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

Rotordynamics, eigenvalues and frequency domain

April 6, 2014, 3:29 am
$
0
0
Hi everybody,
I have two questions about rotordynamics modelling.

First of all, I'm modelling a rotor using a parametric sweep on angular velocity to find its eigenvalues.
After that I would like to extract the system matrices (in LiveLink with Matlab) using the command "mphmatrix", but the damping matrix (which includes the gyroscopic matrix) varies with the angular velocity of the rotor.
So, how can I extract system matrices for a certain step of the parametric sweep? (or for every step of the sweep, it's the same!)

My second question is related with rotordynamics too.
I would like to perform a frequency response analysis of a rotor. In my model the rotor is modelled with beams and the bearings with the Lubrication shell interface. There are also some ODEs so I substituted the time derivatives with i*2*pi*freq and -(2*pi*freq)^2.
The problem is that in this way the ODEs generate an error like this:

Undefined value found.
- Detail: Undefined value found in the stiffness matrix.
There are 1547 equations giving NaN/Inf in the matrix rows for the variable mod1.pf.
at coordinates: (-0.0799129,0.00534406,0), (-0.0768651,0.0223399,0), (-0.0788467,0.0139241,0), (-0.0702572,0.0382926,0), (-0.0739908,0.0304943,0), ...
There are 6 equations giving NaN/Inf in the matrix rows for the variable mod1.thx.
at coordinates: (0,0,0), (0,0,-0.07), (0,0,-0.14), (0,0,-1.3), (0,0,-1.37), ...
There are 6 equations giving NaN/Inf in the matrix rows for the variable mod1.thy.
at coordinates: (0,0,0), (0,0,-0.07), (0,0,-0.14), (0,0,-1.3), (0,0,-1.37), ...
There are 6 equations giving NaN/Inf in the matrix rows for the variable mod1.thz.
at coordinates: (0,0,0), (0,0,-0.07), (0,0,-0.14), (0,0,-1.3), (0,0,-1.37), ...
There are 1 equations giving NaN/Inf in the matrix rows for the variable mod1.tilt1.
at coordinates: (0,0,0), ...
There are 1 equations giving NaN/Inf in the matrix rows for the variable mod1.tilt1B.
at coordinates: (0,0,0), ...
There are 1 equations giving NaN/Inf in the matrix rows for the variable mod1.tilt2.
at coordinates: (0,0,0), ...
..................

I made a simple check about ODE in frequency domain: I modelled (with ODE interface) the frequency response of a 1D mass-spring model, and all work well!
My rotor model is quite complex because there are three physics involved; in particular I think the problem could be in the non linearity of the Lubrication shell Interface used to model the bearings.
Any advice?

Thanks a lot

Best regards
Daniele
↧
Search

mode analysis, not getting the right neff

April 7, 2014, 6:50 am
$
0
0
Hi:
I am solving one file using Comsol 3.5a but I am getting neff to be 3.8. I am supposed to get it 3.88.
Can anyone check why my neff does not go higher?. I have been trying for days and the best thing I could reach is 3.82 . I really need to get 3.88 as it is the mode I am searching for. The file is attached.
Naema
↧
↧

Help for Modeling Transmission and Reflectance of Light for a Stacked Layer Smart Window Device

April 7, 2014, 9:20 am
$
0
0
I am an undergraduate student at University of Maryland, College Park. I am in a capstone research group, where we are designing a smart window device that consists of stacked layers enclosed in glass. When we apply a small voltage to the device, the device becomes partially transparent, and when in the off-state, the device is practically fully reflective.

Our team is required to have modeling work of our device, and we want to use the COMSOL 4.4 Wave Optics Module to model the transmission and reflectance vs. wavelength of light. We are all new to COMSOL, where we just recently have acquired our license, and we have been running through tutorials to learn the software.

I can provide more info on our device if needed.

We were wondering if anyone could steer us in the right direction on how to set up our model, including any parameters, variables, and equations/functions we would be using to solve for the transmission and reflectance. We were also wondering about mesh info.

