DEV Community: Precise Simulation

FEATool Multiphysics 1.18 with New FEA & CFD AI/ML Workflow

Precise Simulation — Sun, 01 Feb 2026 00:00:00 +0000

Precise Simulation is proud to announce the release of FEATool Multiphysics™ version 1.18, a step forward in the_state-of-the-art_ in Finite Element Analysis (FEA) and Computational Fluid Dynamics (CFD) multi-solver simulation software.

The latest release introduces a completely redesigned Graphical User Interface (GUI) making the toolbox more user friendly and_easy-to-use_ than ever before. FEATool Multiphysics also features full support and integration with MATLAB and related toolboxes (such as for Optimization, Control Systems, and_Machine Learning_), and advanced features tailored for the evolving needs of engineers and researchers in industry and academia.

New Standard for FEA, CFD & CAE AI/ML Simulation Workflows

One of the stand-out features of the FEATool toolbox is the Multiphysics Application Programming Interface (API), with one-click export functionality and automatic conversion of simulation models to MATLAB and Python script models. This enables users to quickly define and set up simulation models in a fully integrated and easy-to-use GUI, and later export, modify, and programmatically run them automatically for large scale parametric studies, and data generation and collection for Physics-Informed Neural Network (PINN), Machine Learning (ML) and artificial intelligence (AI) type simulation models.

An example of using FEATool Multiphysics to derive reference data as well as validation for machine learning CFD models can be found in several works on Deep Learning (DL) CFD methodology for flow prediction using with AI and machine learning by Prof. Thi-Thu-Huong Le and coworkers at the at the Blockchain Platform Research Center of the Pusan National University (PNU) in Korea.

Another example, here by Guodong Sa and coworkers, developed a Digital Twin (DT) framework for visualization and design of smart kitchens to facilitate improved kitchen and usability design. In the highlighted work, FEATool Multiphysics was used as a platform for programming and controlling simulation mesh and state-of-the-art CFD flow solvers such as OpenFOAM, and to script, automate, and programmatically generate thousands of sets of simulation data for training the digital twin framework.

And in the medical field, Prof. Zhang H. and coworkers have recently made use of FEATool Multiphysics to develop a method of denoising and improving vascular blood flow imaging for medical diagnosis using a Physics-Informed Neural Network (PINN). The data to train the PINN on was automatically generated by simulating the Navier-Stokes equations for different flow conditions and geometries. See their preprint on Fluid Dynamics and Domain Reconstruction from Noisy Flow Images Using Physics-Informed Neural Networks and Quasi-Conformal Mapping for more information.

For more detailed information on the new features and improvements, and to download the toolbox, please visit:

https://featool.com/news/2026/02/03/FEATool-Multiphysics-v1p18-Simulation-FEA-CFD-CAE-Workflow-Update/

FEATool Multiphysics minor update to version 1.17.5

Precise Simulation — Wed, 13 Aug 2025 04:08:17 +0000

FEATool Multiphysics minor update to version 1.17.5

In particular, significant improvements in graphics and UI performance for larger 3D models
Update for FEniCS external FEA and multiphysics solver
Improved OpenFOAM CFD solver API and documentation
New AC electrostatics plane capacitor script models

Please Visit https://www.featool.com for more information and to download and try the toolbox.

FEATool Multiphysics Update v1.17.4

Precise Simulation — Mon, 14 Jul 2025 04:02:00 +0000

FEATool Multiphysics has now been updated to version v1.17.4 including the following changes:

Update OpenCASCADE geometry engine to v7.9.1 (Windows + Linux)
Support for tabulated/interpolated equation/boundary coefficients
Heat conduction example and tutorial for interpolated nonlinear thermal conductivity
New setinf function (to set infinite values to specific number)
Troubleshooting documentation solver section
Minor bug fixes

See the announcement https://www.featool.com/news/2025/07/14/FEATool-v1p17p4-Multiphysics-Simulation-Toolbox-Update/ for more information.

Gmsh FEA & CFD mesh generation with MATLAB

Precise Simulation — Thu, 12 Jun 2025 01:35:58 +0000

Gmsh Mesh Generator

Gmsh is a very capable and cross platform stand-alone open source mesh generator for FEA and CFD applications. In addition to many built-in mesh generation algorithms, it also includes support for a variety of integrated mesh generators, such as Tetgen and Netgen, and also popular 2D and 3D CAD file formats such as STEP, IGES, and STL. As such is it one of the most flexible and popular mesh generators, and can be used with FEATool Multiphysics from the fully integrated GUI as well as MATLAB® scripting and programming API.

