Artificial neural network-based subgrid-scale models for LES of compressible turbulent channel flow

Fully connected neural networks(FCNNs)have been developed for the closure of subgrid-scale(SGS)stress and SGS heat flux in large-eddy simulations of compressible turbulent channel flow.The FCNNbased SGS model trained using data with Mach number Ma=3.0 and Reynolds number Re=3000 was applied to situations with different Mach numbers and Reynolds numbers.The input variables of the neural network model were the filtered velocity gradients and temperature gradients at a single spatial grid point.The a priori test showed that the FCNN model had a correlation coefficient larger than 0.91 and a relative error smaller than 0.43,with much better reconstructions of SGS unclosed terms than the dynamic Smagorinsky model(DSM).In a posteriori test,the behavior of the FCNN model was marginally better than that of the DSM in predicting the mean velocity profiles,mean temperature profiles,turbulent intensities,total Reynolds stress,total Reynolds heat flux,and mean SGS flux of kinetic energy,and outperformed the Smagorinsky model.

Algebraic Calculation Method of One-Dimensional Steady Compressible Gas Flow

This paper describes a new method of calculation of one-dimensional steady compressible gas flows in channels with possible heat and mass exchange through perforated sidewalls. The channel is divided into small elements of a finite size for which mass, energy and momentum conservation laws are written in the integral form, assuming linear distribution of the parameters along the length. As a result, the calculation is reduced to finding the roots of a quadratic algebraic equation, thus providing an alternative to numerical methods based on differential equations. The advantage of this method is its high tolerance to coarse discretization of the calculation area as well as its good applicability for transonic flow calculations.

Application of shifted lattice model to 3D compressible lattice Boltzmann method

An additional potential energy distribution function is introduced on the basis of previous D3Q25 model,and the equilibrium distribution function of D3Q25 is obtained by spherical function.A novel three-dimensional(3D)shifted lattice model is proposed,therefore a shifted lattice model is introduced into D3Q25.Under the finite volume scheme,several typical compressible calculation examples are used to verify whether the numerical stability of the D3Q25 model can be improved by adding the shifted lattice model.The simulation results show that the numerical stability is indeed improved after adding the shifted lattice model.

A new dynamic subgrid-scale model using artificial neural network for compressible flow

The subgrid-scale(SGS)kinetic energy has been used to predict the SGS stress in compressible flow and it was resolved through the SGS kinetic energy transport equation in past studies.In this paper,a new SGS eddy-viscosity model is proposed using artificial neural network to obtain the SGS kinetic energy precisely,instead of using the SGS kinetic energy equation.Using the infinite series expansion and reserving the first term of the expanded term,we obtain an approximated SGS kinetic energy,which has a high correlation with the real SGS kinetic energy.Then,the coefficient of the modelled SGS kinetic energy is resolved by the artificial neural network and the modelled SGS kinetic energy is more accurate through this method compared to the SGS kinetic energy obtained from the SGS kinetic energy equation.The coefficients of the SGS stress and SGS heat flux terms are determined by the dynamic procedure.The new model is tested in the compressible turbulent channel flow.From the a posterior tests,we know that the new model can precisely predict the mean velocity,the Reynolds stress,the mean temperature and turbulence intensities,etc.

Numerical simulation on centrifugal pump compressible flow field with different gas volume fractions

In order to study the influence of gas-liquid two-phase flow on the performance and internal flow field of a centrifugal pump,the steady three-dimensional flow with different gas volume fractions was simulated by applying the Reynolds-average N-S equation and mixture gas-liquid two-phase flow model,and the compressibility of gas was taken into consideration in the simulation. Then the centrifugal pump characteristic and the gas distribution law in different gas volume fractions were analyzed. The computational results show that gas volume fraction has a certain influence on the performance of the centrifugal pump,and the efficiency and head of the pump are on the decline with the increase of it.Static pressure in the impeller increases in the radial direction,but the pressure gradient in the flow direction is different under the different gas volume fractions. The gas volume is distributed mainly in the ipsilateral direction of impeller back shroud in the flow channel of the volute. On the suction side of the blade inlet there is an obvious low-pressure area,which causes bubbles agglutination and higher gas volume fraction. With the gas entering passage flow,gas volume fraction in the suction decreases and the pressure surface rises gradually. Higher gas volume fraction causes air blocking phenomenon in the flow passage and the discharge capacity reduces. The increase of gas volume makes the turbulent motion within the impeller more and more intense,which leads to more and more energy loss.

