scholarly journals A Novel Algebraic Stress Model with Machine-Learning-Assisted Parameterization

Energies ◽  
2020 ◽  
Vol 13 (1) ◽  
pp. 258
Author(s):  
Chao Jiang ◽  
Junyi Mi ◽  
Shujin Laima ◽  
Hui Li
Keyword(s):  
Machine Learning ◽  
Eddy Viscosity ◽  
Turbulence Modeling ◽  
Data Driven ◽  
High Fidelity ◽  
Stress Model ◽  
Reynolds Stresses ◽  
Mean Flow ◽  
Nonlocal Effects ◽  
Algebraic Stress Model

Reynolds-stress closure modeling is critical to Reynolds-averaged Navier-Stokes (RANS) analysis, and it remains a challenging issue in reducing both structural and parametric inaccuracies. This study first proposes a novel algebraic stress model named as tensorial quadratic eddy-viscosity model (TQEVM), in which nonlinear terms improve previous model-form failure due to neglection of nonlocal effects. Then a data-driven regression model based on a fully-connected deep neural network is designed to determine the TQEVM coefficients. The well-trained data-driven model using high-fidelity direct numerical simulation (DNS) data successfully learned the underlying input-output relationships, further obtaining spatial-dependent optimal values of these coefficients. Finally, detailed validations are made in wall-bounded flows where nonlocal effects are expected to be significant. Comparative results indicate that TQEVM provides improvements both for the stress-strain misalignment and stress anisotropy, which are clear advantages over linear and quadratic eddy-viscosity models. TQEVM extends to the scope of resolution to the wall distance y + ≈ 9 as well as provides a realizable solution. RANS simulations with TQEVM are also carried out and the obtained mean-flow quantities of interest agree well with DNS. This work, therefore, results in a high-fidelity representation of Reynolds stresses and contributes to further understanding of machine-learning-assisted turbulence modeling and regression analysis.

Improved Prediction of Turbomachinery Flows Using Near-Wall Reynolds-Stress Model

2001 ◽  
Vol 124 (1) ◽  
pp. 86-99 ◽  
Author(s):  
G. A. Gerolymos ◽  
J. Neubauer ◽  
V. C. Sharma ◽  
I. Vallet
Keyword(s):  
Experimental Data ◽  
Turbulent Flows ◽  
Reynolds Stress ◽  
Turbulence Models ◽  
Reynolds Stress Model ◽  
Stress Model ◽  
Reynolds Stresses ◽  
Mean Flow ◽  
Flow Equations ◽  
Near Wall

In this paper an assessment of the improvement in the prediction of complex turbomachinery flows using a new near-wall Reynolds-stress model is attempted. The turbulence closure used is a near-wall low-turbulence-Reynolds-number Reynolds-stress model, that is independent of the distance-from-the-wall and of the normal-to-the-wall direction. The model takes into account the Coriolis redistribution effect on the Reynolds-stresses. The five mean flow equations and the seven turbulence model equations are solved using an implicit coupled OΔx3 upwind-biased solver. Results are compared with experimental data for three turbomachinery configurations: the NTUA high subsonic annular cascade, the NASA_37 rotor, and the RWTH 1 1/2 stage turbine. A detailed analysis of the flowfield is given. It is seen that the new model that takes into account the Reynolds-stress anisotropy substantially improves the agreement with experimental data, particularily for flows with large separation, while being only 30 percent more expensive than the k−ε model (thanks to an efficient implicit implementation). It is believed that further work on advanced turbulence models will substantially enhance the predictive capability of complex turbulent flows in turbomachinery.

scholarly journals Reaction Wheels Fault Isolation Onboard 3-Axis Controlled Satel-lite using Enhanced Random Forest with Multidomain Features

2021 ◽  
Vol 12 (2) ◽  
Author(s):  
Afshin Rahimi ◽  
Mofiyinoluwa O. Folami
Keyword(s):  
Machine Learning ◽  
Fault Isolation ◽  
Data Driven ◽  
Machine Learning Techniques ◽  
Superior Performance ◽  
Alternative Methods ◽  
Model Parameters ◽  
High Fidelity ◽  
Novel Approach ◽  
Learning Techniques