Some advantages of using Gmsh compared to the other grid generators is robustness and mesh generation speed (primarily for 3D geometries). Moreover, Gmsh also supports better and more control with a selection of different mesh generation algorithms, and specifying the grid size in different geometry regions, subdomains, as well as on boundaries, allowing for greater flexibility and better grids tuned for the specific problems and geometries. In addition, Gmsh also supports generating boundary layers and unstructured quadrilateral grids which can be useful for fluid dynamics simulations.

Gmsh with the FEATool GUI

Gmsh Installation

Before processing, first ensure that the latest FEATool Multiphysics version is installed (as a MATLAB toolbox Add-On if you wish to use Gmsh with MATLAB).

When using the FEATool Multiphysics toolbox, the Gmsh executable will be automatically downloaded and installed when required. If the Gmsh installation fails, please manually download and install Gmsh from the homepage 1.

Gmsh with MATLAB GUI

Once installed, the Grid Generation Settings dialog box will in 2D and 3D feature a Gmsh selection option from the Grid Generation Algorithm drop-down box. Moreover, the following options also apply to the Gmsh mesh generation algorithm

The Subdomain Grid Size, hmax, indicates the target grid cell diameter, and can either be a single scalar prescribing the grid size for the entire geometry, or a space separated string of numbers (array) where the hmax values correspond to the generated subdomains.

Boundary Grid Size, hmaxb, is analogous to hmax but related to boundaries (edges). hmaxb can consist of a single scalar applicable to all boundaries, for example

0.1

prescribing a mean cell edge length of 0.1 on every boundary. hmaxb can also be a numeric array with entries corresponding to individual boundaries, for example

0.1  0.2  0.3  0.4

specifying cell edge length 0.1 for boundary 1, 0.2 for boundary 2 etc.

The Smoothing parameter specifies the number of post grid smoothing steps to perform (default 3).

Selecting the Internal Boundaries check-box in 2D enables mesh generation of structured boundary layers on boundaries. The height of the boundary layer can be specified with the Layer Size and also Aspect Ratio between cells. With Gmsh it is also possible to specify the number of mesh Cells in the boundary layer, and the boundary numbers to apply boundary layer generation to.

In 2D one can also choose between Triangular and Quadrilateral Cell Types.

The Generate grid button effectively calls the gridgen command line function from the GUI, which in turn makes a system call to Gmsh, after which the generated grid is automatically imported and displayed.

MATLAB Gmsh Programming API

On the MATLAB command line the gridgen function is used to call Gmsh to generate unstructured 2D or 3D triangular grids. The following syntax is used

grid = gridgen( SIN, VARARGIN, 'gridgen', 'gmsh' )

where SIN is a valid FEATool fea model struct, geometry definition struct, or cell array of geometry objects. The gridgen function also accepts the following property/value pairs (VARARGIN).

Property    Value/{Default}            Description
-----------------------------------------------------------------------------------
hmax        scal/arr {0.1}             Target grid size for geometry/subdomains
hmaxb       scal/arr {[]}              Target grid size for boundaries
nsm         scalar   {3}               Number of grid smoothing steps
nref        scalar   {0}               Number of uniform grid refinements
algo2       scalar   {2}               2D mesh algorithm (1=MeshAdapt, 2=Automatic,
                                       5=Delaunay, 6=Frontal, 7=BAMG, 8=DelQuad)
algo3       scalar   {1}               3D mesh algorithm (1=Del, 2=New Del, 4=Front
                                       5=Front Del, 6=Front Hex, 7=MMG3D, 9=R-tree)
blayer      struct   {[]}              Data struct for Gmsh 2D boundary layers
quad        boolean  {false}           Use quad meshing (for 2D)
tol         scalar   {eps*1e3}         Deduplication tolerance
compound    boolean  {true}            Use Gmsh compound boundaries
mshopt      cell     {}                Cell array of Gmsh options
mshall      boolean  {true}            Output/save all meshed entities
mshver      integer  {2}               Gmsh msh file version (1/2/4)
nt          integer  {numcores/2}      Number of concurrent threads
verbosity   integer  {5}               Gmsh verbosity/output level
syscmd      string   {'default'}       Gmsh system call command
                                       (default 'gmsh fdir/fname.geo -')
fname       string   {'fea_gmsh_UID'}  Gmsh imp/exp file name (root)
fdir        string   {tempdir}         Directory to write help files
clean       boolean  {true}            Delete (clean) Gmsh help files
instdir     string   {userpath/.featool}  Gmsh binary installation directory
fid         scalar   {1}               File identifier for output ([]=no output)

Among the properties hmax indicates target grid cell diameters, and is either a numeric scalar valid for the entire geometry or an array with hmax values corresponding to the subdomains. hmaxb is similar to hmax but a numeric array with a hmaxb values corresponding to the boundaries (including internal boundaries).