Gas flow characteristics of argon inductively coupled plasma and advections of plasma species under incompressible and compressible flows

In this work, incompressible and compressible flows of background gas are characterized in argon inductively coupled plasma by using a fluid model, and the respective influence of the two flows on the plasma properties is specified. In the incompressible flow, only the velocity variable is calculated, while in the compressible flow, both the velocity and density variables are calculated. The compressible flow is more realistic; nevertheless, a comparison of the two types of flow is convenient for people to investigate the respective role of velocity and density variables. The peripheral symmetric profile of metastable density near the chamber sidewall is broken in the incompressible flow. At the compressible flow, the electron density increases and the electron temperature decreases. Meanwhile, the metastable density peak shifts to the dielectric window from the discharge center, besides for the peripheral density profile distortion, similar to the incompressible flow.The velocity profile at incompressible flow is not altered when changing the inlet velocity, whereas clear peak shift of velocity profile from the inlet to the outlet at compressible flow is observed as increasing the gas flow rate. The shift of velocity peak is more obvious at low pressures for it is easy to compress the rarefied gas. The velocity profile variations at compressible flow show people the concrete residing processes of background molecule and plasma species in the chamber at different flow rates. Of more significance is it implied that in the usual linear method that people use to calculate the residence time, one important parameter in the gas flow dynamics, needs to be rectified. The spatial profile of pressure simulated exhibits obvious spatial gradient. This is helpful for experimentalists to understand their gas pressure measurements that are always taken at the chamber outlet. At the end, the work specification and limitations are listed.

REVIEW ON MATHEMATICAL ANALYSIS OF SOME TWO-PHASE FLOW MODELS

The two-phase flow models are commonly used in industrial applications, such as nuclear, power, chemical-process, oil-and-gas, cryogenics, bio-medical, micro-technology and so on. This is a survey paper on the study of compressible nonconservative two-fluid model, drift-flux model and viscous liquid-gas two-phase flow model. We give the research developments of these three two-phase flow models, respectively. In the last part, we give some open problems about the above models.

Dynamic modeling of twin roll casting AZ41 magnesium alloy during hot compression processing

A dynamic material model of Mg-4.51Al-1.19Zn-0.5Mn-0.5Ca(AZ41,mass fraction,%)magnesium alloy was put forward.The results show that the dynamic material model can characterize the deformation behavior and microstructure evolution and describe the relations among flow stress,strain,strain rates and deformation temperatures.Statistical analysis shows the validity of the proposed model.The model predicts that lower deformation temperature and higher strain rate cause the sharp strain hardening. Meanwhile,the flow stress curve turns into a steady state at high temperature and lower strain rate.The moderate temperature of 350 ℃and strain rate of 0.01 s-1 are appropriate to this alloy.

全钒液流电池多孔电极非均匀压缩的数值模拟

电极压缩会导致全钒液流电池的多孔电极侵入流道影响电池性能。本文建立了蛇形流道多孔电极非均匀压缩的三维稳态模型,对比研究了多孔电极非均匀压缩和多孔电极矩形压缩的内部压降、电解质流速和过电位分布的不同,发现多孔电极非均匀压缩与实际情况符合程度高。分析了不同压缩比对电极非均匀压缩多孔电极内部压降、电解质速度、反应物质浓度分布、电解质电位、过电位、局部电流密度和浓差过电位的影响。结果表明:在压缩比增大的情况下,电极非均匀压缩多孔电极内部压降和速度增大程度高于电极矩形压缩,电极非均匀压缩侵入部分与流道接触区域处过电位高于电极矩形压缩。电极非均匀压缩多孔电极内部反应物质浓度分布均匀性增大,电解质电位和过电位降低,阴极区域局部电流密度和浓差过电位上升。

粉末冶金Al-5.6Zn-2Mg合金在热压缩过程中加工硬化特性及本构模型的比较

在300~5000℃、应变速率为0.1~0.0001 s−1的条件下,在万能试验机上对Al-5.6Zn-2Mg铝合金进行热压缩试验,以确定动态再结晶开始时的加工硬化速率曲线σ_(c)(ε_(c))l以及关键特征σ_(c)(ε_(c))l、σ_(p)(ε_(p))和σ_(ss)与Z系数之间的相关性。使用了四个本构模型,Arrhenius模型、改进的Johnson-Cook模型(MJC)、改进的Zerilli-Armstrong模型(MZA)和开发的人工神经网络(ANN)。结果表明,ANN型模型和Arrhenius模型的相对误差的绝对平均值最低,分别为0.486%和3.36%,MZA型模型和MJC型模型的相对误差的绝对平均值较高,分别为8.84%和3.93%。由于Arrhenius模型能够处理各种因素之间的非线性关系,因此它被认为是最合适的预测模型,但在材料性质未知或实验数据有限的情况下,MJC模型可能是一种更简单的替代方法。MZA模型不适合估计热压缩时的流变应力。此外,训练最好的神经网络模型的预测性能最好,相对误差的绝对平均值为0.486%,R值为0.99。

