Quantcast
Channel: Solving, solving\ - Latest Discussions - COMSOL Forums
Viewing all 3988 articles
Browse latest View live

Electrical machine 2D to 3D

$
0
0
Hello!

I would like to extend my 2D model of a switched reluctance motor to 3D, to investigate the end effects, but I have some problems solving the 3D model. The 2D model works very well and has been used a lot. I am quite new to comsol and magnetostatic FEM in generel and would like to hear if someone could tell me about generel issues when a 2D magnetostatic models are expanded to 3D.

I use the AC/DC module, the physic is Magnetic fields and I use solve a stationary problem with default solver settings.

Supersonic Expansion into a Laminar Two Phase Flow

$
0
0
Dear all,

I am using Comsol 5.0 and I am trying to solve a problem in 2D which contains a supersonic expansion of CO2 (Mach 1.19) from 2 of 25 micrometer holes into a 6.35 mm tube where water flows laminarly with 0.03 m/s.

I tried to connect High mach flow tool into the Laminar Two Phase flow (phase field). I had lots of problems with complex numbers and I had to make them available. I am using Automatic Newton Damping with Pardiso solver. The convergence is gradually increasing usually at 1e-7, 1e-6 second of the time dependent study.

Could anyone who help me in here?

Kind regards

specifying time step

$
0
0
Hello
I'm new in COMSOL. I wanna solve a transient problem but I don't know how exactly set my time step. actually I got confusion between time step which would defined by study step and time step in the solver.
any hints will be appreciated.

Best regards
Shahin

Symmetric Stress ,Without Fixed Constraint

$
0
0
Hi guys,

I have a symmetric structure which is imposed by symmetric stresses , and i wanna know the displacement of top point of this model without any fixed constraints.

But it cannot be solved ,says
"Failed to find a solution.
The relative error (2.8) is greater than the relative tolerance.
Returned solution is not converged.
- Feature: Stationary Solver 1 (sol1/s1)".

Does anyone have any experience on this kind of model? I don't wanna add fixed constraints, cause it is the most similar one to real situation. Is it possible?

Thanks for any help.
Emma

Time dependent Contact modeling with Automatic Remeshing

$
0
0
Hi
I am trying to model a Sand grain which is pressed completely into a hyperelastic material (rubber).
My sand grain has a diameter of 200 µm.
I am using a prescribed displacement to move it into the rubber.
I also tried to use a stationary study. There I used the Stop condition and remeshing a deformed mesh and a parametric sweep.
I am able to move the sand grain about 105µm out of 200 into the rubber.
After this the mesh starts overlapping and is not converging any more.

As it is complicated to remesh by hand lots of time I tried to put my model in a time dependent study
with an automatic remeshing condition.
I am also using a prescribed displacement there. I want to move my sand grain with 10µm per second into the rubber.
When I try to use the augmented Lagrangian contact it does not converge at all.
Then I tried it with penalty. Now it solves for the first two seconds. After this it starts to remesh but doesn`t get a solution and fails.

I am also not sure which boundary conditions I have to use with the moving mesh.

I attach my two models.
Please have look at it and tell me how to solve my problems.

Any helb would be appreciatet.

Thanks in advance
Carolin Rüss

solving PDE with source term (right-hand side) as matrix from MATLAB

$
0
0
hi, all

I am trying to use coefficient from PDE to solve helmholtz type partial differential equation.
(I know there is helmholtz equation, but both module doesn't working)

So, first I am using COMSOL 4.4 with MATLAB, and I follow the instruction as following link to use MATLAB function as source. (www.comsol.kr/support/knowledgebase/1093/)

since the input matrix is 20x60, I discretized my geometry into 1200 elements,
The problem is after i meshed my geometry, COMSOL continuously says as following

The following feature has encountered problem:
Incorrect size of returned vector, 2000 elements were expected, but the returned matrix was 20x60
- Function: f
- Feature: Stationary Solver 1 (sol1/s1)

So, I have tried 20x100 matrix, then COMSOL says
The following feature has encountered problem:
Incorrect size of returned vector, 400 elements were expected, but the returned matrix was 20x100
- Function: f
- Feature: Stationary Solver 1 (sol1/s1)

any help or comments will be appreciated

inverted mesh element near coordinate

$
0
0
Hi dear IVar

I want to see the mechanical changes due to flow of current through two parallel wires in COMSOL. To do this, first I used two "magnetic field" and "solid mechanic" modules. After getting the results I noticed there is no filed change resulted by mechanical changes which happens due to magnetic forces. After reading previous answers you have commented for others, I used "moving mesh" module to execute the changes. But then I faced this error:
"inverted mesh element near coordinate"
What I have to do in order to solve the error?

