Prediction Of Flow Field Research Articles

Numerical Modeling of Turbulent Biogas Combustion

The primary purpose of this research is to investigate the combustion properties of three different types of biogas using a 300 KW BERL combustor. Biogas is a renewable type of this fossil fuel; it is an intelligent fuel that offers an incredibly environmentally beneficial alternative to existing fuels.In order to study the impact of the biogas compositions on the flow field prediction, we perform the calculation using the FLUENT code, which has been used to present the numerical modeling of turbulent diffusion flames by using the realizable k-–ε model of turbulent flow interacting with a two-dimensional PDF combustion scheme. Some comparisons of biogas performance in turbulent diffusion flame mode and methane performance in conventional mode are shown, in addition to the 9-steps, GRI 2.11 mechanisms and experiment data are used to validate the case investigated.

Reconstruction of Nusselt number in RANS computations with wall function approach

Two novel methods for reconstruction of Nusselt number (Nu) are proposed when the wall function approach is used in the RANS framework for defining wall boundary conditions for high Reynolds number flows. The work is motivated by the observation that the velocity and temperature fields are often well predicted when the wall function formulations are used, while at the same time Nu is not. The reason for discrepancy in predictions of flow fields and temperature, and prediction of Nu, is because Nu estimation is solely dependent on the wall function, while the variables obtained from the transport equations are determined also by the convection and pressure force in the momentum transport. The method based on the correction factor Ψ∗ originates from the simplified analytical wall function approach of Popovac and Hanjalić (2007) and modifies the formulation of so-called wall viscosity. The method based on a wall shape function of exponential type approximates temperature variation in the near-wall region. Once the shape function’s coefficients are calculated using values of temperature in three near-wall consecutive cells, the wall values of temperature or temperature gradient, needed for calculation of Nu, can be determined. The methods are tested for four standard benchmark cases, backward-facing step, impinging jet, differentially heated square cavity, and Rayleigh-Bénard convection. The method based on the correction factor significantly improved Nu predictions for the cases where the velocity field is strongly impacted by pressure force. The method based on the wall shape function improves predictions of Nu in all tested cases, with somewhat less accurate predictions of the maximum value of Nu compared to the correction factor method.

Prediction of transonic flow over supercritical airfoils using geometric-encoding and deep-learning strategies

The Reynolds-averaged Navier–Stokes equation for compressible flow over supercritical airfoils under various flow conditions must be rapidly and accurately solved to shorten design cycles for such airfoils. Although deep-learning methods can effectively predict flow fields, the accuracy of these predictions near sensitive regions and their generalizability to large-scale datasets in engineering applications must be enhanced. In this study, a modified vision transformer-based encoder–decoder network is designed for the prediction of transonic flow over supercritical airfoils. In addition, four methods are designed to encode the geometric input with various information points and the performances of these methods are compared. The statistical results show that these methods generate accurate predictions over the complete flow field, with a mean absolute error on the order of 1 × 10−4. To increase accuracy near the shock area, multilevel wavelet transformation and gradient distribution losses are introduced into the loss function. This results in the maximum error that is typically observed near the shock area decreasing by 50%. Furthermore, the models are pretrained through transfer learning on large-scale datasets and fine-tuned on small datasets to improve their generalizability in engineering applications. The results generated by various pretrained models demonstrate that transfer learning yields a comparable accuracy from a reduced training time.

Research on time series prediction of the flow field in supersonic combustor based on deep learning

The detection of supersonic combustion flow field provides valuable reference information for the efficient combustion of hypersonic vehicles. From the perspectives of scramjet operation during flight as well as experiment where the measured information is limited, flow field prediction using deep learning technology is a promising method that can provide future flow field evolution in scramjet combustors. This study proposes a convolutional neural network model of multi-temporal pressure information fusion to predict combustion flow field based on wall pressure. In addition, it analyzes the effect of various wall pressures on the flow field prediction. An experimental dataset of supersonic ground-pulse wind tunnel tests with different injection pressures is built through the combustion chamber wall under self-ignition conditions of a hydrogen-fueled scramjet. The results showed that the predicted flow field of the combustor agrees with the ground test, average correlation coefficient reached more than 90% and when the wall pressure number is 32, the average peak signal to noise ratio reaches about 23. and the main wave structure and turbulent boundary layer region are restored even if there is a shock boundary layer interaction.