NSM (default 3) the number of post smoothing steps to perform. NREF (default 0) the number of post uniform grid refinement steps. ALGO2 and ALGO3 the Gmsh 2D and 3D mesh generation algorithms. QUAD (default 0) toggles Blossom-Quad conversion for 2D geometries.

Additional Gmsh options can be provided with the cell array MSHOPT. For example MSHOPT could be given as

{{'Mesh', 'CharacteristicLengthMax', '1'}, {'Mesh', 'AnisoMax', '10'}}

More detailed information regarding the mesh generation options can be found in the documentation for Gmsh. Also, for more information about CLI usage access the function help by entering

>> help gridgen

in the MATLAB command line interface.

MATLAB Mesh Generation Examples

The following section presents a few examples to illustrate how one can generate FEA and CFD meshes on the MATLAB command line, and in m-file scripts with the programming API.

MATLAB Mesh Generation Gmsh Grid Examples

Unit square with uniform global grid size set to 0.1.

grid = gridgen( {gobj_rectangle()}, 'hmax', 0.1, 'gridgen', 'gmsh' );
plotgrid( grid ) % Plot and visualize mesh.
Unit square with fine grid along the top boundary.

grid = gridgen( {gobj_rectangle()}, 'hmax', 0.5, ...
'hmaxb', [0.5 0.5 0.01 0.5], 'gridgen', 'gmsh' );
plotgrid( grid )

Domain with curved boundaries meshed with quadrilaterals.

geom.objects = {gobj_rectangle(), gobj_circle([0 0],.6), gobj_circle([1 1],.3,'C2')};
geom = geom_apply_formula( geom, 'R1-C1-C2' );
grid = gridgen( geom, 'hmax', 0.1, 'quad', true, 'gridgen', 'gmsh' );
plotgrid( grid )

Two connected subdomains with refined grid along the shared boundary

geom.objects = { gobj_polygon([-2e-3 -8e-3;0 -8e-3;0 -6e-3;0 6e-3;0 8e-3;-2e-3 8e-3]), ...
                 gobj_polygon([0 -6e-3;2e-3 -5e-3;2e-3 4e-3;0 6e-3]) };
hmax  = 5e-4;
hmaxb = hmax*ones(1,4);
hmaxb(9) = hmax/5;
grid  = gridgen( geom, 'hmax', hmax, 'hmaxb', hmaxb, 'gridgen', 'gmsh' );
plotgrid( grid )

Composite component with several subdomains.

r1 = gobj_rectangle( 0, 0.11, 0, 0.12,  'R1' );
c1 = gobj_circle( [ 0.065 0 ],   0.015, 'C1' );
c2 = gobj_circle( [ 0.11 0.12 ], 0.035, 'C2' );
c3 = gobj_circle( [ 0 0.06 ],    0.025, 'C3' );
r2 = gobj_rectangle( 0.065, 0.16, 0.05, 0.07, 'R2' );
c4 = gobj_circle( [ 0.065 0.06 ], 0.01, 'C4' );
geom.objects = { r1 c1 c2 c3 r2 c4 };
geom = geom_apply_formula( geom, 'R1-C1-C2-C3' );
geom = geom_apply_formula( geom, 'R2+C4' );

grid  = gridgen( geom, 'hmax', [0.0025 0.05 0.0025], 'gridgen', 'gmsh' );
plotgrid( grid )

Complex geometry with several holes and subdomains.

w = 10e-4; L = 3*w; H = 5*w;
p1  = gobj_polygon( [w/10 0;(L-w/4)/2 0;(L-w/4)/2 H;0 H;0 H/3], 'P1' );
p2  = gobj_polygon( [(L+w/4)/2 0;L 0;L H-H/3;L H;(L+w/4)/2 H], 'P2' );
r1  = gobj_rectangle( (L-w/4)/2, (L+w/4)/2, 0, H, 'R1' );
c1  = gobj_circle( [2*w/3 3*w], w/3, 'C1' );
c2  = gobj_circle( [2*w/3 2*w], w/3, 'C2' );
c3  = gobj_circle( [2*w/3 1*w], w/3, 'C3' );
c4  = gobj_circle( [L-w/2 4.5*w], w/8, 'C4' );
c5  = gobj_circle( [L-w   4.5*w], w/8, 'C5' );
c6  = gobj_circle( [L-w/2 4*w], w/8, 'C6' );
c7  = gobj_circle( [L-w   4*w], w/8, 'C7' );
c8  = gobj_circle( [L-w/2 3.5*w], w/8, 'C8' );
c9  = gobj_circle( [L-w   3.5*w], w/8, 'C9' );
c10 = gobj_circle( [L-w/2 3*w], w/8, 'C10' );
c11 = gobj_circle( [L-w   3*w], w/8, 'C11' );

