Deep Learning Applied to Computational Mechanics:A Comprehensive Review,State of the Art,and the Classics

Three recent breakthroughs due to AI in arts and science serve as motivation:An award winning digital image,protein folding,fast matrix multiplication.Many recent developments in artificial neural networks,particularly deep learning(DL),applied and relevant to computational mechanics(solid,fluids,finite-element technology)are reviewed in detail.Both hybrid and pure machine learning(ML)methods are discussed.Hybrid methods combine traditional PDE discretizations with ML methods either(1)to help model complex nonlinear constitutive relations,(2)to nonlinearly reduce the model order for efficient simulation(turbulence),or(3)to accelerate the simulation by predicting certain components in the traditional integration methods.Here,methods(1)and(2)relied on Long-Short-Term Memory(LSTM)architecture,with method(3)relying on convolutional neural networks.Pure ML methods to solve(nonlinear)PDEs are represented by Physics-Informed Neural network(PINN)methods,which could be combined with attention mechanism to address discontinuous solutions.Both LSTM and attention architectures,together with modern and generalized classic optimizers to include stochasticity for DL networks,are extensively reviewed.Kernel machines,including Gaussian processes,are provided to sufficient depth for more advanced works such as shallow networks with infinite width.Not only addressing experts,readers are assumed familiar with computational mechanics,but not with DL,whose concepts and applications are built up from the basics,aiming at bringing first-time learners quickly to the forefront of research.History and limitations of AI are recounted and discussed,with particular attention at pointing out misstatements or misconceptions of the classics,even in well-known references.Positioning and pointing control of a large-deformable beam is given as an example.

A Computational Fluid Dynamics (CFD) Analysis of an Undulatory Mechanical Fin Driven by Shape Memory Alloy

Many fishes use undulatory fin to propel themselves in the underwater environment. These locomotor mechanisms have a popular interest to many researchers. In the present study, we perform a three-dimensional unsteady computation of an undulatory mechanical fin that is driven by Shape Memory Alloy （SMA）. The objective of the computation is to investigate the fluid dynamics of force production associated with the undulatory mechanical fin. An unstructured, grid-based, unsteady Navier-Stokes solver with automatic adaptive remeshing is used to compute the unsteady flow around the fin through five complete cycles. The pressure distribution on fin surface is computed and integrated to provide fin forces which are decomposed into lift and thrust. The velocity field is also computed throughout the swimming cycle. Finally, a comparison is conducted to reveal the dynamics of force generation according to the kinematic parameters of the undulatory fin （amplitude, frequency and wavelength）.

COMPUTATIONAL FLUID DYNAMICS(CFD) SIMULATIONS OF DRAG REDUCTION WITH PERIODIC MICRO-STRUCTURED WALL

Computational fluid dynamics（CFD） simulations are adopted to investigate rectangular microchannel flows with various periodic micro-structured wall by introducing velocity slip boundary condition at low Reynolds number. The purpose of the current study is to numerically find out the effects of periodic micro-structured wall on the flow resistance in rectangular microchannel with the different spacings between microridges ranging from 15 to 60 pm. The simulative results indicate that pressure drop with different spacing between microridges increases linearly with flow velocity and decreases monotonically with slip velocity; Pressure drop reduction also increases with the spacing between microridges at the same condition of slip velocity and flow velocity. The results of numerical simulation are compared with theoretical predictions and experimental results in the literatures. It is found that there is qualitative agreement between them.

Computational fluid dynamics simulation of gas-liquid two phases flow in 320 m^3 air-blowing mechanical flotation cell using different turbulence models

According to the recently developed single-trough floating machine with the world's largest volume(inflatable mechanical agitation flotation machine with volume of 320 m3) in China, the gas-fluid two-phase flow in flotation cell was simulated using computational fluid dynamics method. It is shown that hexahedral mesh scheme is more suitable for the complex structure of the flotation cell than tetrahedral mesh scheme, and a mesh quality ranging from 0.7 to 1.0 is obtained. Comparative studies of the standard k-ε, k-ω and realizable k-ε turbulence models were carried out. It is indicated that the standard k-ε turbulence model could give a result relatively close to the practice and the liquid phase flow field is well characterized. In addition, two obvious recirculation zones are formed in the mixing zones, and the pressure on the rotor and stator is well characterized. Furthermore, the simulation results using improved standard k-ε turbulence model show that surface tension coefficient of 0.072, drag model of Grace and coefficient of 4, and lift coefficient of 0.001 can be achieved. The research results suggest that gas-fluid two-phase flow in large flotation cell can be well simulated using computational fluid dynamics method.