I used this terms in my modeling:
1)free deformed(for all domain, 2-wire and air)
2)prescribe mesh displacement(just for boundary of wire )

With best thanks and regards

"inverted mesh element near coordinates"

$
0
0
Hi dear Ivar,
Thank you for your helpful advices,
Actually, I modified my model by your comments, and it was working correctly when I had chosen "extremely coarse, extra coarse, coarser" for mesh elements that controlled by "physics control mesh". but when I used "coarse, normal and etc," for mesh elements, I saw an error that was "inverted mesh element near coordinates".
So I added "automatic remeshing " to my model. and I selected "General" for "Condition Type", and "(comp1.ale.I1isoMax)>4". but after it, I saw non-convergence solver error. Thank you for your valuable time.
Please inform me what should I do?

Thank you so much and with best regards,
Reza,

Complex Band Strucutre - Acoustics/Structural Mechanics

$
0
0
Hi

In a typical phononic band structure calculation done within COMSOL the procedure is to define Floquet BC's and do a parametric sweep of the wave vector over a particular direction in the irreducible Brillouin zone. The output are frequencies and the input is wave vector. Simple enough.

However, it is also worthwhile to study the evanescent modes in lossy materials, this is done by calculating the complex band structure (usually done with an extended plane wave expansion procedure). The complex band structure is calculated by using a real frequency as input and solving for complex wave number. Thus resulting in two dispersion curves, one is omega vs real(k) and one is omega vs imag(k). My question is then can this easily be done within COMSOL? Even if someone has done this within the scope of optics or electronics, it would be helpful.

thanks
~Chris

Simulating a coil with current above ground, no B field found

$
0
0
Hi,

I am building a simple AC/DC model: a half space as air, a half space as ground and there's a two-turn coil in the air above the ground. I want to put a AC current into the coil and get the magnetic field. I build a helix as the coil and set a current in the global definitions. I used magnetic fields as physics and added a multi-coil domain. However, there is no magnetic field after meshing and computing.
I'm quite new to comsol and wondering if you know which part is wrong in this model? I'll be really appreciated if you could give me a hand.
The model is attached below.

Best regards,
Yu

DC discharge - Undefined value found in the stiffness matrix

$
0
0
Hi everyone,
I’m trying to solve a time-dependent study on a DC discharge 2d model. I inserted some fictitious boundaries in the geometry in order to facilitate the creation of a mapped mesh, but when I try to compute the study the following error occurs:

Undefined value found.
- Detail: Undefined value found in the stiffness matrix.
There are 1000 equations giving NaN/Inf in the matrix rows for the variable comp1.En.
at coordinates: (0.00101,4.3275e-005), (0.00101,2.625e-005), (0.00103,4.3275e-005), (0.00103,2.625e-005), (0.00105,4.3275e-005), ...
and similarly for the degrees of freedom, NaN/Inf in the matrix columns.

I think that BC and meshing are well defined, in fact the model is successfully solved if the fictitious boundaries are not inserted in the geometry and a free triangular mesh is used. But I would prefer to use a mapped mesh. Can someone help me?

LVDT design

$
0
0

HI IVAR,

i am currently working in LVDT Design(Linear Variable Differential Transducer) particularly used for displacement measurement..
i have designed the mechanical model..
i have attached the file what i have done..
my doubt is
how to assign fields such as voltage, magnetic field..
can you give me suggestions???

Issues with Numerical Noise

$
0
0
I am solving a simple classical diffusion equation for 'c' with the diffusion coefficient being directional dependent (function of grad(c)). I am getting some weird results and numerical noise for the diffusion coefficient as it is being calculated.

