feelpp / toolbox Goto Github PK

Is your feature request related to a problem ? Please describe.
Lots of industrial problems require loading data from datafiles and use interpolation to fit the data.

Describe the solution you'd like
use fit interpolation feature of feel++ to add json files.

Additional context
request comes from a ECMI 2018 conference participant

Is your feature request related to a problem ? Please describe.
Would it be possible to have C++ or Python codes as inputs for Feel++ apps in MSO4SC

Describe the solution you'd like
Instead of case option we would use a script/code option which would then be executed by the feel++ app

Describe alternatives you've considered

Additional context
This was discussed with an ECMI 2018 participant

We want to develop a toolbox for fluid seepage through porous media in 2D and 3D using the work already done for the Poisson problems.

One of the aims is to be able to solve Stokes-Darcy and Navier-Stokes-Darcy coupled problems within the context of blood perfusion through the brain.

• Documentation of the model
• Documentation of the mathematical background
• Documentation of the toolbox

The toolbox should be a subclass of the HDG Poisson.

Description of the example

this is related to feelpp/feelpp#1457

add the documentation in the toolbox repo

• #135
• [ ]

Is your feature request related to a problem ? Please describe.
For many problems, it may be useful in a first approximation to consider 2D Axi geometries and formulation. Common examples may be found in physics textbooks.

Describe the solution you'd like
It would be great to have a 2D axi version for each toolbox.

Additional context
Special care shall be taken for PDE with vectorial unknows.

The opus testcase needs to be added to the toolbox benchmarks, we have some results to compare with in some cases (with Comsol)

We would like to enable coupling between a free fluid model and a fluid flow through porous media using existing CFD and soon existing HDG Darcy toolboxes.

We gather at ECMI 2018 user stories to improve our Feel++ apps and pilots as well as the userexperience

This is linked to #16 to provide examples for the Darcy toolbox based on the generic HDG Poisson solver. See #16 for more details.

We should gather here suggestions of examples/test cases of Darcy flows.

Is your feature request related to a problem ? Please describe.
thermo-electric-fluid toolbox would be nice to have in many industrials applications

Additional context
Request from a ECMI participant

SZE have created various testcases/examples to use Feel++ CFD and CSM toolboxes. This needs to be added to Feel++.

add thermal building example

/cc @vincentchabannes

Description of the example

In this example, we consider a solenoid conductor with finite thickness and infinite length.
This allow us to ignore the z components in our equations.
We admit that there is only a radial expansion.

Geometry

The conductor $\Omega$ consists in a rectangular cross section torus.
The geometry also contains an external domain which is an approximation of $\mathbf{R}^3/\Omega$.

Input parameters

current density: $\textbf{j}$ in $A/m^2$.
magnetic field: $\textbf{b}$ in $T$.

Model & Toolbox

• From the momentum conservation equation, we have:
  [
  div\sigma+\textbf{j}\times\textbf{b}=0
  ]
  With the hypothesys specified in introduction, this equation may be rewritten :
  [
  -\sigma_{\theta}+\frac{\partial}{\partial r}(r\sigma_{r})=-rj_{\theta}b_{z}
  ]
  For a solenoid conductor with finite thickness and infinite length, we have for a constant current density
  $j_{\theta}$:
  [

• b_{z} = ...
  ]
  With these hypothesys, we can show that (see [REF002]):
  [

• u_{r} = ...
  ]

• toolbox: elasticity

Materials

|Name |Description | Value | Unit |
|$E$ |Young modulus||$128.10^{9}$|$Pa=kg.m^{-1} .s^{-2}$|
|$\nu$|Poisson's ratio|0.33|- |

Boundary conditions

• entry/exit: $\textbf{u} \dot \texbf{n} = 0$, displacements only in the perpendical plane,
• top:bottom: clamped

Outputs

The output is describe the output set of the example

Fields

add scalar vectorial and matricial fields to be visualized

Measures

add measures, scalar quantities, mean values, performance metrics

Benchmark

Describe Benchmark type:
[X] Verification
[] Validation
[] Performance

The computed values of the displacements on r-axis in the solenoidal mid-plane are compared
with the analytical expression given by Montgomery.

References (articles, papers, reports...)