基于可压缩两相流模型的注射成型填充模拟

基于现有的超算平台及有限体积法计算程序,开发了一种用于注射成型的三维数值模拟计算模型。该模型数值方法基于可压缩两相流,考虑了熔体和气体的密度变化,即将熔体和空气这两种不同性质的流体,采用同一套控制方程进行计算。采用有限体积法进行离散,通过压力隐式算子分割算法(PISO)求解压力-速度耦合方程组,结合流体体积(VOF)法进行界面追踪。首先模拟了带圆柱嵌件的平板型腔填充过程,采用六面体为主导的方法划分了三种不同网格系统,定量比较三种网格下的速度基本一致,从而验证了网格的收敛性。考虑到计算效率,选取M2(网格数为106876个)网格系统,模拟了温度、剪切速率和压力在不同截面上的分布。结果表明,在填充时间(t)为0.6,0.8,1.4 s时刻下熔体流体前沿的界面追踪模拟与实验吻合;在更复杂的不均匀厚度凹槽平板熔体填充模拟过程中,塑料熔体注入型腔时,会先在薄壁凹槽内流动,随后流进厚壁区域。在t为0.125,0.25,0.375 s时刻下的流动前沿会产生形状像喷泉的流锋,这种流动现象可以通过三维模型预测,与商业软件仿真结果呈现基本一致。随着填充速率的增大,注射压力数值与Moldflow结果更接近。

不同环境参数对300 MW压缩空气储能系统运行的影响

压缩空气储能系统是提高电力系统稳定性的有效手段,为研究环境温度和湿度对压缩空气储能系统性能的影响,基于MSP平台建立300 MW压缩空气储能系统热力学模型,研究环境温度和湿度对压缩过程性能的影响规律。结果表明,随空气温度增大,压缩机总功率增大,压缩过程时间增大;随空气湿度增大,压缩机总功率增大,压缩过程时间增大且变化幅度极小。研究结论对压缩空气储能系统的实际运行具有一定的参考意义。

基于机器学习的Ti-5Al-1.5Mo-1.8Fe低成本钛合金热加工图预测

为了研究Ti-5Al-1.5Mo-1.8Fe低成本钛合金的热变形行为,运用Instron 5869压缩实验机进行热压缩实验。构建了以变形温度、应变速率、应变为输入变量和流变应力为输出变量的6种机器学习模型,预测不同条件下该合金的流变应力值并评估检验模型的预测性能。根据预测表现最好的LSTM神经网络模型的预测数据绘制预测加工图,对照实验加工图评估检验其预测能力。结果表明:预测加工图能够较为准确地反映出Ti-5Al-1.5Mo-1.8Fe合金在应变为0.499时的可加工区域,与实验加工图的吻合程度较高,该方法能较好地预测Ti-5Al-1.5Mo-1.8Fe合金的热变形行为。

Investigation on strong nonlinear interactions between underwater explosion and water surface based on compressible multiphase flow with phase transition

The strong nonlinear interactions between underwater explosion and water surface were numerically investigated using a phase transition model based on a four-equation system,which can deal with the complex deformable interface among different phases,including water,air,explosion bubble,and cavitation.The numerical method is verified by comparing the numerical results with experimental results,and good agreements are found.This study considers an ideal sine wave for simulating the shape of water surface.Two examples of different detonation depths of charge are investigated.In each example,the first case is the basic simulation without surface wave,and the other three cases are the simulations with sine waves of different wavelengths.Unique characteristics of the interactions,such as shock wave propagation,explosion bubble expansion,and the generation,development,and collapse of cavitation,are observed in the numerical simulations.By capturing the detailed density and pressure contours during the interaction process,we can better understand the underlying mechanisms of the explosion bubble,cavitation,and surface waves.These numerical results demonstrate that geometric nonlinearity impacts cavitation evolution and the explosion bubble movement mechanism.Additionally,the secondary cavitation phenomenon has been found in the cases without surface wave,and its fundamental physical mechanism is presented in detail.The present results can expand the existing database of multiphase flow in the underwater explosion and provide an insight into the strong nonlinear interaction between the underwater explosion and water surface.

