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.

Anisotropy Properties of Turbulence in Flow Over Seepage Bed

The present study analyses the Reynolds stress anisotropy in the non-uniform sediment beds under the condition of no seepage and downward seepage flow. The results show the estimation of the deviation measure from the isotropic turbulence in view of Reynolds stress tensor for turbulent flow in the presence of seepage through the channel bed. The investigation presents the Lumley triangle for flow turbulence, Eigen values, and the invariant functions for the whole flow depth subjected to no seepage and seepage beds. The longitudinal profile of anisotropy tensor within the near-bed zone for seepage flow provides the higher anisotropic stream than those of no seepage flow, while the remaining (transverse and vertical) profiles of anisotropy tensor in the vicinity of bed for seepage flows provides lower anisotropic stream. The anisotropic invariant maps show the near bed anisotropy inclining to be a two-component isotropy subjected to no seepage and seepage flow. With the increase in vertical distance from bed surface that is close to the water surface, the data sets of anisotropic invariant maps for no seepage and seepage flows show a trend of one-component isotropy, while it has an affinity to develop a three-component isotropy in the vicinity of mid zone of the flow depth. Invariant function data sets present a well two-component isotropy in the near bed region of flow and a quasi-three component isotropy in the outer region of flow for seepage flows as compared to no seepage flow.

Turbulence Anisotropy Investigations in an Internal Combustion Engine

The assumption of isotropic turbulence is commonly incorporated into models of internal combustion engine (ICE) in-cylinder flows. While preliminary analysis with two-dimensional velocity data indicates that the turbulence may tend to isotropy as the piston approaches TDC, the validity of this assumption has not been fully investigated, partially due to lack of three-component velocity data in ICEs. In this work, the velocity was measured using two-dimensional, three-component (2D-3C) particle image velocimetry in a single-cylinder, motored, research engine to investigate the evolution of turbulence anisotropy throughout the compression stroke. Invariants of the Reynolds stress anisotropy tensor were calculated and visualized, through the Lumley triangle, to investigate turbulence states. Results showed the turbulence to be mostly anisotropic, with preferential tendency toward 2D axisymmetry at the beginning of the compression stroke and approaching isotropy near top-dead-center. Findings provide new insights into turbulence in dynamic, bounded flows to assist with the development of physics-based, quantitative models.

Effects of Axial Casing Grooves on the Structure of Turbulence in the Tip Region of an Axial Turbomachine Rotor

Challenges in turbulence modeling in the tip region of turbomachines include anisotropy, inhomogeneity, and non-equilibrium conditions, resulting in poor correlations between Reynolds stresses and the corresponding mean strain rate components. The geometric complexity introduced by casing grooves exacerbates this problem. Taking advantage of a large database collected in the refractive index-matched liquid facility at JHU, this paper examines the effect of axial casing grooves on the distributions of turbulent kinetic energy (TKE), Reynolds stresses, anisotropy tensor, and TKE production rate in the tip region of an axial turbomachine. Comparisons are performed at flow rates corresponding to prestall and best efficiency points of the untreated machine. Common features include high TKE near the tip leakage vortex center, and in shear layer connecting it to the blade suction side tip corner. The turbulence is highly anisotropic and inhomogeneous, with the anisotropy tensor shifting from one dimensional (1D) to 2D and to 3D structures over small distances. With the grooves, the flow structure, hence the distribution of Reynolds stresses, becomes more complex. Additional sites with elevated turbulence include the corner vortex that develops at the entrance to the grooves, and in the flow jetting out of the grooves into the passage. Consistent with trends of the production rates of normal Reynolds stress components, the grooves increase the axial but reduce the radial velocity fluctuations as the inflow and outflow from the groove interacts with the passage flow. These findings might assist the development of Reynolds stress models suitable for tip flows.

Anisotropy of Reynolds stress tensor in combined wave-current flow

This study examines the Reynolds stress anisotropy tensor based on the turbulence triangle technique and Eigen values in wave-current combined flows. The invariant functions are also presented at four different levels from the bed to comprehend the level of anisotropy in combined flow. The spectral variation of the ratio of principal Reynolds shear stress to the turbulent kinetic energy is examine and discuss in comparison to the canonical value. The combined wave-current data exhibit ratios much smaller than the canonical value of 0.3 across all frequencies. To characterize the behaviours of eddies in wave-current turbulent flow the Taylor and Kolmogorov length and time scales were analysed and discussed. Further, to enumerate degree of organization of complex eddy motions in combined flow the normalized Shannon entropy is also evaluated using discrete probability distribution.