Computational Fluid Dynamics(CFD) Analysis and Optimization of Reconstructed Intake System of Cylinder Head Based on Slicing Reverse Method

To find out and improve the flow characteristics inside the intake system of cylinder head,the application of computational fluid dynamics(CFD)in the evaluation and optimization of the reconstructed intake system based on slicing reverse method was proposed.The flow characteristics were found out through CFD,and the velocity vector field,pressure field and turbulent kinetic energy field for different valve lifts were discussed,which were in good agreement with experimental data,and the quality of reconstruction was evaluated.In order to improve its flow characteristic,an optimization plan was proposed.The results show that the flow characteristics after optimization are obviously improved.The results can provide a reference for the design and optimization of the intake system of cylinder head.

Computational Fluid Dynamics (CFD) Modelling to Estimate Fluvial Bank Erosion—A Case Study

River bank erosion models are an important prerequisite for understanding the development of river meanders and for estimating likely land-loss and potential danger to floodplain infrastructure. Although bank erosion models have been developed that consider large-scale mass failure, the contribution of fluvial erosion (the process of particle-by-particle erosion due to the shearing action of the river flow) to bank retreat has not received as much consideration. In principle, such fluvial bank erosion rates can be quantified using excess shear stress formulations, but in practice, it has proven difficult to estimate the parameters involved. In this study, a series of three-dimensional Computational Fluid Dynamics (CFD) simulations for a meander loop on the River Asker (200 m long) at Bridport in southern England were undertaken to elucidate the overall flow structures and in particular to provide estimates of the applied fluid shear stress exerted on the riverbanks. The CFD models, which simulated relatively low and relatively high flow conditions, were established using Fluent 6.2 software. The modelling outcomes show that the key qualitative features of the flow endure even as flow discharge varies. At bank full, the degrees of velocity and simulated shear stresses within the inner bank separation zones are shown to be higher than those observed under low flow conditions, and that these elevated shear stresses may be sufficient to result in the removal of accumulated sediments into the main downstream flow.

Painlevé Analysis,Soliton Solutions and Bcklund Transformation for Extended (2 + 1)-Dimensional Konopelchenko-Dubrovsky Equations in Fluid Mechanics via Symbolic Computation

This paper is to investigate the extended(2+1)-dimensional Konopelchenko-Dubrovsky equations,which can be applied to describing certain phenomena in the stratified shear flow,the internal and shallow-water waves, plasmas and other fields.Painleve analysis is passed through via symbolic computation.Bilinear-form equations are constructed and soliton solutions are derived.Soliton solutions and interactions are illustrated.Bilinear-form Backlund transformation and a type of solutions are obtained.

Introducing CFD and Wind Tunnel Testing in an Undergraduate Fluid Mechanics Course

CFD （computational fluid dynamics） is following the trend of CAD and FEA （finite element analysis） to undergraduate education especially with recent advances in commercial codes. It will soon take its place as an expected skill for new engineering graduates. CFD was added as a component to an experiment in a junior level fluid mechanics course. The objectives were to introduce CFD, as an analysis tool, to the students and to support the theoretical concepts of the course. The students were asked to complete an experimental two-dimensional study for a wing in a wind tunnel, to use CFD to simulate the flow, and to predict the aerodynamic lift using CFD as well as the experimentally obtained pressure distribution. In addition, they had to compare their results to published data for the studied wing. Details of the course, the wind tunnel test and the CFD simulations are presented. Samples from the students＇ work are used in the discussion. The lab activities were successfully completed by the students and the learning objectives were well addressed. One of the valuable outcomes from this lab was the opportunity for the students to integrate multiple fluid mechanics analysis tools and learn about the limits for each tool. CFD also enhanced the learning in the lab activities and increased students＇ interest in the subject.