One of the tutorials we went through seemed similar to what we want to do.
The description is: A plane electromagnetic wave propagating through free space is incident at an angle
upon an infinite dielectric medium. This model computes the reflection and
transmission coefficients and compares the results to the Fresnel equations.
The link to it is: www.comsol.com/model/download/....woptics.fresnel_equations.pdf

It seems to need to add boundary conditions to our model, as well as ports to set-up the wave propagation into our material. The model we followed used a parametric sweep with a direct solver.

Any help/info on how to model transmission vs wavelength for a stack of layers would be much appreciated.!
↧

linear system Solvers

April 8, 2014, 1:04 am
$
0
0
Hi all,

Basic question about solvers in comsol, does anybody know that what is the difference between direct solvers:
1) UMFPACK 2)SPOOLES 3)PARADISO 4)TAUCS

I tried to read manual, but it seems vague to me.
Thanks for your help
Sepideh
↧

COMSOL 4.0a : how to set maximum Java heap size??

September 9, 2010, 9:34 pm
$
0
0
Hi,
I have problem to set comsol java heap.
In v3.5a, I know I can modify java heap in comsol.opts.
But in v4.0a, I can't find the comsol.opts, and I have no idea about setting the java heap.
Please give me help~~~
THX~
↧
↧

Viewing the whole equation system

January 4, 2012, 5:00 am
$
0
0
Hello,

is it somehow possible to take a look at the whole equation system, COMSOL has built, after it was solved?

Thanks is advance.
↧

Error: Failed to find time step so that given tolerances are satisfied

May 3, 2012, 11:03 pm
$
0
0
Error: Failed to find time step so that given tolerances are satisfied

This is the error that Generalized-alpha gives. I think the BDF solver in similar cases gives the error
"Cannot meet error tolerances. Increase absolute or relative tolerance."

This happens in odd situations. For example, I have a parameter that, if I set it to 0.1, the error doesn't happen, 0.15, it DOES happen, for 0.2, 0.3, 1.0, it doesn't happen. If I adjust the absolute tolerance, then for 1.0 it DOES happen, but for 1000 it doesn't happen. There seems to be no clear way to characterise when it occurs. All I know, is that it happens around the same time in the simulation as a boundary condition changes according a (smoothed) step function (in time). There is a generous amount of smoothing.

My question is: Why does it happen at all?

In my general understanding of solvers, there should always be a reasonable time step at which the error will be small enough. If there is too much error, then we reduce the time step. And this is what the solvers seem to be doing. But all of a sudden, there is a time in the simulation, when reducing the time step by 5 or 6 orders of magnitude, compared to what's normally needed, does not reduce the error enough.

I'm trying to build an understanding of this behaviour.

I've tried adjusting many of the solver parameters. So far, most of them don't have an effect on the issue, except Amplification for high frequency. (This is for Generalized alpha.) What is this, and why does it make a difference? I've checked the User and Reference documentation, and there is very little information, except to say that it's "damping of higher frequencies".

This issue is really causing a lot of trouble. Even at fairly modest amplitudes of the step in BCs.
↧

auxiliary sweep

April 10, 2014, 4:16 am
$
0
0
Hello everyone,

I am simulating a reacting flow with a rather fast reaction. This problem does not converge, and I found out that I should try an auxiliary sweep with a parameter in front of the reaction rate that is increased step by step using last steps solution as initial guess.
For the first step this is working perfectly, I get a good solution for 1/1000 of the reaction rate. But after the first step the study stops and does not continue with the next step. Anybody has an idea what I am doing wrong?

Here is what I entered so far:
I checked auxiliary sweep in my stationary solution and added the parameter (which i before defined in global definitions as 1[1].) I chose Specified combinations, gave a range of 4 steps and chose run continuation for last parameter (but it is the same for no parameter and then reuse solution yes)
under solver - stationary solver - parametric I did not change anything.