Turbulence Anisotropy Investigations in an Internal Combustion Engine

The assumption of isotropic turbulence is commonly incorporated into models of internal combustion engine (ICE) in-cylinder flows. While preliminary analysis with two-dimensional velocity data indicates that the turbulence may tend to isotropy as the piston approaches TDC, the validity of this assumption has not been fully investigated, partially due to lack of three-component velocity data in ICEs. In this work, the velocity was measured using two-dimensional, three-component (2D-3C) particle image velocimetry in a single-cylinder, motored, research engine to investigate the evolution of turbulence anisotropy throughout the compression stroke. Invariants of the Reynolds stress anisotropy tensor were calculated and visualized, through the Lumley triangle, to investigate turbulence states. Results showed the turbulence to be mostly anisotropic, with preferential tendency toward 2D axisymmetry at the beginning of the compression stroke and approaching isotropy near top-dead-center. Findings provide new insights into turbulence in dynamic, bounded flows to assist with the development of physics-based, quantitative models.

Effects of Axial Casing Grooves on the Structure of Turbulence in the Tip Region of an Axial Turbomachine Rotor

Challenges in predicting the turbulence in the tip region of turbomachines include anisotropy, inhomogeneity, and non-equilibrium conditions, resulting in poor correlations between the Reynold stresses and the corresponding mean strain rate components. The geometric complexity introduced by casing grooves exacerbates this problem. Taking advantage of a large database collected in the refractive index-matched liquid facility at JHU, this paper examines the evolution of turbulence in the tip region of an axial turbomachine with and without axial casing grooves, and for two flow rates. The semi-circular axial grooves are skewed by 45° in the positive circumferential direction, similar to that described in Müller et al. [1]. Comparison to results obtained for an untreated endwall includes differences in the distributions of turbulent kinetic energy (TKE), Reynolds stresses, anisotropy tensor, and dominant terms in the TKE production rate. The evolution of TKE at high flow rates for blade sections located downstream of the grooves is also investigated. Common features include: with or without casing grooves, the TKE is high near the tip leakage vortex (TLV) center, and in the shear layer connecting it to the blade suction side tip corner. The turbulence is highly anisotropic and inhomogeneous, with the anisotropy tensor demonstrating shifts from one dimensional (1D) to 2D and to 3D structures over small distances. Furthermore, the correlation between the mean strain rate and Reynolds stress tensor components is poor. With the grooves, the flow structure, hence the distribution of Reynolds stresses, becomes much more complex. Turbulence is also high in the corner vortex that develops at the entrance to the grooves and in the flow jetting out of the grooves into the passage. Consistent with trends of production rates of normal Reynolds stress components, the grooves increase the axial and reduce the radial velocity fluctuations compared to the untreated endwall. These findings introduce new insight that might assist the future development of Reynolds stress models suitable for tip flows.

Beyond the second-order magnetic anisotropy tensor: higher-order components due to oriented magnetite exsolutions in pyroxenes, and implications for palaeomagnetic and structural interpretations

SUMMARY Exsolved iron oxides in silicate minerals can be nearly ideal palaeomagnetic recorders, due to their single-domain-like behaviour and the protection from chemical alteration by their surrounding silicate host. Because their geometry is crystallographically controlled by the host silicate, these exsolutions possess a shape preferred orientation that is ultimately controlled by the mineral fabric of the silicates. This leads to potentially significant anisotropic acquisition of remanence, which necessitates correction to make accurate interpretations in palaeodirectional and palaeointensity studies. Here, we investigate the magnetic shape anisotropy carried by magnetite exsolutions in pyroxene single crystals, and in pyroxene-bearing rocks based on torque measurements and rotational hysteresis data. Image analysis is used to characterize the orientation distribution of oxides, from which the observed anisotropy can be modelled. Both the high-field torque signal and corresponding models contain components of higher order, which cannot be accurately described by second-order tensors usually used to describe magnetic fabrics. Conversely, low-field anisotropy data do not show this complexity and can be adequately described with second-order tensors. Hence, magnetic anisotropy of silicate-hosted exsolutions is field-dependent and this should be taken into account when interpreting isolated ferromagnetic fabrics, and in anisotropy corrections.