As the number of satellite launches increases each year, it is only natural that an interest in the safety and monitoring of these systems would increase as well. However, as a system becomes more complex, generating a high-fidelity model that accurately describes the system becomes complicated. Therefore, imploring a data-driven method can provide to be more beneficial for such applications. This research proposes a novel approach for data-driven machine learning techniques on the detection and isolation of nonlinear systems, with a case-study for an in-orbit closed loop-controlled satellite with reaction wheels as actuators. High-fidelity models of the 3-axis controlled satellite are employed to generate data for both nominal and faulty conditions of the reaction wheels. The generated simulation data is used as input for the isolation method, after which the data is pre-processed through feature extraction from a temporal, statistical, and spectral domain. The pre-processed features are then fed into various machine learning classifiers. Isolation results are validated with cross-validation, and model parameters are tuned using hyperparameter optimization. To validate the robustness of the proposed method, it is tested on three characterized datasets and three reaction wheel configurations, including standard four-wheel, three-orthogonal, and pyramid. The results prove superior performance isolation accuracy for the system under study compared to previous studies using alternative methods (Rahimi & Saadat, 2019, 2020).

3D Reynolds Stress Modelling of Turbulent Flow in Linear Turbine Cascade

1997 ◽  
Author(s):  
M. Kanniche ◽  
R. Boudjemadi ◽  
F. Déjean ◽  
F. Archambeau
Keyword(s):  
Reynolds Stress ◽  
Eddy Viscosity ◽  
Strain Tensor ◽  
Secondary Flows ◽  
Turbulence Closure ◽  
Turbine Cascade ◽  
Reynolds Stresses ◽  
Mean Flow ◽  
Principal Axes ◽  
The Mean

The flow in a linear turbine cascade (Gregory-Smith et al. (1990)) is numerically investigated using a Reynolds Stress Turbulence closure. A particular attention is given to secondary flows where the normal Reynolds stresses are expected to play an important role. The most classical turbulence closure, the k-epsilon model uses the Boussinesq Eddy Viscosity concept which assumes an isotropic turbulent viscosity. The Reynolds stresses are then related to local velocity gradients by this isotropic eddy viscosity. Corollary, the principal axes of the Reynolds stress tensor are colinear with those of the mean strain tensor. The advantage of Reynolds Stress Turbulence closure is the calculation of Reynolds stresses by their own individual transport equations. This leads to a more realistic description of the turbulence and of its dependance on the mean flow. The most classical Second Order turbulence model (Launder et al. (1975)) is applied to a linear turbine cascade, and the results are compared to secondary velocity and turbulence measurements at cross-passage planes.

Numerical Simulation of a Three-Dimensional Axisymmetric Hill: Performance Evaluation of Hybrid RANS-LES Models

2021 ◽  
Author(s):  
Tausif Jamal ◽  
Varun Chitta ◽  
Dibbon K. Walters
Keyword(s):  
Eddy Viscosity ◽  
Turbulence Modeling ◽  
Three Dimensional ◽  
Boundary Layer Separation ◽  
Wake Flow ◽  
Dynamics Simulation ◽  
Recirculation Region ◽  
Mean Flow ◽  
Large Eddy ◽  
Rans Models

Abstract Computational fluid dynamics simulation of flow over a three-dimensional axisymmetric hill presents a unique set of challenges for turbulence modeling. The flow past the crest of the hill is characterized by boundary layer separation, complex vortical structures, and unsteady wake flow. As a result, traditional eddy-viscosity Reynolds-averaged Navier-Stokes (RANS) models have been found to perform poorly for this benchmark test case. Recent studies have focused on the use of large-eddy simulation (LES) and hybrid RANS-LES (HRL) methods to improve accuracy. In this study, several different HRL models are investigated and results from the different models are evaluated relative to each other, to an eddy-viscosity RANS model, and to previously documented high-fidelity large-eddy simulations and experimental data. Results obtained from the simulations in terms of mean flow statistics, surface pressure distribution, and turbulence characteristics are presented and discussed in detail. Results indicate that HRL models can significantly improve predictions over RANS models, but only when the development of turbulent velocity fluctuations in the separated shear layer and recirculation region are well resolved.