geom.objects = { p1 p2 r1 c1 c2 c3 c4 c5 c6 c7 c8 c9 c10 c11 };
geom = geom_apply_formula( geom, 'P1-C1-C2-C3' );
geom = geom_apply_formula( geom, 'P2-C4-C5-C6-C7-C8-C9-C10-C11' );

hmaxb = zeros(1,21);
hmaxb([6 20]) = w/50;
grid = gridgen( geom, 'hmax', w./[5 20 5], 'hmaxb', hmaxb, 'gridgen', 'gmsh' );
plotgrid( grid )

Notes

For geometries with multiple and overlapping geometry objects the actual subdomain numbering will generally not correspond to the geometry object numbering (two intersecting geometry objects will for example create three or more subdomains and several internal boundaries). In this case the actual subdomain and boundary numbering for vector valued hmax and hmaxb arrays can easiest be visualized and determined by first creating a coarse grid and switching to Equation/Subdomain and Boundary modes, respectively.

Gmsh propagates the hmax and hmaxb values down to the specific nodes in the mesh which means that it is currently not possible to exactly define mesh sizes for subdomains and boundaries.

Gmsh also generates temporary geo and msh data files which, if required, can be found in the directory specified by the fdir parameter (default given by the MATLAB tempdir command).

Please visit https://www.featool.com for more information and toolbox download.

References

[1] Gmsh - A three-dimensional finite element mesh generator with built-in pre- and post-processing facilities.

[2] Geuzaine C., Remacle J.-F. Gmsh: a 3D finite element mesh generator with built-in pre- and post-processing facilities, International Journal for Numerical Methods in Engineering 79 (11), pp. 1309-1331, 2009.

Building FEA simulation tools using FEATool Multiphysics MATLAB API

Precise Simulation — Tue, 10 Jun 2025 09:18:00 +0000

ADRANOS (short for Advanced Divertor paRametric Analysis for coolaNt Operating Scenarios) is a specialized MATLAB based simulation tool, developer by researchers at the Department of Engineering of the University of Palermo, to assess steady-state performance diverter cooling circuits in fusion reactor applications.
The specialized tool was developed as full scale 3D Computational Fluid Dynamics (CFD) simulation have shown not to be feasible, as a large number of parametric configurations and simulations are required to be performed in order to analyze and optimize diverter performance and design.

ADRANOS simulation framework using FEATool Multiphysics for heat transfer FEM simulationsThe ADRANOS simulation framework consists of two modules that work together:

Lumped-Parameter Module - This module looks at the steady-state of the entire cooling circuit, breaking it down into major sections or volumes to quickly figure out overall flow rates, average coolant temperatures, and pressure drops across the sections.
2D FEM Thermal Analyses Module - This module employs the FEATool Multiphysics toolbox to perform thermal simulations of the most crucial parts of the divertor, specifically the Plasma Facing Unit (PFU) monoblocks.

The 2D FEM module is a crucial part of ADRANOS and works in conjunction with the lumped-parameters module. The integration of the FEM solver allows ADRANOS to assess detailed temperature distributions and check for compliance with material temperature limits, especially for critical PFU materials like tungsten, copper interlayer, and alloys in the pipes.
In particular, as FEATool Multiphysics features a native MATLAB Application Programming Interface (API) making it easy to program and set up custom parametric simulations runs varying geometry, mesh, and physics definitions to programmatically generate parallel sets of simulation data. Moreover, the easy to use API also allowed for integrating and embedding the FEATool FEM simulations in the overarching ADRANOS framework.

The ADRANOS code was developed in the MATLAB programming language with an object-oriented (OOP) approach, to make it highly flexible in evaluating different cooling system topologies. Moreover, it is also optimized for parallel High Performance Computing (HPC), in order to greatly reduce overall time required to perform complex simulations, and enable quick and automatic execution of thousands of simulations under various coolant operating conditions.

See the linked references for more detailed information about this research with data available from the authors upon request.