A novel model with an improved loss function to predict the velocity field from the pressure on the surface of the hydrofoil

Reconstructing the flow field from limited observations is critical, as it can be used in marine applications which monitor flow changes with limited observations. With the development of deep learning, prediction models based on neural networks have been used for rapid prediction of flow fields. The work proposes models based on U-structure networks (U-net) and Signed Distance Function (SDF) to predict the velocity field from sparse pressure on the surface of the hydrofoil. An improved loss function is employed to improve the prediction accuracy. Through training and testing on a numerical simulation dataset, it is found that the prediction of models has a good agreement with the true values. The improved loss function redirects attention of the model to areas with poor prediction accuracy near the surface of the hydrofoil. For the dataset unknown to models, the prediction ability of the model is limited as input features are not enough to support strong generalization.

Physics-informed data-driven model for fluid flow in porous media

A physics-informed machine learning model is developed that can replace the numerical simulations of porous media. By learning the communications among grid cells in the numerical domain, this model is capable of accurately predicting flow fields for new sets of simulation runs. Because of the many possible random arrangements of particles and their orientation with respect to each other, generalization of permeability with high accuracy is not trivial — nor is it practical using conventional means. Furthermore, building a comprehensive database for different grain/pore arrangements is not possible because of the cost of running numerical simulations to generate the database that represent all possible arrangements. The objective is to predict grid-level flow fields in porous media as a priori to determining permeability of porous media. The rationale is that once the detailed grid-level dynamics can be accurately predicted using data-driven approach, for any configuration/topology of the porous media, the detailed dynamics could be predicted without any need for new expensive numerical simulation runs. In this work, a physics-informed deep learning model is developed by using the results from Lattice-Boltzmann simulations of randomly distributed circular grains to represent the porous media. A variety of porous structures are developed by changing the number, size, and location of circular grains. The deep U-Net and ResNet neural network architectures are combined to train a deep learning model which avoids the vanishing gradient issues. The continuity and momentum conservation equations are embedded into the loss function of deep learning architecture. Robustness of the developed model is then tested for numerous variations of porous media which have not been used for developing the model.

Comparison of Thermal Effectiveness and Crevice Corrosion Risk of Fin Geometry on All-Aluminum Microchannel Heat Exchangers

All-aluminum microchannel heat exchangers have recently gained popularity in the heating, ventilation, and air conditioning industry. Despite their attractive thermal performance design, these heat exchangers make coils used in automotive, commercial, and residential applications prone to crevice corrosion. This study uses high-fidelity conjugate heat transfer simulations to model a micro channel heat exchanger system that includes fins and tubes with crossflow to compare their thermal effectiveness to gain insight into potential crevice corrosion of the MCHE alloy. This study considers three fin geometries (louver, step, and saw) with the same tube and circular shape microchannel and identifies the corrosion hot spot and thermal effectiveness. A predicted flow field also identifies crevices between fins and tube surfaces as critical corrosion hot spots often associated with low-velocity regions. The crevice volumes for the louver, step, and saw fin shapes are calculated as respectively. Results also show that the same circular microchannel louver shape fin has higher effectiveness than the step and saw shape fin. The thermal effectiveness for microchannel tubes with louver, step, and saw shape fins are 0.337, 0.20737, and 0.2895, respectively.

The prediction of external flow field and hydrodynamic force with limited data using deep neural network

The prediction of external flow field and hydrodynamic force with limited data using deep neural network

Improved prediction of coherent structure in an intermediate turbine duct

Modeling the unsteady flow in an intermediate turbine duct in an aero-engine is highly challenging partly owing to the complex physics subjected to unsteady vortical flow upstream. In this paper, the high-fidelity detached eddy simulations of an aggressive intermediate turbine duct are performed and validated against published data. The dominating coherent flow motions are then captured with sparsity-promoting dynamics mode decomposition under different tip gap sizes and rotating speeds of the rotor blade. Furthermore, a novel low-order spectrum reconstruction method based on selected flow modes and Gaussian fitting is proposed for fast prediction of the instantaneous flowfield in the intermediate turbine duct. The sources of error are discussed and a potential approach to improve the prediction accuracy is proposed. It reveals that the proposed method achieves accurate and generalized performance in reconstructing and predicting flow fields with periodic features for a small cost of calculation time. The maximum root mean squared error in the scenario of tip gap size and rotating speed variation is below 5.96 and 8.41, respectively. The error analysis shows that the variation of flow modes under different inflow conditions and nonlinearity of interacting flow structure accounts for the major part of prediction error. By applying a Proper orthogonal decomposition (POD) interpolation on the flow modes, the maximum root mean squared error in the scenario of tip gap size and rotating speed variation can drop to 4.76 and 2.69 without breaking the physical meaning of the representative modes. The novelty of the present work lies at the proposed prediction method combining DMDSP and POD-interpolation, of which the majority of the physical information are maintained and the prediction accuracy is improved. It may provide a promising approach for turbomachinery design, flow prediction, and investigating underlying mechanisms with low data requirements.