Any idea/help/hint? Thanks in advance!
Best regards,
Miriam
↧
Search

Simulation time does not scale with processor

April 10, 2014, 11:40 am
$
0
0
Hello all,

Just as a preface, I do not have significant experience with Comsol in general, and the bulk of my manipulations are occurring within Matlab, but my question is of a general sort.

I am running simulations in Comsol, though Matlab, of a two dimensional silica wedge with azimuthal symmetry at various radii for the purpose of obtaining the form of the mode profiles and various properties related to loss, dispersion, etc, within with wedge for an arbitrarily sized ring resonator.

The Matlab code that I'm using to run the simulation simply controls the inputs into Comsol, and does not do any real computation itself.

My situation is this:
I had previously been running the simulations on a laptop with an i7-2720QM (quad core) CPU which CPU-Z shows clocking up at 3GHz, and has 8GB of RAM.

I recently built a new computer with an i7-4930K (hex core) which clocked out of the box to about 3.7GHz, and I further OCed to 4.4GHz. It has 32GB of RAM.

In both cases, Comsol is not using all of the RAM, since each individual simulation isn't very big (it takes 30-60 seconds to complete), I just have to do a lot of them back to back.

Both machines are running Comsol 4.4 and a 64-bit OS, the laptop is Windows 7 with Matlab R2011a, and the desktop is Windows 8.1 with Matlab R2014a.

Even straight out of the box, the new CPU should be about 2.5-3x more powerful than the laptop CPU in terms of simply cores and clock speed (in reality it scales even higher). However, the speed at which it runs the simulations was only about 25% faster. Furthermore, after I overclocked the CPU, the speed of the simulations scaled almost perfectly with the increased clock speed, becoming about 50% faster than the laptop overall.

So essentially, my issue is that the simulation speed scaled horribly with the switch to a more powerful processor, but perfectly with the clock speed of the processor itself.

I should also add that I monitored the total CPU usage during runs for both cases, and for both machines, Comsol consistently made use of 50-60% of the processing power during computation, so it is not as if the new chip just isn't being utilized effectively.

I apologize if I have not provided sufficient information, and if more specifics on the simulation itself are required, I can try to obtain them from the person who wrote it.

Is there any particular reason for the speed to scale so horribly, such as bad support for Ivy-E chips?

Any and all insight is appreciated.
↧
↧

Electrostatics module doesnt work

April 11, 2014, 1:52 pm
$
0
0
Hello ,
I am trying to simulate Sharland et al's paper with 6 species corrosion process. After long weeks of attempts I have used constraints to enforce equilibrium constraints.
Please find the attached model. It doesnt seem to generate any V(from electrostatics module) I tried to make mesh finer, doesnt converge.With a little coarser mesh , no V is captured.Face "undefined Value found' with a defined initial value for V.I have tried to replicate an existing model in 3.5a to 4.3a.
I will appreciate any inputs !
↧

element order and convergence

July 27, 2011, 2:43 pm
$
0
0
I have a FSI problem that converges when I use first order elements for pressure and velocity and second order for displacements. But I am not happy with the results I am seeing. Hence I changed the element order.

I am using second order elements for the velocity and displacement and first order for pressure. Now the problem is not converging.

It should not be behaving this way because I think that increasing the order of the element should give better convergence. Am I wrong? Did anybody face a similar problem?

Thanks,
Sirisha
↧

Modeling a droplet formation

April 13, 2014, 12:22 am
$
0
0
Hello,

I am currently trying to model in 3D droplet formation such as T-junction.
Attached file is my model.
This model was simplified to calculate my comuputer.
I want to calculate large model if possible.
However, a large amount of memory is needed to calculate the large model.
I have adjusted the solver setting.
As one example, I used to segregated solver for save memory.
But, my approach was not sufficiently effective for saving memory.
Please help me.
Thank you so much.
↧