References:

Quartararo A., Di Maio P.A., Vallone E. ADRANOS: A numerical tool developed to analyse coolant operating conditions of the EU-DEMO divertor, Fusion Engineering and Design 197, 114055, 2023, doi: 0.1016/j.fusengdes.2023.114055.

CFD simulations for a real-time smart home digital twin (DT) framework

Precise Simulation — Sun, 02 Mar 2025 00:00:00 +0000

multiscale simulation through multiple sensors collaborative sensing of physical reality. In the context of a smart home, DT technology uses sensors, internet of things (IoT) devices, deep learning (DL), and artificial intelligence (AI) algorithms to collect and analyze data in real-time, creating a virtual model of the home environment.

Digital twin frameworks typically integrate data processing, flow field simulation, equipment monitoring, interactive elements, and visualization techniques to provide real-time insights into operations. For a smart kitchen, DT offers significant benefits for designers, enabling them to design appropriate kitchen appliances, improve development efficiency, and make more accurate predictions. For users, the DT offers valuable insights into the current state and dynamic evolution of the kitchen environment, thereby elevating the overall user experience to a higher level.

Smart home geometry (left) and simplified FEATool Multiphysics CFD simulation for digital twin (right)

One of the most computationally expensive components of the proposed digital twin framework are computational fluid dynamics (CFD) flow simulations. In order to overcome this limitation, the authors implemented an online and cloud based residual-based Fourier neural operator (RFNO) simulation method to achieve accurate results faster.

RFNO works by encoding input data, such as initial conditions, boundary conditions, and physical parameters, along with geometry and location information. The input is then processed through multiple residual modules, each containing Fourier layers. Within these layers, RFNO applies a Fourier transform to extract flow field information, filters high-frequency modes using a linear transform. Given new inputs, an inverse Fourier transform can then be used to convert the data back without performing costly simulations. The proposed RFNO method could be shown to be 1000 times faster than traditional CFD simulations on a flow around a cylinder benchmark.

In the presented work, the native MATLAB and Python programming API available with FEATool Multiphysics, as well as easy to use OpenFOAM GUI for mesh and case file generator, made it a uniquely suitable tool to script, automate, and programmatically generate thousands of sets of simulation data for training the RFNO digital twin framework.

See the linked references for more detailed information about this research.

References:

Sa G., Wu C., Liu Z., et al. A visual digital twin framework based on residual-based fourier neural operator online simulation method, Journal of Simulation, 1–21, 2024, doi: 10.1080/17477778.2024.2394063.

Originally published at https://www.featool.com.

Design of Transmission Tubes for Surgical Concentric Push-Pull Robots

Precise Simulation — Sun, 16 Feb 2025 00:00:00 +0000

This paper explores the design of “concentric push-pull robots” designed to make endoscopic surgery easier, such as biopsies or to removing polyps. In contrast to using colonoscopes which are difficult to maneuver, and may lead to longer procedure times and a higher chance of needing additional more invasive procedures, the authors designed concentric push-pull robots, which consist of two thin tubes, one inside the other, that are made of a material that can bend. The tubes are moved by pushing and pulling them from the end outside the body, which causes the tubes to bend in specific ways allowing for more precise control of the instruments inside the body.

The authors used the FEATool Multiphysics finite element (FEA) simulation toolbox to create 3D models of the tubes and analyze the stiffness while simulating how the tube would bend and twist in response to different forces. These simulations were then used to calculate the bending and torsional stiffness of the tube. They also used FEATool to investigate how the shape of the slots affected the stiffness properties. stiffness. Finally, they used simulation to optimize the design of the slots to achieve the desired balance of bending and torsional stiffness.

See the linked references for more detailed information about this research.

References:

Dang K.T., Qiu S., et al. Design of Transmission Tubes for Surgical Concentric Push-Pull Robots, International Symposium on Medical Robotics (ISMR), 2024, doi: 10.1109/ISMR63436.2024.10585572.

FEATool Multiphysics v1.17.3 Update for MATLAB 2025a

Precise Simulation — Fri, 14 Feb 2025 09:32:47 +0000

FEATool #Multiphysics FEA & #CFD simulation toolbox update v1.17.3 now available.

Including compatibility fixes for #MATLAB 2025a runtime. However, using MATLAB 2025a or later is NOT Recommended as the old Java GUI has been deprecated, while the new Web UI generally does not perform well and has some UI incompatibilities.

Visit https://www.featool.com/ for more information and toolbox download.