• https://github.com/feelpp/hifimagnet/blob/develop/MagnetTools/docs/Stress/stress.tex
• [REF001] Montgomery, Solenoid Magnet Design, Wiley-Interscience, New-York. 1969.
• [REF002] M N. Wilson, Superconducting Magnets, Oxford University Press, London, 1987.

we have examples and benchmarks, I am thinking that it might be best to actually merge them into cases or examples and have Feel++ Cases Guide. Then we would need to identify the benchmarks (which are just a subset of a case or example, they provide verification,validation or scalability checks) in each modules in a specific section. That way we gather all examples in one place and don’t have to remember whether it is a benchmark or an example

Another side effect is that we show more examples in one go for each toolbox (edited)
Yet another advantage is that it simplifies the documentation and avoid duplication between example and benchmark templates

@vincentchabannes what do you think ?

• cfpdes update
• electric update
• multifluid update
• maxwell update
• ROOT update

Description of the example

simulate a Stokes flow in a pipe with a poiseuille inflow

Geometry

the geometry is a 2D pipe with length L=5 and height H=1

Input parameters

The input parameters can be the

• viscosity
• length
• height
• velocity magnitude

Model & Toolbox

• toolbox fluid CFD using the Stokes

Materials

constant set to eg. blood viscosity

Boundary conditions

First case

• inlet : poiseuille (dirichlet)
• outlet : free (neumann)
• wall : no slip condition (homogeneous dirichlet)

Outputs

Fields

velocity and pressure

Measures

l2 norm of error with respect to exact solution

Benchmark

this is a Verification benchmark

/cc @vincentchabannes

Is your feature request related to a problem ? Please describe.
When a end-user is reading the documentation, he may want to execute the case or have access to the case string to pass to MSO4SC

Describe the solution you'd like
To facilitate the usage, we can provide a copy to clipboard button that allows to retrieve either the command line or the case option

we may use https://clipboardjs.com/ and add it to supplemental-ui. it is small and seems to fit the bill.

Also we need to

• update the template case file and add this feature
• update current cases

Description of the example

In this example, we will estimate the rise in temperature due to Joules losses in a stranded conductor. An electrical potential $V_0$ is applied to the entry/exit of the conductor which is also cooled by a force flow.
The geometry of the conductor is choosen as to have an analytical expression for the temperature.

Geometry

• The conductor consists in a rectangular cross section torus which is somehow "cut" to allow for applying electrical potential. The conductor is cooled with a force flow along its cylindrical faces.
• /images/learning/thermoelectric/quarter-turn3D.png
• upload CAD file if available

Input parameters

Model & Toolbox

• This problem is fully described by a ThermoElectricModel, namely a poisson equation for the electrical potential $V$ and a standard heat equation for the temperature field $T$ with Joules losses as a source term.
  Assuming that the physical properties of the materials forming the conductor are not dependent on $T$, we can show that $(V,T)$ are solutions of:
  [
  V = V0*\frac{\theta}{2\pi}
  \rho C_{p}\frac{\partial T}{\partial t} - \nabla.(k \nabla T)=\sigma(\frac{V}{2\pi r})^{2}
  ]

In our geometry and owing to the considered thermal boundary conditions, $T$ only depends on the radius $r$.
We can show that the $T$ static solution is of the form:
[
T=-a \log(\frac{r}{r_{0}})^{2} + T_{max}
]
with:

• $a=sigma/(2k)(V0/(2*pi))**2$

• $b=k*(1/(h1r1)+1/(h2r2))+log(r2/r1)$

• $c=log(r2/r1)log(r2r1)+2k(log(r1)/(h1r1)+log(r2)/(h2r2))$

• $r0=exp( ((Tw2-Tw1)/b+ac/b)/(2a) )$

• [Tm = 2ak/(h1r1+h2r2)log(r2/r1)
  + (h1r1Tw1+h2r2Tw2)/(h1r1+h2r2)
  + a(h1r1log(r1/r0)**2+h2r2log(r2/r0)**2)/(h1r1+h2r2)
  ]

• toolbox: thermoelectric

Materials

Boundary conditions

The boundary conditions for the electrical probleme are introduced as simple Dirichlet boundary conditions for the electric potential on the entry/exit of the conductor. For the remaining faces, as no current is flowing througth these faces, we add Homogeneous Neumann conditions.

where $f$ is a factor representing the portion of the torus considered (eg f=0.25 is only 1/4th of the torus is modeled, f=1 for the complete torus).

