-driven models for the pressure strain correlation for turbulent channel flow using neural networks. The input features of the neural networks are chosen using physics-based rationale. The networks are trained with the high-resolution DNS data of turbulent channel flow at different friction Reynolds numbers (Reλ). The testing of the models is performed for unknown flow statistics at other Reλ and also for turbulent plane Couette flows. Based on the results presented in this article, the proposed machine learning framework exhibits considerable promise and may be utilized for the development of accurate Reynolds stress models for flow prediction.

Asymptotic behaviour at the wall in compressible turbulent channels

The asymptotic behaviour of Reynolds stresses close to walls is well established in incompressible flows owing to the constraint imposed by the solenoidal nature of the velocity field. For compressible flows, thus, one may expect a different asymptotic behaviour, which has indeed been noted in the literature. However, the transition from incompressible to compressible scaling, as well as the limiting behaviour for the latter, is largely unknown. Thus, we investigate the effects of compressibility on the near-wall, asymptotic behaviour of turbulent fluxes using a large direct numerical simulation (DNS) database of turbulent channel flow at higher than usual wall-normal resolutions. We vary the Mach number at a constant friction Reynolds number to directly assess compressibility effects. We observe that the near-wall asymptotic behaviour for compressible turbulent flow is different from the corresponding incompressible flow even if the mean density variations are taken into account and semi-local scalings are used. For Mach numbers near the incompressible regimes, the near-wall asymptotic behaviour follows the well-known theoretical behaviour. When the Mach number is increased, turbulent fluxes containing wall-normal components show a decrease in the slope owing to increased dilatation effects. We observe that $R_{vv}$ approaches its high-Mach-number asymptote at a lower Mach number than that required for the other fluxes. We also introduce a transition distance from the wall at which turbulent fluxes exhibit a change in scaling exponents. Implications for wall models are briefly presented.

Flow-Particle Coupling in a Channel Flow Laden with Elongated Particles: The Role of Aspect Ratio

A turbulent channel flow laden with elongated, fiber-like particles is investigated experimentally by optical techniques. The flow-particle inter-coupling is analyzed in the case of particles with an aspect ratio of 40 and 80, at two volume fractions, 10−5 and 10−4. An image processing technique is presented, which is employed to simultaneously obtain carrier flow velocimetry data and distribution and orientation data of dispersed particles. Turbulence enhancement is reported in the near-wall region, with a higher level of increase associated with higher aspect ratio particles. Comparison to fiber data suggests that this mechanism of turbulence modulation stems from a particles orientational behavior. The preferential particle distribution is reported to be dependent on the aspect ratio in the region close to the wall. The probability density function of the fibers’ orientation angle appears to be independent of the particle aspect ratio once it is conditioned to the fibers’ characteristic size.

Data-driven turbulence modelling using symbolic regression

The study is focused on the performance of machine-learning methods applied to improve the velocity field predictions in canonical turbulent flows by the Reynolds-averaged Navier–Stokes (RANS) equation models. A key issue here is to approximate the unknown term of the Reynolds stress (RS) tensor needed to close the RANS equations. A turbulent channel flow with the curved backward-facing step on the bottom has the high-fidelity LES data set. It is chosen as the test case to examine possibilities of GEP (gene expression programming) of formulating the enhanced RANS approximations. Such a symbolic regression technique allows us to get the new explicit expressions for the RS anisotropy tensor. Results obtained by the new model produced using GEP are compared with those from the LES data (serving as the target benchmark solution during the machine-learning algorithm training) and from the conventional RANS model with the linear gradient Boussinesq hypothesis for the Reynolds stress tensor.