Thrombosis risk assesment using CFD simulation from PIV data

Precise Simulation — Thu, 02 Jan 2025 00:00:00 +0000

The following research, conducted by Borowski F., Kaule S., and colleagues concerns a type of heart valve that can be inserted without open-heart surgery (transcatheter aortic valve replacement TAVR). The researchers wanted to see if the way the valve is inserted affects how well blood flows through it, specifically the risk of thrombosis, or blood clot formation, due to misalignment of the artificial valve. To perform the study an experimental artificial model of the human heart was created after which they then inserted the valve in two different ways, measuring the flow field with particle image velocimetry (PIV).

To evaluate the risk of thrombosis two fluid mechanic predictors where chosen, shear induced effect and residence time, which are commonly associated with the thrombosis formation. These two transport equations were calculated and solved using the numerical solvers provided by FEATool Multiphysics. Geometry and time dependent boundary conditions in the fluid domain could be implemented using the measured PIV velocity fields, using custom MATLAB functions in FEATool, after which a transient numerical simulation was performed to solve the transport equations.

The results of the study show that the circumferential positioning of TAVR has an impact on the local flow field. The authors conclude that the implantation strategy of a TAVR should not only focus on implantation height with respect to regurgitation but also on the alignment of native and prosthetic commissures.

FEATool’s easy to use GUI, and integration with a MATLAB programming and scripting API allowed for easy manipulation and analysis of the imported PIV data, streamlining the process of CFD simulation. Once FEATool completed the simulations, the researchers extracted the velocity and pressure data at each grid point. This data was then preprocessed and formatted and analyzed.

See the linked references for more detailed information about this research.

References

Borowski F., Kaule S., et al. Analysis of thrombosis risk of commissural misaligned transcatheter aortic valve prostheses using particle image velocimetry, Technisches Messen, vol. 91, no. 6, pp. 291-304, 2024, doi: 10.1515/teme-2022-0100.

AI & Deep Learning for CFD Flow Prediction

Precise Simulation — Wed, 20 Nov 2024 00:00:00 +0000

In a series of publications Prof. Thi-Thu-Huong Le, Hoyeun Kang, and
colleagues from Pusan National University (PNU) in Korea, have
successfully developed a new deep learning (DL) CFD methodology for
flow prediction using FEATool Multiphysics with AI and machine learning.

Although traditional Computational Fluid Dynamics (CFD) solvers are
becoming increasingly common tool in aerospace, automotive, and other
engineering fields, while effective, face limitations, primarily
related to high computational cost. Especially when high spatial and
temporal resolution is needed for accurate simulation, traditional CFD
solvers become computationally expensive, and can sometimes require
days or even weeks to perform accurate analysis.

Deep learning have potential to offer an alternative approach that may
significantly reduce computational cost, creating an AI CFD model on
an accurate fluid dynamics dataset. The autors propose such a new CFD
model, named CFDformer, which combines a Vision Transformer (ViT)
and a U-shaped Convolutional Neural Network (U-Net) in an
encoder-decoder architecture to predict fluid flow on 2D geometries.

A DL-CFD model agent can offer a number of advantages compared to
traditional CFD simulations

Act as Surrogate Models - Deep learning models, trained on data
from CFD simulations, can act as surrogate models, effectively
approximating the solutions to the Navier-Stokes equations. This
approach bypasses the need for computationally intensive iterative
calculations.
Speed Up Simulations - Deep learning models can significantly
reduce simulation times. For example, CFDformer, a hybrid model
combining a Vision Transformer and a U-Net, was shown to decrease
analysis time by up to 99.94% compared to standard CFD solvers.
Complex Flow Scenarios - Deep learning models can handle
complex flow scenarios with relatively high accuracy, and are
particularly well-suited for modeling flows around obstacles, such
as found in aerodynamics applications.
Adapt to New Conditions - Some deep learning models can generalize
well to new, unseen conditions not encountered during training. This
allows them to make reasonable predictions for conditions within a
specific range, even without explicit training data for those
conditions.
Enhancing Feature Extraction - Deep learning models can be
designed to extract both local and global features from input data,
improving the accuracy of flow approximations. For instance,
CFDformer uses convolutional layers to capture local spatial
features and a Vision Transformer to analyze global flow features.

The FEATool Multiphysics toolbox was used to generate datasets of 2D
incompressible and laminar flows around various obstacles, including
cylinders, triangles, rectangles, and pentagons. In particular, as
FEATool supports multiple CFD solvers,
such as OpenFOAM, using multiple solvers is ideal way to generate
accurate and trustworthy, benchmark and validation CFD studies and
datasets. These datasets served as the ground truth for training and
evaluating deep learning models.