The application of machine learning to structural health monitoring

2006 ◽  
Vol 365 (1851) ◽  
pp. 515-537 ◽  
Author(s):  
Keith Worden ◽  
Graeme Manson
Keyword(s):  
Machine Learning ◽  
Normal Condition ◽  
Damage Identification ◽  
Measured Data ◽  
Data Driven ◽  
High Fidelity ◽  
Element Analysis ◽  
Model Driven ◽  
Statistical Representation ◽  
Data Driven Approach

In broad terms, there are two approaches to damage identification. Model-driven methods establish a high-fidelity physical model of the structure, usually by finite element analysis, and then establish a comparison metric between the model and the measured data from the real structure. If the model is for a system or structure in normal (i.e. undamaged) condition, any departures indicate that the structure has deviated from normal condition and damage is inferred. Data-driven approaches also establish a model, but this is usually a statistical representation of the system, e.g. a probability density function of the normal condition. Departures from normality are then signalled by measured data appearing in regions of very low density. The algorithms that have been developed over the years for data-driven approaches are mainly drawn from the discipline of pattern recognition, or more broadly, machine learning. The object of this paper is to illustrate the utility of the data-driven approach to damage identification by means of a number of case studies.

Calculation of turbulence-driven secondary motion in non-circular ducts

1984 ◽  
Vol 140 ◽  
pp. 189-222 ◽  
Author(s):  
A. O. Demuren ◽  
W. Rodi
Keyword(s):  
Eddy Viscosity ◽  
Three Dimensional ◽  
Numerical Procedure ◽  
Momentum Equation ◽  
Shear Stresses ◽  
Reynolds Stresses ◽  
Calculation Methods ◽  
Mean Flow ◽  
Square Duct ◽  
Secondary Motion

Experiments on and calculation methods for flow in straight non-circular ducts involving turbulence-driven secondary motion are reviewed. The origin of the secondary motion and the shortcomings of existing calculation methods are discussed. A more refined model is introduced, in which algebraic expressions are derived for the Reynolds stresses in the momentum equations for the secondary motion by simplifying the modelled Reynolds-stress equations of Launder, Reece & Rodi (1975), while a simple eddy-viscosity model is used for the shear stresses in the axial momentum equation. The kinetic energy k and the dissipation rate ε of the turbulent motion which appear in the algebraic and the eddy-viscosity expressions are determined from transport equations. The resulting set of equations is solved with a forward-marching numerical procedure for three-dimensional shear layers. The model, as well as a version proposed by Naot & Rodi (1982), is tested by application to developing flow in a square duct and to developed flow in a partially roughened rectangular duct investigated experimentally by Hinze (1973). In both cases, the main features of the mean-flow and the turbulence quantities are simulated realistically by both models, but the present model underpredicts the secondary velocity while the Naot-Rodi model tends to overpredict it.

Analysis of Turbulent Flow Around Horizontal Axis Wind Turbines Using Algebraic Stress Model

2015 ◽  
Author(s):  
Saman Beyhaghi ◽  
Ryoichi S. Amano
Keyword(s):  
Turbulent Flow ◽  
Wind Turbine ◽  
Reynolds Stress ◽  
Wind Turbines ◽  
Cost Effective ◽  
Horizontal Axis ◽  
Stress Model ◽  
Reynolds Stresses ◽  
Algebraic Stress Model ◽  
Near Wall