Can anyone that has possibly had this issue please aide me in possibly fixing this issue. I have attached the simple model file to examine.

Thank you​

how many boundary condition should be imposed on spatially first order differential equation in comsol?

$
0
0
I'd like to solve a 1st order differential equation like this in 1d domain:

du/dx =u
x in [0,2]
u(0) = 1000

while I was using the General Form PDE in stationary , there are two boundary point in x=0 and x=2.
the default condition is zero flux condition at both ends. However, for a 1st order equation there should be only one boundary condition. zero flux at x = 2 is over-determinant for this problem.

The results obtained is also oscillating as shown in the attachment.

So, my question is: what condition should I impose at x=2? for this kind of equation?

AC/DC simulation_Space charge added to spheres

$
0
0
Hello,
I am new to the AC/DC modelling using Comsol.
Basically, I have a 3D unit block with 8 nano-particles inside. Each nano-particles is consisted of four different layers. I applied the positive space charge to the inner sphere and the negative charge to the outer three. In the meantime, I applied the voltage and ground to two opposite surfaces of the unit block. Ideally, the space charges I added should be an exponential function 5*exp(w1*(10^9)-7.5 and -5*exp(7.5-w1*(10^9) where W1 is the distance between any point on the sphere to the centre. These two functions are supposed to be symmetrical but different polarities. However, once I plotted the space charge of the cut line passing through the centre of the sphere. It does not seem to be symmetrical, even without any voltage applied around the unit block.

I am not sure if the space charge I added has the correct formula or the way of adding the charge is correct. I have attached two graphs which are the ideal space charge and the actual space charge generated from my model, respectively. As I mode is too large, it can be found from the Dropbox:

www.dropbox.com/s/6cw7d8iozdsu...2wt%25c%2815.01.27%29.mph?dl=0

Appreciate any suggestions and comments on these problems.

Best,

Yang

Evaluate derived values

$
0
0
Hello,

In the Results node, I would like to evaluate several expressions of derived values simultaneously in the same table. For example, I have to evaluate the Surface Maximum of stresses sx, sy and sz and I would like to obtain each one in a column in the same table, without creating a distinct subnode "Surface Maximum" for each expression.

Thank you in advance.

3 D Electrostatic Force Calculation

$
0
0
HI,
I am a newbie to Comsol, and i am using 4.3 version and working on a mems gyroscope. i need to calculate electrostatic actuation force from the model,i am using a comb drive structure as an actuator.
I am not sure which physics i should use and how to calculate the force produced after giving an excitation voltage.
Thanks in advance

input configuration time discrete solver

$
0
0
hello,

how can i input parameter or variables to config the time discrete solver, especially the parameter prevlevels? What is the max number to store previous data?

Thanks. jojo

Undefined value found in the equation residual vector.

$
0
0
I am trying to model a pem fuel cell. Although I have entered all parameters and values, i have some errors as follows;

Undefined value found.
- Detail: Undefined value found in the equation residual vector.
There are 4 degrees of freedom giving NaN/Inf in the vector for the variable mod1.wH2Oc
at coordinates: (0.0022813,0), (0.0022432,0), (0.0022813,7e-005), (0.0022432,7e-005), ...
There are 4 degrees of freedom giving NaN/Inf in the vector for the variable mod1.wO2
at coordinates: (0.0022813,0), (0.0022432,0), (0.0022813,7e-005), (0.0022432,7e-005), ...

Can you help me please?

Import results from a different simulation as a boundary condition or solve simulation for each geometry separetly

$
0
0
Hi folks,

I am simulating a pipe flow with 2 sections, and due to the dense mesh, the cluster is running out of memory.

So I was wondering: is it possible to simulate the first section and save the results, then later on import the profiles from this simulation into another file?

I just need the profiles at the pipe exit so I can use it as a boundary condition in the inlet of the second pipe.

Edit:

Another option would be to solve for each geometry but not at the same time, in other words, solve for the various
geometry parts but one after the other.

Is this possible or something similar?

Best Regards,

Rui Silva
Viewing all 3988 articles
Browse latest View live


<script src="https://jsc.adskeeper.com/r/s/rssing.com.1596347.js" async> </script>