Transverse Velocity Field Measurements in High-resolution Solar Images Based on Deep Learning

To address the problem of the low accuracy of transverse velocity field measurements for small targets in high-resolution solar images, we proposed a novel velocity field measurement method for high-resolution solar images based on PWCNet. This method transforms the transverse velocity field measurements into an optical flow field prediction problem. We evaluated the performance of the proposed method using the Hα and TiO data sets obtained from New Vacuum Solar Telescope observations. The experimental results show that our method effectively predicts the optical flow of small targets in images compared with several typical machine- and deep-learning methods. On the Hα data set, the proposed method improves the image structure similarity from 0.9182 to 0.9587 and reduces the mean of residuals from 24.9931 to 15.2818; on the TiO data set, the proposed method improves the image structure similarity from 0.9289 to 0.9628 and reduces the mean of residuals from 25.9908 to 17.0194. The optical flow predicted using the proposed method can provide accurate data for the atmospheric motion information of solar images. The code implementing the proposed method is available on https://github.com/lygmsy123/transverse-velocity-field-measurement.

Fast flow field prediction of hydrofoils based on deep learning

Conventionally, the flow field over the hydrofoil is solved by computational fluid dynamics (CFD), which is a computationally expensive task. As an alternative, deep learning (DL) approximates the flow field without solving the complicated Navier-Stokes (NS) equations. Consequently, this paper proposes a data-driven methodology using Convolutional Neural Networks (CNN) for hydrofoils performance prediction, which automatically detects fundamental features and accurately predicts the flow field of the hydrofoil. The CFD simulation is used to generate the dataset for neural network training and testing. The dataset containing hydrofoil geometry and operating conditions is fed into the CNN to approximate the model for predicting the flow field. Furthermore, we use novel attention gates (AGs) that concentrate on hydrofoils of different shapes and sizes. Models trained with AGs not only learn to restrain unrelated features of the input information, but also emphasize significant features useful for flow field prediction. In addition, the effects of various hyperparameters of models on convergence have been extensively investigated. The experimental results demonstrate that the flow field over a hydrofoil can be obtained within 0.5 s, with a mean relative error of less than 1.5%.

TransCFD: A transformer-based decoder for flow field prediction

The computational fluid dynamics (CFD) method is computationally intensive and costly, and evaluating aerodynamic performance through CFD is time-consuming and labor-intensive. For the design and optimization of aerodynamic shapes, it is essential to obtain aerodynamic performance efficiently and accurately. This paper proposed TransCFD, a Transformer-based decoding architecture for flow field prediction. The aerodynamic shape is parameterized and used as input to the decoder, which learns an end-to-end mapping between the shape and the flow fields. Compared with the CFD method, the TransCFD was evaluated to have a mean absolute error (MAE) of less than 1%, increase the speed by three orders of magnitude, and perform very well in generalization capability. The method simplifies the input requirements compared to most existing methods. Although the object of this work is a two-dimensional airfoil, the setup of this scheme is very general. TransCFD is promising for rapid aerodynamic performance evaluation, with potential applications in accelerating the aerodynamic design.

A hybrid experimental-computational study: Prediction of flow fields and full-field pressure distributions on measured shapes of three-tab asphalt roofing shingles subjected to hurricane velocity winds

A novel hybrid experimental-computational study is performed to predict the flow fields and pressure distributions on the measured three-dimensional shapes of flexible, three-tab asphalt roofing shingles undergoing increasing uplift when exposed to hurricane velocity winds for two hours. To quantify the evolution of shingle shapes, StereoDIC analysis is used to measure the transient, full-field deformed shapes of full-sized, three-tab asphalt shingles that did not show separation or failure when subjected to hurricane velocity winds for two hours. Based on physical observations during wind loading, the authors performed steady state computational fluid dynamics (CFD) simulations to predict the full-field pressure distributions on as-measured, uplifted three-dimensional shingle shapes at selected time instances during wind loading.Simulation predictions clearly show flow recirculation regions on both the front and top of the shingles that remain attached throughout wind loading and control the full-field uplift pressure distribution. For low velocity flow with maximum uplift ≤ 8.4 mm, CFD-predicted pressures are in good agreement with prior measurements. For both low and high-speed flows, the model predictions indicate that high pressures are formed at the leading-edge, upstream of the sealant layer, with maximum pressure occurring near the tab cutouts along the leading-edge of the shingle, providing a physical basis for the observed higher uplift and increased potential for shingle failure in these regions. The combined experimental-computational studies provide a contemporary way to eliminate the difficulties associated with attachment of pressure sensors to flexible materials that can alter shingle response, providing the basis for future design improvements by delineating the physical processes controlling pressure loading and shingle uplift in hurricane velocity winds.