FEATool's easy to use GUI, and integration with a MATLAB programming
and scripting API allowed for easy manipulation and analysis of the
simulation data, streamlining the process of preparing the data for
deep learning. Once FEATool completed the simulations, the researchers
extracted the velocity and pressure data at each grid point. This data
was then preprocessed and formatted as input for training deep
learning models.

Please visit https://www.featool.com for more information and toolbox download.

References

Kang H., et al. A new fluid flow approximation method using a vision transformer and a U-shaped convolutional neural network, AIP Advances, Volume 13, Issue 2, 2023, doi: 10.1063/5.0138515.
Prihatno A.T., et al. 2D Fluid Flows Prediction Based on U-Net Architecture, Proceedings of International Conference on Artificial Intelligence in Information and Communication (ICAIIC), 2023, doi: 10.1109/ICAIIC57133.2023.10066980.
Le T.T.H, Kang H. et al. CFD Prediction of Indoor Airflow using Deep Learning, Conference paper, 2022.
Le T.T.H., Kang H., Kim H. Towards Incompressible Laminar Flow Estimation Based on Interpolated Feature Generation and Deep Learning, Sustainability, 14, 11996, 2022, doi: 10.3390/su141911996.
Kang H. CFDformer: Novel Fluid Flow Approximation based on ViT and U-Net, GitHub repository for CFDformer MATLAB dataset generation, doi: 10.5281/zenodo.7527624, 2023.

CAD File Import and Mesh Generation with Gmsh

Precise Simulation — Wed, 13 Nov 2024 11:43:32 +0000

Gmsh is a popular cross platform and open source mesh generation software which supports many CAD file formats such as STEP, IGES, and STL. In addition to its own mesh generation algorithms, Gmsh also includes support for a variety of integrated mesh generators, such as Netgen, Tetgen, and Triangle. As such is it one of the most flexible mesh generators, and can be used with FEATool Multiphysics to allow meshing complex geometries and models than the default built-in grid generator.

Although FEATool includes built-in support for import and automatic meshing of CAD geometries, controlling the mesh generation steps manually offer more control and can result in higher quality meshes, and more accurate simulation results. Manual meshing can for instance be useful for more difficult 3D geometries, which may require repair or defeaturing in order to allow for creation of good meshes for simulation.

The following section explains how to use Gmsh to manually load and import a CAD geometry from a STEP file, and then generate a mesh which can be imported and used with FEATool Multiphysics.

Gmsh CAD File Import and Mesh Generation Tutorial

This is a tutorial showing how to import a STEP model into Gmsh, and
generate a 3D volume mesh for finite element analysis (FEA)
simulations. The STEP file and CAD model used in the example can be
downloaded from the link below (but any STEP model should technically
work)

Fixed Spanner CAD STEP Model

1. Start Gmsh (using version 2 Mesh File Format)

If the resulting mesh is to be used with the FEATool Multiphysics toolbox (or the Gmsh MATLAB API), it is important that the Gmsh mesh file is saved in Gmsh ASCII mesh file format version 2. This format is default in Gmsh versions 2-3, but not in the current Gmsh versions 4.x.

To make sure Gmsh saves and exports meshes in version 2 format, one can either start Gmsh with the command line argument -format msh2, for example

gmsh.exe -format msh2

or set the following option in the general or mesh specific Gmsh .opt options file

Mesh.MshFileVersion = 2;

The option files can be generated by selecting Save Model Options or Save Options As Default in the File menu of the Gmsh GUI.

If there are issues importing the mesh into FEATool one can also optionally try to add the -save_all command line argument which corresponds to the Mesh.SaveAll = 1; .opt file option.

2. CAD Model Import with the Merge Operation

The second step is to use the Merge… option, from the File menu in the Gmsh menubar. Choosing this menu option opens a file selection dialog box where the CAD file to import can be chosen. In this example, select the spanner.step file and press the Open button to automatically let Gmsh load, import, and construct edges and faces for the geometry. If the model has been imported successfully it should be displayed and shown in the main GUI window of Gmsh (you can use mouse controls to rotate and zoom in/out).

3. Setting Mesh Options and Specifying Grid Size

Gmsh supports several options to specify mesh and grid sizes. One can for example prescribe specific sizes at various points, or define functions to determine variable mesh sizing. Here the Min/Max element size option is used to set a global maximum mesh size of 6 mm (as the units in this CAD model is defined in millimeters). This option can be found on the General tab, in the Mesh section of the Options dialog box (which is accessed by selecting Options from the Tools menu).

A description of the different mesh options is available in the Mesh options section of the Gmsh reference guide.