As for the heat equation, the forced water cooling is modeled by robin boundary condition with $Tw$ the temperature of the coolant and $h$ an heat exchange coefficient.

Outputs

The output is describe the output set of the example

Fields

add scalar vectorial and matricial fields to be visualized

Measures

add measures, scalar quantities, mean values, performance metrics

Benchmark

Describe Benchmark type:
[X] Verification
[] Validation
[] Performance

The computed values of the temperature are compared with the analytical expression given above for several mesh sizes.

References (articles, papers, reports...)

• add any article in pdf or html links related to the example
• [REF001] authors..., title, ... journal,... year...

this is in angiotk repo

/cc @vincentchabannes

Magnetic Field on the Z-Axis of a Cross rectangular section solenoidal magnet

In this example, we will compute the magnetic field generated by a stranded conductor.
The geometry of the conductor is chosen such that we can derive the analytical expression
of the magnetic field.

Geometry

The conductor stem:[\Omega] consists in a rectangular cross-section torus.
The geometry also contains an external domain which is an approximation of stem:[\mathbf{R}^3/\Omega].

Input parameters

current density: stem:[\mathtbf{j}] in stem:[A/m^2].

Model & Toolbox

• The magnetostatic formulation in terms of magnetic potential stem:[\mathtbf{A}] may be found in xxx:
  [stem]
  ++++
  \nabla^{2}\mathbf{A}=-\mu\mathbf{j}
  ++++
  with stem:[\mathbf{A}] the magnetic potential and stem:[\mathbf{j}] the current density in $\Omega$.
  The general solution of this equation is :
  [stem]
  ++++
  \mathbf{A}(\mathbff{r})=-\mu\int_{\Omega}G(\mathtbf{r},\mathbf{r'})\mathbf{j}(\mathbf{r'})
  ++++
  with stem:[G(\mathbf{r},\mathbf{r'})] the Green's function defined as :
  [stem]
  ++++
  G(\mathbf{r},\mathbf{r'})=\frac{1}{4\pi}\frac{1}{|\mathbf{r}-\mathbf{r'}|}
  ++++
  with stem:[\mathtbf{r}\in\mathbf{R}^3] andstem: [\mathbf{r'}\in\Omega_{cond}]

Then it follows that the magnetic field stem:[\mathbf{B}], defined as the curl of stem:[\mathtbf{A}], is:
[stem]
++++
\mathbf{B}(\mathbf{r})=\frac{\mu_{0}}{4\pi}\int_{\Omega_{cond}}\frac{\mathbf{j}(\mathtbf{r'})\times(\mathbf{r}-\mathbf{r'})}{|\mathbf{r}-\mathbf{r'}|^{3}} d\mathbf{r'}\qquad\forall\mathbf{r}\in\Omega_{mgn}
++++
This is the so-called Biot & Savart's law.

For simple conductor geometry, analytical expressions of the magnetic field may be found in physics textbooks.
As a classical result, we have the following expression for a "solenoidal" conductor with a constant current density
distribution for the magnetic field along the Z-Axis:
xxxxx

• toolbox: maxwell

Boundary conditions

• stem:[\mathbf{A} = \mathbf{0}] on the boundaries
  // - stem:[\mathbf{A} \times mathbf n = \mathbf{0}]
• stem:[\mathbf{A} \dot \mathbf{n} = 0 ] on symetry plane

Outputs

The output is described the output set of the example

Fields

add scalar vectorial and matricial fields to be visualized

Measures

add measures, scalar quantities, mean values, performance metrics

Benchmark

Describe Benchmark type:
[X] Verification
[] Validation
[] Performance

The computed values of the magnetic field on the Z-Axis is compared with the analytical expression given above
for several mesh sizes.

References (articles, papers, reports...)

• add any article in pdf or html links related to the example
• [REF001] authors..., title, ... journal,... year...

the example is in angiotk repo

This case fails using

feelpp_toolbox_fluid --case "github:{repo:toolbox, branch:master, path:examples/modules/cfd/examples/spring_cfd}"

The mesh is missing (it should be downloaded from girder).