Fast prediction of compressor flow field in nuclear power system based on proper orthogonal decomposition and deep learning

Research and development on digital twins of nuclear power systems has focused on high-precision real-time simulation and the prediction of local complex three-dimensional fluid dynamics. Traditional computational fluid dynamics (CFD) methods cannot take into consideration the efficiency and accuracy of fluid dynamics. In this study, a fast-flow field-prediction framework based on proper orthogonal decomposition (POD) and deep learning is proposed. Compressed data containing the original flow field information are obtained using POD and deep neural network (DNN) is used to construct the POD-DNN flow field reduction model to achieve fast flow field prediction. The calculation accuracy and speed of the reduced-order model are analyzed in detail, considering the flow field of the nuclear compressor and key flow equipment of the nuclear power system as objects. The results show that the average relative deviation of the POD-DNN is <10% and calculation time is <1% when compared to those of CFD. This research shows that the high-fidelity model constructed using model reduction and deep learning is a feasible method for the realization of digital twins of the nuclear power system in engineering.

A Study on Using Location-Information-Based Flow Field Reconstruction to Model the Characteristics of a Discharging Valve in a Hydrodynamic Retarder

In modeling the characteristics of a discharging valve in a hydrodynamic retarder, it is commonly required to determine the value of the flow area to calculate the force on the spool. However, the flow area often relies heavily on empirical or simulation data, which leads to increased uncertainty and computational cost, especially with the variation in the spool displacement. To overcome these shortcomings, Res-SE-U-Nets (networks that combine residual connections, squeeze-and-excitation blocks, and U-Net) are used to reconstruct the velocity field, and they have shown exceptional performance in image-to-image mapping tasks. The dataset of computational fluid dynamics (CFD) results for the velocity field is collected and verified using particle image velocimetry (PIV). The results show that Res-SE-U-Nets can capture the location information of the flow field using a training set of only 120 data points. By utilizing location information in velocity field reconstruction, the flow area can be directly obtained under different spool displacements and pressures to calculate the spool force. The valve characteristics calculated with this method show an error of less than 2% when compared with CFD results, which confirms the validity and effectiveness of this method. The proposed method, which utilizes location information extracted from flow field prediction results, is capable of calculating valve characteristics. This approach also demonstrates the feasibility of using Res-SE-U-Nets for flow field reconstruction.

Fast and reliable prediction of scramjet flowfields via Gaussian process latent variable model and deep learning

Fast and accurate prediction of high-speed flowfields is of particular interest to researchers in fluid science and engineering to enable efficient design exploration and knowledge discovery. The reliability of prediction is another important metric for the performance of prediction models. While predictive modeling approaches with and without reduced-order modeling (ROM) via machine learning techniques have been proposed, they are inherently subject to loss of information for ROM-based approaches and substantial computational costs in modeling for non-ROM-based approaches. This paper proposes an accurate ROM-based predictive framework with minimum information loss enabled by incorporating Gaussian process latent variable modeling (GPLVM) and deep learning. The stochastic nature of GPLVM allows for uncertainty quantification that indicates the degree of prediction error or reliability of prediction without requiring validation data. The applicability for supersonic/hypersonic viscous flowfields has been examined for two cases including axisymmetric intakes and two-dimensional fuel injection in scramjet engines by comparison with other predictive models. Comparable or superior prediction accuracy over the other models has been achieved by the proposed approaches, demonstrating its high potential to serve as a new competent, data-driven technique for fast, accurate, and reliable prediction of scramjet flowfields.

