scholarly journals A Data Augmentation-Based Technique for Deep Learning Applied to CFD Simulations

Mathematics ◽  
2021 ◽  
Vol 9 (16) ◽  
pp. 1843
Author(s):  
Alvaro Abucide-Armas ◽  
Koldo Portal-Porras ◽  
Unai Fernandez-Gamiz ◽  
Ekaitz Zulueta ◽  
Adrian Teso-Fz-Betoño

The computational cost and memory demand required by computational fluid dynamics (CFD) codes simulations can become very high. Therefore, the application of convolutional neural networks (CNN) in this field has been studied owing to its capacity to learn patterns from sets of input data, which can considerably approximate the results of the CFD simulations with relative low errors. DeepCFD code has been taken as a basis and with some slight variations in the parameters of the CNN, while the net is able to solve the Navier–Stokes equations for steady turbulent flows with variable input velocities to the domain. In order to acquire extensive input data to the CNN, a data augmentation technique, which considers the similarity principle for fluid dynamics, is implemented. As a consequence, DeepCFD is able to learn the velocities and pressure fields quite accurately, speeding up the time-consuming CFD simulations.

2012 ◽  
Vol 184-185 ◽  
pp. 944-948 ◽  
Author(s):  
Hai Jun Gong ◽  
Yang Liu ◽  
Xue Yi Fan ◽  
Da Ming Xu

For a clear and comprehensive opinion on segregated SIMPLE algorithm in the area of computational fluid dynamics (CFD) during liquid processing of materials, the most significant developments on the SIMPLE algorithm and its variants are briefly reviewed. Subsequently, some important advances during last 30 years serving as increasing numerical accuracy, enhancing robustness and improving efficiency for Navier–Stokes (N-S) equations of incompressible fluid flow are summarized. And then a so-called Direct-SIMPLE scheme proposed by the authors of present paper introduced, which is different from SIMPLE-like schemes, no iterative computations are needed to achieve the final pressure and velocity corrections. Based on the facts cited in present paper, it conclude that the SIMPLE algorithm and its variants will continue to evolve aimed at convergence and accuracy of solution by improving and combining various methods with different grid techniques, and all the algorithms mentioned above will enjoy widespread use in the future.


1998 ◽  
Vol 120 (03) ◽  
pp. 76-78
Author(s):  
Nathalie Hamill

More-intuitive pre-processors and advanced solvers are making computational fluid dynamics (CFD) software easier to use, more accurate, and faster. CFD techniques involve the solution of the Navier-Stokes equations that describe fluid-flow processes. Using MSC/ PATRAN as a starting point, AEA Technology plc, Harwell, Oxfordshire, England, has developed a pre-processor for its software that is fully computer-aided design (CAD)-compatible and works with native CAD databases such as CADDS 5, CATIA, Euclid3, Pro /ENG INEER, and Unigraphics. The simplicity of modeling complex geometries in CFX allows more details to be included in models, such as gangways between coaches, bogies, and even some parts of the pantograph. CFX 5's coupled solver offers a radically different approach that solves all the hydrodynamic equations as a single system. CFX 5 has demonstrated its ability to deliver much faster pre-processing and shorter run times, thus increasing productivity for its users. CFX 5.2 should be a further step forward in commercial CFD, with its mixed element types combining the accuracy of prismatic meshes adjacent to surfaces with the speed and geometric flexibility of tetrahedral elements in the remainder of the grid.


2003 ◽  
Author(s):  
V. Jalili ◽  
C. Bailey ◽  
M. K. Patel

This is a novel application of Computational Fluid Dynamics (CFD), in the vacuum De-zincing process. The complete modeling process would involve the solution of the following equations: a) Navier-Stokes Equations; b) The Energy Equation; c) The Solution of the Species Concentration. The aim of this research as a novel approach in vacuum Dezincing process has been to gain an understanding in terms of the actual complicated physics involved de-zincing process such, as phase change and solidification. The results in this paper have contributed to a better understanding of the vacuum De-zincing process, hence identifying parameters, which would aid the efficient recovery of the zinc from the molten metal bath.


2019 ◽  
Author(s):  
Alexander M. Molchanov

This book is intended for beginners in the field of Computational Fluid Dynamics (CFD), studying in English. If you have never studied CFD before, if you have never worked in the area, and if you have no real idea as to what the discipline is all about, then this book is for you. Although the material has been developed from first principles wherever possible, the book will be of greatest benefit to those who are familiar with the ideas of calculus, elementary vector and matrix algebra and basic numerical methods. The main purpose in writing this book is to provide a simple, satisfying, and motivational approach toward presenting the subject to the reader who is learning about CFD for the first time. In the workplace, CFD is today a mathematically sophisticated discipline. The book is focused on the problems related to aviation and aerospace topics. However, the proposed methods can be easily applied to a wider sphere of science.