Is your feature request related to a problem ? Please describe.
In Hifimagnet the field of concern is more the current density (conductivity times electric field) than the electric field itself.
It would be great to be able to export it.

Describe the solution you'd like
Add current-density to the list of exported fields.

Describe alternatives you've considered
Since the conductivity and the electric field can be exported, we could compute ourselves the current density.

Additional context
I started implementing it from the same model than the electric field, with a member to store it and a function updateCurrentDensity().
For the linear case, there is no issue, but in the non linear case, the conductivity depending on the temperature, it is the thermoelectric toolbox that should handle this.

Possible fix:

• template the updateCurrentDensity function and pass the temperature as a parameter
• the electric::updateCurrentDensity does nothing if it depends on the temperature and the thermoelectric toolbox has its own function to do it.

What is the best solution ?
@vincentchabannes @Trophime

• backwardstep.cfg symbolic link to backwardstep3d.cfg
• backwardstep2d.jon update
• backwardstep3d.jon update
• file pages/backwardstep/Readme.adoc doesn't contain results and conclusions

We need to describe the process of contributing an example or benchmark:

• template documentation for example and benchmark (see all issue templates)
• where to put the documentation and example/benchmarks (directory layout)
• producing and uploading to DMP vtk.js files

add Chorus AIRBUS testcase to feelpp/toolbox

Is your feature request related to a problem ? Please describe.
we should support embedded graph plot to add for example convergence graphs and outputs plots

Describe the solution you'd like
plotly.js is a nice framework to embed graphics in html page
this can be easily supported in antora framework

Describe the bug
test case fails in singularity on atlas

To Reproduce
Steps to reproduce the behavior:

1. Go to Atlas
2. Start singularity
3. run `feelpp_toolbox_solid --case "github:{repo:toolbox,path:examples/modules/csm/examples/rotating-winch}"
   4.See error

Reading /feel/downloads/feelpp_toolbox_solid/cases/rotating-winch/biele.cfg...
[ Starting Feel++ ] application feelpp_toolbox_solid version 0.105.0-rc.10 date 2018-Jun-12
 . feelpp_toolbox_solid files are stored in /feel/toolboxes/solid/rotating-winch/biele/P2/np_1
 .. logfiles :/feel/toolboxes/solid/rotating-winch/biele/P2/np_1/logs
[modelProperties] Loading Model Properties : "/feel/downloads/feelpp_toolbox_solid/cases/rotating-winch/biele.json"
[loadMesh] Loading Gmsh compatible mesh: "/feel/downloads/feelpp_toolbox_solid/cases/rotating-winch/biele.msh"
[loadMesh] Loading Gmsh compatible mesh: "/feel/downloads/feelpp_toolbox_solid/cases/rotating-winch/biele.msh" done
F0612 06:48:46.079028 42072 nullspace.cpp:128] Check failed: this->hasOrthonormalBasisVectors() basis vectors are not orthonormal
*** Check failure stack trace: ***
    @     0x7fdb4b3d9acc  google::LogMessageFatal::~LogMessageFatal()
    @     0x7fdb5598cbfa  Feel::NullSpace<>::orthonormalizeBasisVector()
    @     0x7fdb51001944  _ZN4Feel9NullSpaceIdE12updateForUseINS_13FunctionSpaceINS_4MeshINS_7SimplexILt3ELt1ELt3EEEdLi0EEENS_5basesIJNS_8LagrangeILt1ENS_9VectorialENS_10ContinuousENS_14PointSetFeketeELt0EEEEEEN5boost9parameter5void_ESH_SH_E7ElementIdNS_11VectorUblasIdNSF_7numeric5ublas6vectorIdNSM_15unbounded_arrayIdSaIdEEEEEEEEEEEvRKSt16initializer_listIT_EN4mpl_5bool_ILb1EEE
    @     0x7fdb50ffb1d7  _ZN4Feel9NullSpaceIdEC2INS_13FunctionSpaceINS_4MeshINS_7SimplexILt3ELt1ELt3EEEdLi0EEENS_5basesIJNS_8LagrangeILt1ENS_9VectorialENS_10ContinuousENS_14PointSetFeketeELt0EEEEEEN5boost9parameter5void_ESH_SH_E7ElementIdNS_11VectorUblasIdNSF_7numeric5ublas6vectorIdNSM_15unbounded_arrayIdSaIdEEEEEEEEEEERKSt16initializer_listIT_E
    @     0x7fdb50b31edf  _ZN4Feel10FeelModels6detail12getNullSpaceIN5boost10shared_ptrINS_13FunctionSpaceINS_4MeshINS_7SimplexILt3ELt1ELt3EEEdLi0EEENS_5basesIJNS_8LagrangeILt1ENS_9VectorialENS_10ContinuousENS_14PointSetFeketeELt0EEEEEENS3_9parameter5void_ESI_SI_EEEEEENS_9NullSpaceIdEERKT_N4mpl_4int_ILi3EEE
    @     0x7fdb50b2d113  Feel::FeelModels::SolidMechanics<>::init()
    @           0x458522  Feel::runApplicationSolid<>()
    @           0x43dc8f  main
    @     0x7fdb22396830  __libc_start_main
    @           0x43d0b9  _start
    @              (nil)  (unknown)
Aborted (core dumped)