periodic function - picewise works, analytic doesn't

April 14, 2014, 12:55 am
$
0
0
Hey,
I’m fairly new to COMSOL and trying to use the Electric Currents and Heat Transfer Interface to simulate the heating process caused by an applied voltage in a solid.
If I implement the applied Voltage using the piecewise function the simulation works just fine, but if I try to implement the signal using a rectangular function and making it periodic by using the analytic function (like I read it in a few other discussions), the simulation works a lot worse (meaning: it takes much longer to solve, the time steps need to be a lot smaller (0.01 -> 0.005) and the convergence plots don’t look very good).
I want to be able to change the signal (“on”-time, “off”-time, amplitude, and the overall pulse train time) without having to change all the values in the piecewise function. I would really appreciate it if anyone could tell me what the difference is between the piecewise and analytic function since I seem to be missing something here. Like I said it is not working properly with the analytic function and to me they look exactly the same.

This is what I have done so far:
The applied Voltage is a periodic rectangular function for about 15 seconds (the last pulse should finish regularly before the signal stops), after that it is zero before the simulations stops (to look at the cooling due to convection).
The amplitude is 50 V, the pulse is turned “on” for 500 ms and turned “off” for 150 ms.

Version 1 (works perfectly)
Piecewise function (pw1):
-> see pw1.jpg
I implement the signal with the terminal at the electric currents interface (terminal type: voltage): 50*pw1(t[1/s])
I use a userdefined mesh and the time dependent study.

Version 2 (doesn't rally work) - I use the rectangle and analytic function to implement the pulse:
Global Parameters:
rt: rise time = 25 ms
ampl: amplitude of the signal = 50
p: period = 650 ms
V_an: time for how long the pulse sequence should be turned on = 15 s

rectangle (rect3):
-> see rect3.jpg

analytic (an5):
-> see an5.jpg

analytic (an6):
-> see an6.jpg

I implement the signal with the terminal: an6(t[1/s])
I use a userdefined mesh and the time dependent study. Also I changed the time stepping in the time-dependent solver intermediate.

Thanks a lot for your help!
Yasmin
↧
↧

How to implement the Beer-Lambert absorption law when ablating a solid.

March 8, 2011, 6:52 am
$
0
0
Hi,

I have a model which I am heating with a laser heat source.
The laser is absorbed in the solid according to Beer's law, that is the intensity of the radiation exponentially decays with path length, [ exp(-alpha*x) ] where x is path length.
This all works fine.

I am trying to implement albation in my model.
When the temperature goes above the vaporization temperature the material is 'ablated'.

My problem is that Beer's law scales with x so that it starts at x = 0 even when part of the material is ablated the equation still scales from x=0.

Basically I need to redefine my coordinate x as the ablated region moves across the material so that x starts at zero and increases linearly.

I have tried using a condition (T>300)*x where 300 is the vaporization temperature.
This condition works in that the x is zero up until the edge of the ablated region but then x jumps to whatever value it had before i.e it doesn't scale linearly.

If anyone has any information that might help me I'd really appreciate you letting me know.

Sorry if the explanation is extremely confusing!

Regards,
-Adam
↧

Geometry Parametric sweep

July 11, 2013, 7:52 am
$
0
0
Hello,
I want to run a geometry parametric sweep.
Consider the following geometry sweep parameters:
H range(20[mm],10[mm],60[mm])
Z range(0[mm],10[mm],20[mm]+H)

The idea is that Z maximum value is against the height H of the object in the current nested iteration of Z against H.
In the first round (H=20) maximum Z will be 20+H=40. When first round of Z sweep finishes, H will become 30 and the maximum Z will be 50[mm].
This is not happening and ComSol is using the H global defined parameter.
Is it possible to force ComSol to use in the Z parameter sweep range the H value as calculated in the for the current parameter sweep step?

A simple code for that will be
for (H=1 to N)
for (Z= 1 to H)
mesh.....
next Z
next H

Thank you,
Dimitris
↧
Search

time dependent study for sinusoidal wave

April 2, 2014, 10:34 pm
$
0
0
Dear all,

I'm trying to simulate my module in 2D axis of COMSOL version 4.3 in joule heating physics with sinusoidal input voltage of { V0*sin(2*pi*50*t)} and for the frequency of 50 Hz it will takes 0.02 second for full wave . so i must use small time step of around 0.0001sec. . i have problem when i increase the time to more than 5 second with step of 0.0001 sec .