Uncertainty quantification analysis for simulation of wakes in wind-farms using a stochastic RANS solver, compared with a deep learning approach

Quantification of uncertainties in Reynolds–Averaged Navier–Stokes (RANS) simulations has gained a considerable interest in turbulence modeling. We present two different approaches for the quantification and propagation of model-form and operational uncertainties in context of wind turbine RANS simulations. The first approach is based on a stochastic RANS solver in OpenFOAM using intrusive polynomial chaos method (Parekh and Verstappen, 2023). Here the uncertainties are propagated through a single (large) simulation for the coupled coefficients of the polynomial expansion. The second approach is a surrogate based uncertainty quantification (SBUQ) method. The surrogate model comprises of a 3D U-Net neural network (trained over a single wind turbine) combined with a wake superposition model in order to the prediction of flow field in an array of wind turbines. The above-mentioned approaches are applied for uncertainty quantification analysis in RANS simulations of two turbulent engineering flow problems — (i) a wake past a single wind turbine, and (ii) wake interactions and power losses in an array of wind turbines. The results show that the uncertain RANS solutions from the two approaches are able to reasonably capture the reference high-fidelity solution. We also discuss comparisons between the two approaches including computational cost, applicability, generality etc. The two methods can be further explored and applied to engineering applications where it is critical to compute the turbulent RANS solution in presence of various sources of uncertainties.

End-to-end wind turbine wake modelling with deep graph representation learning

Wind turbine wake modelling is of crucial importance to accurate resource assessment, to layout optimisation, and to the operational control of wind farms. This work proposes a surrogate model for the representation of wind turbine wakes based on a state-of-the-art graph representation learning method termed a graph neural network. The proposed end-to-end deep learning model operates directly on unstructured meshes and has been validated against high-fidelity data, demonstrating its ability to rapidly make accurate 3D flow field predictions for various inlet conditions and turbine yaw angles. The specific graph neural network model employed here is shown to generalise well to unseen data and is less sensitive to over-smoothing compared to common graph neural networks. A case study based upon a real world wind farm further demonstrates the capability of the proposed approach to predict farm scale power generation. Moreover, the proposed graph neural network framework is flexible and highly generic and as formulated here can be applied to any steady state computational fluid dynamics simulations on unstructured meshes.

Fast prediction of mine flow field based on convolution neural network

Accurately describing the flow state of mine gas is the basis for the prevention and control of mine dust and toxic gases, computational fluid dynamics (CFD) method is often used to solve mine flow field, but CFD method is usually an iterative process with high calculation cost, time-consuming and high memory requirements. In order to realize the fast solution of mine flow field, a prediction model of mine flow field based on convolutional neural network (CNN) is proposed. The geometry flow field simulated by lattice Boltzmann method (LBM) is used as the training data. CNN is used to extract the basic characteristics of the flow field in the training data, establish the mapping relationship between the geometric boundary and the flow field, and quickly predict the mine flow field. The simulation experiment and the particle image velocimetry (PIV) test experiment of the air window flow field were carried out respectively. In LBM simulation experiment, the average values of mean square error (MSE), mean absolute error (MAE), R2, explained variance score (EVC), Pearson correlation coefficient (PCC) and cosine similarity between CNN predicted values and LBM simulated values of 10 groups of test samples are 0.2633, 0.2449, 0.9595, 0.9697, 0.9827, 0.9933, respectively. It shows that the accuracy of CNN flow field prediction model is similar to that of LBM simulation, and the computational speed of CNN model is increased by three orders of magnitude compared with LBM simulation. In PIV test experiment, the predicted values of CNN model are basically consistent with the PIV experiment results, no matter the velocity distribution of a section or the flow trend of the overall flow field. The LBM simulation experiment and PIV test experiment strongly prove the reliability and generalization ability of CNN model for predicting mine flow field.

Simulation of acoustic oscillatory flows around a curvature controlled by a plasma actuator

In this study, simulations based on the three-dimensional compressible Navier–Stokes equations were conducted to clarify a control method with a plasma actuator to reduce the acoustic power loss in the oscillatory flow under acoustic excitation (acoustic oscillatory flow) in a curved duct. The acoustic power loss for the curved duct exceeds that for the straight duct, particularly at a high acoustic Reynolds number based on the amplitude of the velocity fluctuations in the duct. Flow control by a plasma actuator, which was installed near the curvature in the straight section, was applied to suppress the periodically occurring flow separation around the curvature, which leads to energy loss in the baseline flow. The plasma actuator was intermittently driven at the same frequency as the acoustic oscillatory flow. It was thus demonstrated in the present numerical simulations that the acoustic power loss around the curvature was reduced by the flow control with the plasma actuator. Furthermore, efficient control conditions for the number of plasma actuators and the driving voltage are discussed considering the power consumption of the plasma actuator, although the power efficiency of the control needs to be improved for actual application. Moreover, the relatively effective intermittent control conditions of the phase and duty ratio for the reduction of acoustic power loss are presented with the predicted flow fields around the curvature.