The porosity appears to be considered scalar in the mixed-Poisson sources. It could be interesting to generalize to tensors in order to manage anisotropic situations.

add FSI examples initially describe in
A solution for the incompressibility dilemma in partitioned fluid–structure interaction with pure Dirichlet fluid domains ( Ulrich KüttlerChristiane FörsterWolfgang A. Wall)
with the square and the bended domain.
Others description/results can be found in :

• Generalized Robin-Neumann explicit coupling schemes for incompressible fluid-structure interaction: stability analysis and numerics ( Miguel Angel Fernández, Jimmy Mullaert, Marina Vidrascu)

add doc + example of python code to access girder

Describe the bug
the images are missing in

• https://docs.feelpp.org/cases/latest/electric/rect/readme.html
• https://docs.feelpp.org/cases/latest/electric/README.html
• https://docs.feelpp.org/cases/latest/electric/quarterturn/readme.html

@Trophime would it be possible to fix this ?

• building fails
• motor example fails

Description of the example

This example shows the distribution of the electric potential $V$ in a nanoscale floating-gate nMOS (Metal-Oxide-Semiconductor) transistor working in inversion conditions.
The prefix "n" indicates that the electric current is due to negatively charged electrons.
This simulation involves a non-homogeneous problem with internal interfaces and it has been solved with the HDG method applied to the Darcy problem.

Geometry

A scheme of a realistic floating-gate nMOS transistor used as a nonvolatile memory device is shown with the schematic representation of the two-dimensional cross-section in which the control gate is now included for sake of simplicity in Figure ??

The device is composed of a pair of n-doped source and drain regions, a p-doped substrate region and a silicon dioxide (SiO2) region in which a control gate and a floating gate are buried.
The floating gate is isolated from the control gate by an inter-polysilicon dielectric (IPD) layer beneath the control gate.

Input parameters

Physical parameters:

• electron charge
• permittivity of vacuum
• relative permittivity of silicon
• relative permittivity of silicon dioxide
• thermal voltage
• intrinsic concentration

Device parameters:

• horizontal length
• vertical length
• width
• oxide thickness
• source and drain lengths
• channel length
• junction depth
• source potential
• bulk potential
• source doping concentration
• bulk doping concentration
• drain doping concentration

Model & Toolbox

• The main goal of the simulations is to determine, via the integral boundary condition, the value attained on $\Gamma_G$ at inversion conditions by the electric potential.
  This value is the threshold voltage of the device, denoted henceforth by $V_T$ , and is a fundamental design parameter in integrated circuit nanoelectronics.
  The accurate prediction of the threshold voltage may be significantly influenced by the numerical treatment of the heterogeneous material device structure across the interface $\Gamma_{int}$.
  With this respect the HDG framework appears to be the ideal computational approach because it allows to enforce, at the same time and in a strong manner, (a) the jump of the normal component of the electric displacement vector across the interface and (b) the continuity of the electric potential, at suitable quadrature nodes on each face belonging to $\Gamma_{int}$, unlike standard displacement-based finite element discretization schemes which satisfy property (b) but fail at satisfying property (a).