4. 3D Mesh and Grid Generation

To generate a 3D volume mesh, first click on the [+] Mesh node in the left hand side tree list to expand it, then click to the 3D node to start and run the selected mesh generation algorithm. Once the mesh generation algorithm has finished the resulting mesh should be displayed in the Gmsh GUI window.

5. Mesh Info, Saving, and Mesh Export

Progress information and messages (and errors) are displayed in the bottom line of the Gmsh UI. Clicking on this line also expands and opens the message and log window, where one can see information and statistics from the mesh generation process.

Once a mesh has been generated it can be saved and exported by using the Save Mesh option from the File menu (the file will be saved in the same directory and named as the imported CAD file but with the .msh file extension).

Gmsh Import and Export in FEATool Multiphysics

To import and use a Gmsh .msh file with FEATool Multiphysics use the Import Grid > Gmsh Format… option found under the Grid mode menu of the toolbox GUI (note that the mesh file has to have the same 2D or 3D space dimension as the model, and be in Gmsh version 2 ASCII format to be compatible with the toolbox).

An illustrative example on how to import the spanner CAD model, define, set up, and run simulations in FEATool Multiphysics is explained in the linked tutorial for a structural mechanics simulation with a parametric displacement study.

MATLAB Gmsh Import and Export Functionality

Alternatively, the impexp_gmsh function, from the MATLAB programming and scripting API, can be used directly on the command line to import a Gmsh mesh directly into MATLAB. The following commands imports, exports, and visualizes the mesh in MATLAB

input_file = 'C:\temp\spanner.msh';   % Full path to input mesh file.
mesh = impexp_gmsh( input_file, 'import' );

plotgrid( mesh )   % Plot and visualize mesh

plotsubd( mesh )   % Plot and visualize subdomains

plotbdr( mesh )    % Plot and visualize boundaries

output_file = 'C:\temp\output.msh';   % Full path to output mesh file.
impexp_gmsh( output_file, 'export', mesh );

Toolbox Availability and Download

The FEATool Multiphysics™ toolbox is available as stand-alone
Desktop App, and also as MATLAB toolbox Add-Ons, with fully interactive GUI and cross-platform support for the Microsoft Windows, Linux, and MacOS operating systems. The toolboxes can be downloaded directly from the homepage https://www.featool.com (or installed with one-click from the MATLAB Add-Ons Toolbar).

Multi-body System Simulation with FEATool Multiphysics for Design of Optical Grating Tiling Devices

Precise Simulation — Tue, 12 Nov 2024 13:19:52 +0000

A grating tiling device is an optical element which can separate light
into different wavelength components. Such devices are for example
used in spectroscopy, and in optical laser devices for fusion
confinement. Due to the high sensitivity and precision required in
these devices, stability with respect to vibrations and temperature is
of the most important aspects in their design.

Modeling and design of grating devices using traditional finite
element analysis (FEA) require increasing computational effort, and is
often limited by available computing capabilities. In the last two
decades, multi-body system (MBS) simulations have instead become the
most effective tool in modeling and analysis of such complex
mechanical systems.

Research presented in the paper “Development of dynamics for design
procedure of novel grating tiling device with experimental
validation” make use of the FEATool Multiphysics toolbox, and its
native MATLAB® FEA API, to develop a new efficient process for
design of tiled grating devices. The authors suggest and explore an
alternative MBS procedure in order to obtain an efficient design cycle
(labeled dynamics for design DFD), by integrating the dynamics of
interconnected systems, including nonlinearities, vibration analysis,
and parameter estimation, with current design methodologies.

Specifically, a multi-body dynamics approach was used to calculate
reaction forces acting on the flexible and flexure bodies of the
grating tiling device. This approach defined a parametric feedback
loop in MATLAB, where forces where used as input to a FEATool
Multiphysics stress-strain structural mechanics script model in each
iteration of the loop.

The optimum design could be achieved by changing materials in the
flexible and flexure bodies and system dimensions and iterating until
reaching the maximum range of grating movement parameter. As seen in
the diagram below the FEATool simulation software was used as a
central key component in the simulation loop made possible due to its
MATLAB scripting and programming API, and built-in translation of GUI
models to MATLAB simulation scripts.

The authors concluded that the dynamics for design (DFD) procedure,
which includes the use of the FEATool Multiphysics toolbox, was a very
effective way to design a grating tiling device which can achieve the
required performance without breaking down.

References

Bai Q., et al. Development of Dynamics for Design Procedure of Novel Grating Tiling Device with Experimental Validation, Journal of Applied Sciences, 11(24), 11716, 2021, doi: 10.3390/app112411716.