基于CFD-DEM耦合的梯级溜槽的设计与分析

针对传统物料转运过程中溜槽和输送带磨损严重、出口处粉尘浓度过高的问题,建立含臂架的梯级溜槽几何模型,采用基于计算流体力学与离散单元法(CFD-DEM)耦合的数值模拟方法,分析了臂架对转运溜槽的磨损以及对其出口处粉尘排放浓度的影响。仿真结果表明:含臂架的梯级溜槽可以有效控制物料流的速度和方向,降低对溜槽内表面的冲击磨损,降低出口处的粉尘量。

Application of Computational Fluid Dynamics and Fluid Structure Interaction Techniques for Calculating the 3D Transient Flow of Journal Bearings Coupled with Rotor Systems

Journal bearings are important parts to keep the high dynamic performance of rotor machinery. Some methods have already been proposed to analysis the flow field of journal bearings, and in most of these methods simplified physical model and classic Reynolds equation are always applied. While the application of the general computational fluid dynamics （CFD）-fluid structure interaction （FSI） techniques is more beneficial for analysis of the fluid field in a journal bearing when more detailed solutions are needed. This paper deals with the quasi-coupling calculation of transient fluid dynamics of oil film in journal bearings and rotor dynamics with CFD-FSI techniques. The fluid dynamics of oil film is calculated by applying the so-called ＂dynamic mesh＂ technique. A new mesh movement approacb is presented while the dynamic mesh models provided by FLUENT are not suitable for the transient oil flow in journal bearings. The proposed mesh movement approach is based on the structured mesh. When the joumal moves, the movement distance of every grid in the flow field of bearing can be calculated, and then the update of the volume mesh can be handled automatically by user defined function （UDF）. The journal displacement at each time step is obtained by solving the moving equations of the rotor-bearing system under the known oil film force condition. A case study is carried out to calculate the locus of the journal center and pressure distribution of the journal in order to prove the feasibility of this method. The calculating results indicate that the proposed method can predict the transient flow field of a journal bearing in a rotor-bearing system where more realistic models are involved. The presented calculation method provides a basis for studying the nonlinear dynamic behavior of a general rotor-bearing system.

Computational Fluid Dynamics Approach to the Effect of Mixing and Draft Tube on the Precipitation of Barium Sulfate in a Continuous Stirred Tank

The effect of mixing on the precipitation of barium sulfate in a continuous stirred tank is simulated numerically with different feeding location, feed concentration, impeller speed and residence time through solving the standard momentum and mass transport equations in combination with the moment equations for crystal population balance. The numerical method was validated with the literature data. The simulation results including the distribution of the local supersaturation ratio distribution in the precipitator, mean crystal size and coefficient of variation under different operating conditions compared well with experimental data in the literature. The effect of the presence of a draft tube on precipitation were also investigated, and it is suggested that the installation of a draft tube increased the mean crystal size, in general agreement with experimental work in the literature.

Analysis of a Propeller Wake Flow Field Using Viscous Fluid Mechanics

The computational fluid dynamics （CFD） method is used to numerically simulate a propeller wake flow field in open water. A sub-domain hybrid mesh method was adopted in this paper. The computation domain was separated into two sub-domains, in which tetrahedral elements were used in the inner domain to match the complicated geometry of the propeller, while hexahedral elements were used in the outer domain. The mesh was locally refined on the propeller surface and near the wake flow field, and a size function was used to control the growth rate of the grid. Sections at different axial location were used to study the spatial evolution of the propeller wake in the region ranging from the disc to one propeller diameter （D） downstream. The numerical results show that the axial velocity fluctuates along the wake flow; radial velocity, which is closely related to vortices, attenuates strongly. The trailing vortices interact with the tip vortex at the blades＇ trailing edge and then separate. The strength of the vortex shrinks rapidly, and the radius decreases 20% at one diameter downstream.