2019 ◽  
Vol 7 (11) ◽  
pp. 384 ◽  
Author(s):  
Yingfei Zan ◽  
Ruinan Guo ◽  
Lihao Yuan ◽  
Zhaohui Wu

In underwater towing operations, the drag force and vertical offset angle of towropes are important considerations when choosing and setting up towing equipment. The aim of this paper is to study the variation in drag force, vertical offset angle, resistance, and attitude for towing operations with a view to optimizing these operations. An underwater experiment was conducted using a 1:8 scale physical model of a subsea module. A comprehensive series of viscous Computational Fluid Dynamics (CFD) simulations were carried out based on Reynolds-averaged Navier–Stokes equations for uniform velocity towing. The results of the simulation were compared with experimental data and showed good agreement. Numerical results of the vorticity field and streamlines at the towing speeds were presented to analyze the distribution of vortexes and flow patterns. The resistance components were analyzed based on the numerical result. It was found that the lateral direction was a better direction for towing operations because of the smaller drag force, resistance, and offset angle. Similar patterns and locations of streamlines and vortexes were present in both the longitudinal and lateral directions, the total resistance coefficient decreases at a Reynolds number greater than that of a cylinder.


2011 ◽  
Vol 64 (2) ◽  
Author(s):  
Giancarlo Alfonsi

The direct numerical simulation of turbulence (DNS) has become a method of outmost importance for the investigation of turbulence physics, and its relevance is constantly growing due to the increasing popularity of high-performance-computing techniques. In the present work, the DNS approach is discussed mainly with regard to turbulent shear flows of incompressible fluids with constant properties. A body of literature is reviewed, dealing with the numerical integration of the Navier-Stokes equations, results obtained from the simulations, and appropriate use of the numerical databases for a better understanding of turbulence physics. Overall, it appears that high-performance computing is the only way to advance in turbulence research through the front of the direct numerical simulation.


Water ◽  
2019 ◽  
Vol 11 (1) ◽  
pp. 82
Author(s):  
Getnet Kebede Demeke ◽  
Dereje Hailu Asfaw ◽  
Yilma Seleshi Shiferaw

Hydraulic structures are often complex and in many cases their designs require attention so that the flow behavior around hydraulic structures and their influence on the environment can be predicted accurately. Currently, more efficient computational fluid dynamics (CFD) codes can solve the Navier–Stokes equations in three-dimensions and free surface computation in a significantly improved manner. CFD has evolved into a powerful tool in simulating fluid flows. In addition, CFD with its advantages of lower cost and greater flexibility can reasonably predict the mean characteristics of flows such as velocity distributions, pressure distributions, and water surface profiles of complex problems in hydraulic engineering. In Ethiopia, Tendaho Dam Spillway was constructed recently, and one flood passed over the spillway. Although the flood was below the designed capacity, there was an overflow due to superelevation at the bend. Therefore, design of complex hydraulic structures using the state-of- art of 3D hydrodynamic modelling enhances the safety of the structures. 3D hydrodynamic modelling was used to verify the safety of the spillway using designed data and the result showed that the constructed hydraulic section is not safe unless it is modified.


2014 ◽  
Vol 2014 ◽  
pp. 1-12 ◽  
Author(s):  
Yong Liu ◽  
Jia Li ◽  
Yu Tian ◽  
Xia Yu ◽  
Jian Liu ◽  
...  

The application of two-phase computational fluid dynamics (CFD) for simulating crater-like Taylor cone formation dynamics in a viscous liquid is a challenging task. An interface coupled level set/volume-of-fluid (CLSVOF) method and the governing equations based on Navier-Stokes equations were employed to simulate the crater-like Taylor cone formation process. The computational results of the dynamics of crater-like Taylor cone slowly formed on a free liquid surface produced by a submerged nozzle in a viscous liquid were presented in this paper. Some experiments with different air pressures were carried out to evaluate the simulation results. The results from both CFD and experimental observations were compared and analyzed. The numerical results were consistent with the experimental results. Our study showed that the CLSVOF method gave convincing results, and the computational method is robust to extreme variations in interfacial topology.


Author(s):  
Zhang Lisheng ◽  
Jiang Jin ◽  
Xiao Zhihuai ◽  
Li Yanhui

In this paper numerical simulations were conducted to analyze the effects of design parameters and distribution of balancing-hole on the axial-force of a partial emission pump. The studied pump is a single stage pump with a Barske style impeller. Based on the original impeller, we designed 7 pumps with different balancing-hole diameters and the partial emission pump equipped with different impellers were simulated employing the commercial computational fluid dynamics (CFD) software Fluent 12.1 to solve the Navier-Stokes equations for three-dimensional steady flow. A sensitivity analysis of the numerical model was performed with the purpose of balancing the contradiction of numerical accuracy and the cost of calculation. The results showed that, with increasing of the capacity, the axial force varies little. The diameter of the inner balancing-hole plays a dominant role of reducing axial-force of partial emission pump, the axial-force decreases with increasing of inner balancing-hole diameter on the whole range of operation, the axial-force of impeller without inner balancing-hole is approximately 3 times larger than that of impeller with inner balancing-hole. While the diameter of outer balancing-hole has a reverse effects compared with that of inner balancing-hole. With increasing of outer balancing-hole, the axial force increases accordingly.


2017 ◽  
Vol 2 (6) ◽  
Author(s):  
Maciej Staszak

AbstractThe article presents the different approaches to polymer membrane mathematical modeling. Traditional models based on experimental physicochemical correlations and balance models are presented in the first part. Quantum and molecular mechanics models are presented as they are more popular for polymer membranes in fuel cells. The initial part is enclosed by neural network models which found their use for different types of processes in polymer membranes. The second part is devoted to models of fluid dynamics. The computational fluid dynamics technique can be divided into solving of Navier-Stokes equations and into Boltzmann lattice models. Both approaches are presented focusing on membrane processes.


Sign in / Sign up

Export Citation Format

Share Document