Lattice Boltzmann modeling and simulation of compressible flows

In this mini-review we summarize the progress of Lattice Boltzmann （LB） modeling and simulating compressible flows in our group in recent years. Main contents include （i） Single-Relaxation-Time （SRT） LB model supplemented by additional viscosity, （ii） Multiple-Relaxation-Time （MRT） LB model, and （iii） LB study on hydrodynamic instabilities. The former two belong to improvements of physical modeling and the third belongs to simulation or application. The SRT-LB model sup- plemented by additional viscosity keeps the original framework of Lattice Bhatnagar-Gross Krook （LBGK）. So, it is easier and more convenient for previous SRT-LB users. The MRT-LB is a com- pletely new framework for physical modeling. It significantly extends the range of LB applications. The cost is longer computational time. The developed SRT-LB and MRT-LB are complementary from the sides of convenience and applicability.

Discrete Boltzmann model for implosion- and explosion- related compressible flow with spherical symmetry

To kinetically model implosion- and explosion-related phenomena, we present a theoretical framework for constructing a discrete Boltzmann model （DBM） with spherical symmetry in spherical coordinates. To achieve this goal, a key technique is to use local Cartesian coordinates to describe the particle velocity in the kinetic model. Therefore, geometric effects, such as divergence and convergence, are described as a ＂force term＂. To better access the nonequilibrium behavior, even though the corre- sponding hydrodynamic model is one-dimensional, the DBM uses a discrete velocity model （DVM） with three dimensions. A new scheme is introduced so that the DBM can use the same DVM regard- less of whether or not there are extra degrees of freedom. As an example, a DVM with 26 velocities is formulated to construct the DBM at the Navier-Stokes level. Via the DBM, one can study simulta- neously the hydrodynamic and thermodynamic nonequilibrium behaviors in implosion and explosion processes that are not very close to the spherical center. The extension of the current model to a multiple-relaxation-time version is straightforward.

Method of coupling 1-D unsaturated flow with 3-D saturated flow on large scale

A coupled unsaturated-saturated water flow numerical model was developed. The water flow in the unsaturated zone is considered the one-dimensional vertical flow, which changes in the horizontal direction according to the groundwater table and the atmospheric boundary conditions. The groundwater flow is treated as the three-dimensional water flow. The recharge flux to groundwater from soil water is considered the bottom flux for the numerical simulation in the unsaturated zone, and the upper flux for the groundwater simulation. It connects and unites the two separated water flow systems. The soil water equation is solved based on the assumed groundwater table.and the subsequent predicted recharge flux. Then, the groundwater equation is solved with the predicted recharge flux as the upper boundary condition. Iteration continues until the discrepancy between the assumed and calculated groundwater nodal heads have a certain accuracy. Illustrative examples with different water flow scenarios regar.ding the Dirichlet boundary condition, the Neumann boundary condition, the a.tmospheric boundary condition, and the source or sink term were calculated by the coupled model. The results are compared with those of other models, including Hydrus-lD, SWMS-2D, and FEFLOW, which demonstrate that the coupled model is effective and accurate and can significantly reduce the computational time for the large number of nodes in saturated-unsaturated water flow simulation.

On the Liquid-Vapor Phase-Change Interface Conditions for Numerical Simulation of Violent Separated Flows

Numerous models have been proposed in the literature to include phase change into numerical simulations of two-phase flows.This review paper presents the modeling options that have been taken in order to obtain a model for violent separated flows with application to sloshing wave impacts.A relaxation model based on linear non-equilibrium thermodynamics has been chosen to compute the rate of phase change.The integration in the system of partial differential equations is done through a non-conservative advection term.For each of these modelling choices,some alternative models from the literature are presented and discussed.The theoretical framework for all phase change model(conservation equations and entropy growth)is also summarized.

Development and Comparative Studies of Three Non-free Parameter Lattice Boltzmann Models for Simulation of Compressible Flows

This paper at first shows the details of finite volume-based lattice Boltzmann method(FV-LBM)for simulation of compressible flows with shock waves.In the FV-LBM,the normal convective flux at the interface of a cell is evaluated by using one-dimensional compressible lattice Boltzmann model,while the tangential flux is calculated using the same way as used in the conventional Euler solvers.The paper then presents a platform to construct one-dimensional compressible lattice Boltzmann model for its use in FV-LBM.The platform is formed from the conservation forms of moments.Under the platform,both the equilibrium distribution functions and lattice velocities can be determined,and therefore,non-free parameter model can be developed.The paper particularly presents three typical non-free parameter models,D1Q3,D1Q4 and D1Q5.The performances of these three models for simulation of compressible flows are investigated by a brief analysis and their application to solve some one-dimensional and two-dimensional test problems.Numerical results showed that D1Q3 model costs the least computation time and D1Q4 and D1Q5 models have the wider application range of Mach number.From the results,it seems that D1Q4 model could be the best choice for the FVLBM simulation of hypersonic flows.