调谐液柱阻尼器-结构系统风致振动响应的CFD/CSD耦合分析方法

针对调谐液柱阻尼器(tuned liquid column damper, TLCD)难以建立其精确的非线性理论分析模型,且其力学性能试验成本高和耗时长等问题,首先采用计算流体动力学(CFD)数值模拟方法,对TLCD系统的力学性能和动力特征进行仿真模拟,在此基础上进一步提出了基于计算流体动力学/计算结构动力学(CFD/CSD)耦合分析方法,求解带TLCD系统的高层建筑结构的风致动力响应。通过开展某一TLCD系统在特定底部激励下的力学性能和动力特性试验,得到其内液体晃荡的自由液面波高和晃动力时程,验证了CFD数值模拟方法可以准确地分析TLCD水箱内液体的非线性晃动特征。随后对风工程领域广泛采用的76层建筑结构振动控制Benchmark模型,假设其顶部设置TLCD系统时主体结构在三种风速重现期(10、50和100年)风速对应的横风向动力风荷载激励下的风致控制效率,采用提出的CFD/CSD耦合分析方法,进行了数值仿真模拟分析。耦合分析结果表明,TLCD系统对Benchmark模型的风致加速度、速度和位移响应均有一定的控制效果,对加速度响应的控制效果要优于对位移响应的控制效果。该研究方法可为复杂TLCD系统对高层建筑的风振控制分析提供有效的参考。

Application of computational fluid dynamics simulation for submarine oil spill

Computational fluid dynamics （CFD） codes are being increasingly used in the simulation of submarine oil spills. This study focuses on the process of oil spills, from damaged submarine pipes, to the sea surface, using numerical models. The underwater oil spill model is developed, and a description of the governing equations is proposed, along with modifications required for the particalization of the control volume. Available experimental data were introduced to evaluate the validity of the CFD predictions, the results of which proved to be in good agreement with the experimental data. The effects of oil leak rate, leak diameter, current velocity, and oil density are investigated, by the validated CFD model, to estimate the undersea leakage time, the lateral migration distance, and surface diffusion range when the oil reaches the sea surface. Results indicate that the leakage time and lateral migration distance increase with decreasing leak rates and leak diameter, and increase with increasing current velocity and oil density. On the other hand, a large leak diameter, high density, high leak rate, or fast currents result in a greater surface diffusion range. The findings and analysis presented here will provide practical predictions of oil spills, and guidance for emergency rescues.

Characterization of the Engineering Flow in a Miniaturised Bioreactor by Computational Fluid Dynamics

Three approaches based on computational fluid dynamics(CFD) techniques have been assessed for their ability to describe the engineering flow environment in a miniaturized mechanically agitated bioreactor. The three approaches tested were the source-sink(SS), the multiple reference frames(MRF) and the sliding grids(SG). In all the cases, the predictions of the velocity components agree with reported experimental data. However, the analysis of the results of the turbulent intensities predicted by the three approaches indicates the MRF and the SG techniques under predicted turbulent intensities are comparable to both experimental measurements and the SS method. The predicted power number and pumping number based on the SS approach are closer to typical reported experimental values compared to those obtained from the MRF and SG methods.

The influence of temperature on flow-induced forces on quartz-crystal-microbalance sensors in a Chinese liquor identification electronic-nose: three-dimensional computational fluid dynamics simulation and analysis

An electronic-nose is developed based on eight quartz-crystal-microbalance (QCM) gas sensors in a sensor box, and is used to detect Chinese liquors at room temperature. Each sensor is a highly-accurate and highly-sensitive oscillator that has experienced airflow disturbances under the condition of varying room temperatures due to unstable flow-induced forces on the sensors surfaces. The three-dimensional (3D) nature of the airflow inside the sensor box and the interactions of the airflow on the sensors surfaces at different temperatures are studied by computational fluid dynamics (CFD) tools. Higher simulation accuracy is achieved by optimizing meshes, meshing the computational domain using a fine unstructural tetrahedron mesh. An optimum temperature, 30 ℃, is obtained by analyzing the distributions of velocity streamlines and the static pressure, as well as the flow-induced forces over time, all of which may be used to improve the identification accuracy of the electronic-nose for achieving stable and repeatable signals by removing the influence of temperature.