• poisson: give toolbox name

Materials

already in the geometry and in the input parameters

Boundary conditions

Boundary conditions:

• $V = \bar{V}j + \bar{V|{bi,j}$ on $\Gamma_j, j=S,D,B$
• $\underline{D} \cdot \underline{n} = 0$ on $\Gamma^{lat}_si \cup \Gamma^{lat}_ox$
• $\int_{\Gamma_G} \underline{D} \cdot \underline{n} d\Sigma = q \bar{N}B \delta L{ch} L_z $ on $\Gamma_G$

Interface conditions:

• $[V]{\Gamma{int}} = 0$
• $[\underline{D}]{\Gamma{int}} = q \bar{N}_B \delta$

The quantities \bar{V|_{bi,j} are the built-in potentials associated with the subdomain regions
$\Omega_j , j = S,D,B$, while $\bar{N}_B$ is the concentration of ionized dopants in the bulk region and �$\delta$ is the width of the accumulation region in the $y$ direction of the channel.

Outputs

describe the output set of the example

Fields

add scalar vectorial and matricial fields to be visualized

Measures

add measures, scalar quantities, mean values, performance metrics

References (articles, papers, reports...)

• section of the HDG article in preparation

Is your feature request related to a problem ? Please describe.
Currently the way we get to output from Feel++ apps is cumbersome and not very user friendly
We should be able to retrieve the data from the simulation in an easier way

Describe the solution you'd like
Several solutions were discussed:

• automated upload to girder
• Download buttons (log files, restarts files, result files paraview, results files csv)

Additional context
The duscussion was with a participant of the ECMI 2018 conference

Description of the benchmark

compare the performances of different linear solvers in linear elasticity

Geometry

• cantilever

Input parameters

• mesh size
• FEM order
• pc-type
• ksp-type

Model & Toolbox

• toolbox: csm - linear elasticity

Materials

compressible

Boundary conditions

describe set of boundary conditions

Outputs

performance metrics:

• number of iterations
• time
• relative time

Fields

add scalar vectorial and matricial fields to be visualized

Measures

add measures, scalar quantities, mean values, performance metrics

Benchmark

Describe Benchmark type:

• Performance

The performance will be compared with SOFA framework

Is your feature request related to a problem ? Please describe.
There is some boilerplate when adding a new case documentation and input files.
This could be simplified/accelerated via a template system

Describe the solution you'd like
Ruby provides some nice tools and in particular supports Liquid templating.

the testcase is described in the Feel++ courses

@olmes-sch you will work with @celineCQ

/cc @vincentchabannes

2D building aerothermal flow

/cc @vincentchabannes

In the research area of building physics, the so-called HAMSTAD (Heat, Air and Moisture STAnDardization) project is a very well known benchmark for the testing of simulation tools

see e.g.

• van schijndel_paper.pdf
• Sander_Goesten_MSc_Final.pdf
• HamstadBench1TUEComsol.pdf

use toolbox application report and add it to the antora site along the description of the testcase

the report is at the toplevel of the application repository

see course example

/cc @vincentchabannes

Is your feature request related to a problem ? Please describe.
We don't have CI support for toolbox. We should be able to know the status of all the cases stored

Describe the solution you'd like
we should have a quick way to assess the health of the case files and also be able to run all cases
I suggest using --check-init to run simulation up to init() stage included
we then use buildkite to run all the registered cases

Is your feature request related to a problem ? Please describe.
A ECMI 2018 participant would like to enable reduced order methods from Feel++ in MSO4SC

Describe the solution you'd like
The offline stage can definitely be deployed and the output databases can be then uploaded to Girder.

Additional context
Request from a ECMI 2018 participant

Description of the example

add the 2D Chorus AIRBUS testcase

The documentation is here

http://docs.feelpp.org/cases/0.105/heatfluid/cabin/README/

Add the magnet thermo-electric example in https://github.com/feelpp/feelpp/tree/develop/toolboxes/thermoelectric/ElectroMagnets in feelpp/toolboxes.
It could start as an example or could be in benchmarks if we compare the results with what is obtained with current hifimagnet models and the thermo-electric toolbox

Is your feature request related to a problem ? Please describe.
is it possible to compute outputs such as

• min
• max
• mean
  either global or local(markers)

Describe the solution you'd like

if yes, how is it done in the PostProcessing section?

Additional context
This was asked by one participant of ECMI 2018 conference.