Due to the problems associated with increase of greenhouse gases, and the limited supply of fossil fuels, switching to clean and renewable sources of energy seems necessary. Wind energy is a very suitable form of renewable energy which can be a good choice for those areas around the world with sufficient amount of wind annually. However, in order for the commercial wind turbines to be cost-effective, they need to operate at very high elevations, and therefore, blades with the length as high as 60–70 m are common. Because of the high manufacturing and transportation costs of the wind turbine components, it is necessary to evaluate and predict the performance of the turbine prior to shipping it to the installation site. Computational Fluid Dynamics (CFD) has proven to be a simple, cheap and yet relatively accurate tool for prediction of wind turbine performance, where suitability of different designs can be evaluated at a low cost. Total lift and drag forces can be calculated, from which one can estimate the torque, and ultimately the output power. Reynolds Stress Model (RSM) is a well-known Reynolds Averaged Navier-Stokes (RANS) turbulence model, which is typically more accurate than eddy viscosity models, but it comes with higher computational cost. In the present work, turbulent flow of air around a horizontal axis wind turbine blade is modeled computationally by using a modified version of RSM, known as Algebraic Stress Model (ASM) for the near-blade region. Because of the periodicity nature of the flow domain, only one of three blades is modeled by applying the periodic conditions on the sides of a 120 degree sector of the domain. While the flow is solved in the bulk fluid using the k-epsilon model, in order to better capture the near-wall effects and to make the computations cost effective, it is proposed to apply ASM only in the locations very close to the blade surface. A number of reasonable assumptions are made in ASM in order to convert the transport differential equations of the Reynolds stresses into an algebraic form. The highly coupled system of non-linear equations is then solved concurrently for six Reynolds stress components. Turbulent kinetic energy, turbulent dissipation rate, and mean velocity gradients are calculated from the k-epsilon model and used as initial values and iterated through the ASM computations. To the best of our knowledge, this is the first time that ASM is used for analysis of Reynolds stress for flow around rotating wind turbines blades. Reynolds stresses are obtained at several locations (heights) along the blade, and at different radial distances from the blade. Different variations of implicit and explicit ASM are examined and compared in terms of accuracy. Results indicate that the implicit ASM method using the full form of pressure-strain term tends to show predictions that are closer to the predictions of the fully-resolved RSM simulation, as compared to the other ASM models examined. Therefore, there seems to be a good potential for reducing computational costs for determination of near wall Reynolds stresses and ultimately calculating torque and power generated from wind turbines without sacrificing the accuracy.

scholarly journals Hybrid Modelling by Machine Learning Corrections of Analytical Model Predictions towards High-Fidelity Simulation Solutions

Materials ◽  
2021 ◽  
Vol 14 (8) ◽  
pp. 1883
Author(s):  
Frederic E. Bock ◽  
Sören Keller ◽  
Norbert Huber ◽  
Benjamin Klusemann
Keyword(s):  
Neural Network ◽  
Machine Learning ◽  
Analytical Model ◽  
Data Driven ◽  
High Fidelity ◽  
Prediction Errors ◽  
Hybrid Modelling ◽  
Model Predictions ◽  
Physical Laws ◽  
Modelling Approach

Within the fields of materials mechanics, the consideration of physical laws in machine learning predictions besides the use of data can enable low prediction errors and robustness as opposed to predictions only based on data. On the one hand, exclusive utilization of fundamental physical relationships might show significant deviations in their predictions compared to reality, due to simplifications and assumptions. On the other hand, using only data and neglecting well-established physical laws can create the need for unreasonably large data sets that are required to exhibit low bias and are usually expensive to collect. However, fundamental but simplified physics in combination with a corrective model that compensates for possible deviations, e.g., to experimental data, can lead to physics-based predictions with low prediction errors, also despite scarce data. In this article, it is demonstrated that a hybrid model approach consisting of a physics-based model that is corrected via an artificial neural network represents an efficient prediction tool as opposed to a purely data-driven model. In particular, a semi-analytical model serves as an efficient low-fidelity model with noticeable prediction errors outside its calibration domain. An artificial neural network is used to correct the semi-analytical solution towards a desired reference solution provided by high-fidelity finite element simulations, while the efficiency of the semi-analytical model is maintained and the applicability range enhanced. We utilize residual stresses that are induced by laser shock peening as a use-case example. In addition, it is shown that non-unique relationships between model inputs and outputs lead to high prediction errors and the identification of salient input features via dimensionality analysis is highly beneficial to achieve low prediction errors. In a generalization task, predictions are also outside the process parameter space of the training region while remaining in the trained range of corrections. The corrective model predictions show substantially smaller errors than purely data-driven model predictions, which illustrates one of the benefits of the hybrid modelling approach. Ultimately, when the amount of samples in the data set is reduced, the generalization of the physics-related corrective model outperforms the purely data-driven model, which also demonstrates efficient applicability of the proposed hybrid modelling approach to problems where data is scarce.

Sign in / Sign up

Export Citation Format

Share Document