基于CFD茶叶烘干机的优化设计与试验

针对目前6CH型自动链板式茶叶烘干机存在结构设计欠合理、温度分布不均匀性等问题,对6CH型自动链板式茶叶烘干机进行优化设计与试验。将导流板的布置进行优化设计,对优化后的茶叶烘干机采用计算流体力学软件CFD建立干燥机部件的稳态温度场仿真模型,通过截面温度、速度云图,分析得到3组工况下(进风速度20 m·s^(-1)温度130℃、进风速度25 m·s^(-1)温度140℃、进风速度30 m·s^(-1)温度150℃),干燥室内部均能满足所需高温环境,并且能满足物料快速烘干再掉落至下一层的要求。采用三因素三水平正交试验,确定实际工况下的最优参数组合,试验结果表明,当进风温度为144.81℃、鼓风机风速为20.62 m·s^(-1)、输送速度为1.78 m·s^(-1),基于此组合下的降水幅度为40.75%,碎茶率为2.07%,均满足且超过农艺要求,该研究为茶叶烘干与烘干机关键参数优化提供参考。

COMPUTATIONAL FLUID DYNAMICS RESEARCH ON PRESSURE LOSS OF CROSS-FLOW PERFORATED MUFFLER

The pressure loss of cross-flow perforated of physical modeling, simulation and data processing. muffler has been computed with the procedure Three-dimensional computational fluid dynamics （CFD） has been used to investigate the relations of porosities, flow velocity and diameter of the holes with the pressure loss. Accordingly, some preliminary results have been obtained that pressure loss increases with porosity descent as nearly a hyperbolic trend, rising flow velocity of the input makes the pressure loss increasing with parabola trend, diameter of holes affects little about pressure loss of the muffler. Otherwise, the holes on the perforated pipes make the air flow gently and meanly, which decreases the air impact to the wall and pipes in the muffler. A practical perforated muffler is used to illustrate the available of this method for pressure loss computation, and the comparison shows that the computation results with the method of CFD has reference value for muffler design.

基于CFD的FPSO风载荷规范计算适用性研究

[目的]目前,设计人员多使用针对大型油轮的OCIMF规范和针对海上平台的API规范对FPSO风载荷进行计算,但因FPSO船舶上层建筑更为复杂,需进一步研究这2种规范对于FPSO风载荷计算的适用性。[方法]建立具有通用型上部模块的某30万吨级大型FPSO数值模型,对恶劣海况下不同风向角、横倾角下的FPSO所受风载荷进行数值模拟,分析其中存在的遮蔽效应;与规范计算结果进行对比分析,讨论在风载荷作用下FPSO受到的横倾力矩。[结果]结果显示,船舶正浮状态受到的最大风载荷和横倾力矩出现在270°风向角;船舶横倾状态下受到的风载荷和横倾力矩比正浮状态更大,最大横倾力矩出现在10.5°横倾角280°风向角;采用API规范和OCIMF规范得到的FPSO风载荷计算结果与CFD计算结果相差较大,二者在270°风向角的结果与CFD分别相差13.6%和24.5%。[结论]数值仿真给出的流场细节有利于分析上部模块间的遮蔽效应,能够较为准确地预报船舶所受到的风载荷,可以为考虑遮蔽效应的FPSO稳性设计提供参考。

Computational Fluid Dynamics Based Bulbous Bow Optimization Using a Genetic Algorithm

Computational fluid dynamics （CFD） plays a major role in predicting the flow behavior of a ship. With the development of fast computers and robust CFD software, CFD has become an important tool for designers and engineers in the ship industry. In this paper, the hull form of a ship was optimized for total resistance using CFD as a calculation tool and a genetic algorithm as an optimization tool. CFD based optimization consists of major steps involving automatic generation of geometry based on design parameters, automatic generation of mesh, automatic analysis of fluid flow to calculate the required objective/cost function, and finally an optimization tool to evaluate the cost for optimization. In this paper, integration of a genetic algorithm program, written in MATLAB, was carried out with the geometry and meshing software GAMBIT and CFD analysis software FLUENT. Different geometries of additive bulbous bow were incorporated in the original hull based on design parameters. These design variables were optimized to achieve a minimum cost function of ＂total resistance＂. Integration of a genetic algorithm with CFD tools proves to be effective for hull form ootimization.