I need to simulate my module for time between 0 sec until 300 second with time step of 0.0001. as times: range(0,0.0001,300)s
So it needs around 3 million step to get the full result and i think it's difficult to that .

is ther are any way to get it in the transient study so i can check the time that need to arrive the steady state temperature .

best regards ,




↧

How to solve the shutdown problem due to "flmklblas.DLL"?

April 16, 2014, 3:50 pm
$
0
0
I was running a parametric sweep model with direct solver, and my COMSOL 4.1 keeps shutting down when it goes about 60% of the process. The windows 7 logo says the error is due to "flmklblas.DLL". Does anyone know how to solve it? Thank you.
↧
↧

Convergence problem ( PID heater controller)

April 17, 2014, 12:26 am
$
0
0

Hi,

I have a problem with the convergence of my model. I have this message d'error


""
there are 876 degrees of freedom returning NaN / Inf in the matrix rows for the variable comp1.T.
the coordinates (0.021,0.00055) (0.0212876,0.00055) (0.0215991,0.00055) (0.0219364,0.00055) (0.0223016,0.00055) ...
and for the same degrees of freedom, NaN Inf in columns of the matrix.
The linear solver has sent an error message.
The relative residual (6th 051) is larger than the tolerance.
The returned solution has not converged correctly"""

When I consider the power of heater as a consatnt my model ( MOCVD Chamber) converges. Unfortunately,
when I want to control the power of heater( P_adjustable) I had always the same message d'error.
I defined in EDO et EAD global equations P_adjustable=T_consigne-intop1(T)
intop1(T) is temperature in the control point
T_consigne is desired temperature in the control point
I defined T_consigne in parameters
I used a refined mesh, and a resolution with auxilary sweep kappa.
Should i change values of PID Coeffecients ?

I attach my model version 4.


Any idea, help ? Thanks in advance !
Regards,
↧

How to solve the shutdown problem due to "flmklblas.DLL"?

April 17, 2014, 6:42 am
$
0
0
Dear Ivar Kjelberg,
Hi! My name is Shiyi Wang, a PhD student majored in Electro-optics. Your clear response and kind help impress me a lot when I check the solutions on the discussion forum. If possible, could you please give me any ideas about solving the shutdown problem of COMSOL due to its "flmklblas.DLL"? We were using 12 cores 96GB Dell server to run a parametric sweep RF model with direct solver MUMPs based on COMSOL 4.1, which causes the CPU occupation to be almost 100% for most of the time. The system is win7 64-bit. It seems that it can not bear for more than 1 hour. We have tried to update Java, BIOS and windows to the newest version and the problem is still there. What is more ridiculous, another server with less computation capability processes the same model pretty well. For every time it shuts down automatically, we checked the windows log and there is an error caused by "flmklblas.DLL" under COMSOL folder. I have been dealing with this problem for almost half a month. If you could respond to my private message, that will be really great!

Thank you so much! Have a great Easter day!
↧

Failed to find a solution

April 18, 2014, 8:22 am
$
0
0
Hello
I am a new user, and I am very impressed, Comsol is perfectly simulated in chemical reactor design
And when I began using of Comsol I have some problems with it.
And I have a set of PDE's in Dehydrogenation Reactor, and we decided to solved it by comsol
But I get this matter (( Failed to find a solution. Maximum number of Newton iterations reached.
There was an error message from the linear solver. The relative residual (5.8e+006) is greater than the relative tolerance. Returned solution is not converging.))

plz. Help me...
--
Omar S. Lateef
omar@greenpcbtronics.com
00964 7704852473
↧
Viewing all 3988 articles
Browse latest View live
Search
<script src="https://jsc.adskeeper.com/r/s/rssing.com.1596347.js" async> </script>