Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.
Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.
what is wrong with the ALE
Posted 22 déc. 2011, 03:27 UTC−5 Interfacing, Heat Transfer & Phase Change, Studies & Solvers Version 4.2a 7 Replies
Please login with a confirmed email address before reporting spam
(all are in the .doc attachment. Thanks)
Hi,everyone,
I ‘m building a simple model based on the “heat transfer” module and the “ALE” module. The geometry is a simple rectangle. In the heat transfer node, the upper boundary is a heat flux BC, and other three are thermal insulation. In the ALE node, a prescribed normal mesh velocity is implemented on the upper boundary, other three are fixed.
Recently, I have encountered some problems relative to the ALE. Due to this reason, I deactivate the ALE module and solve the model where there is only a “heat transfer” module with some local variable to see where the error and mistake is (to ensure whether there is any wrong variable definition). In the ALE node, a “prescribed normal mesh velocity” is implemented on the upper boundary, where the ”normal mesh velocity Vn” is setting “to roeB_r+rolB_r” (m/s), After solving the model without the ALE, the theoretical normal mesh velocity (I defined a local variable with the expression of “roeB_r+rolB_r” to visualize the variable) is illustrated in the Fig. 1. In the Fig.1, we can see there should be a moving of the boundary.
(see the Fig. 1 in the attachment)
But when I activate the ALE node, the following error occurs (also as the Fig.2 shown).
Attempt to evaluate non-integral power of negative number. - Function: ^ Failed to evaluate temporary symbolic derivative variable. - Variable: mod1.rolB_r@VDN${mod1.genext1@13} - Defined as: ((1/(mod1.dB*((unit_m_cf^2)*unit_s_cf)))*((((((((log((mod1.TB/T_m))*(unit_s_cf*(mod1.DB*2)))/(r_0*unit_m_cf))^(-0.5))*0.5)*((1/(r_0*unit_m_cf))*((((((1/(mod1.dB*mod1.hcB))*(((1/(unit_K_cf*unit_m_cf))*unit_W_cf))))*2)*unit_s_cf)*log((mod1.TB/T_m)))))*unit_kg_cf)*((mod1.Pr_sB/unit_Pa_cf)^0.25))*((((unit_m_cf^3)*mod1.dB)/unit_kg_cf)^0.75))) Failed to evaluate temporary symbolic derivative variable. - Variable: mod1.ale.vn@VDN${mod1.genext1@13} - Defined as: (((1/(mod1.dB*((unit_m_cf^2)*unit_s_cf)))*((((((((log((mod1.TB/T_m))*(unit_s_cf*(mod1.DB*2)))/(r_0*unit_m_cf))^(-0.5))*0.5)*((1/(r_0*unit_m_cf))*((((((1/(mod1.dB*mod1.hcB))*(((1/(unit_K_cf*unit_m_cf))*unit_W_cf))))*2)*unit_s_cf)*log((mod1.TB/T_m)))))*unit_kg_cf)*((mod1.Pr_sB/unit_Pa_cf)^0.25))*((((unit_m_cf^3)*mod1.dB)/unit_kg_cf)^0.75)))) Failed to evaluate expression. - Expression: d(d(((mod1.ale.xt_free*mod1.ale.nx+mod1.ale.yt_free*mod1.ale.ny-mod1.ale.vn)*test(-x_lm))*(dvol),{test@0}),{mod1.genext1@13})
I’m quite puzzled about that, I can solve the model and evaluate all the relative variables without the ALE. This means there is no any wrong with the parts except the ALE section. While the above-mentioned error “fail to evaluate the relative variable” puzzles me a lot.
I have attached my .mph model, is anyone kind to check and give me some suggest?
If you wish, we can cooperate together and co-author, plz contact me.
Thanks sincerely in advance for your response.
Best wishes to you and your family.
Yours
FM Huang
Dec.22th.2011
Hi,everyone,
I ‘m building a simple model based on the “heat transfer” module and the “ALE” module. The geometry is a simple rectangle. In the heat transfer node, the upper boundary is a heat flux BC, and other three are thermal insulation. In the ALE node, a prescribed normal mesh velocity is implemented on the upper boundary, other three are fixed.
Recently, I have encountered some problems relative to the ALE. Due to this reason, I deactivate the ALE module and solve the model where there is only a “heat transfer” module with some local variable to see where the error and mistake is (to ensure whether there is any wrong variable definition). In the ALE node, a “prescribed normal mesh velocity” is implemented on the upper boundary, where the ”normal mesh velocity Vn” is setting “to roeB_r+rolB_r” (m/s), After solving the model without the ALE, the theoretical normal mesh velocity (I defined a local variable with the expression of “roeB_r+rolB_r” to visualize the variable) is illustrated in the Fig. 1. In the Fig.1, we can see there should be a moving of the boundary.
(see the Fig. 1 in the attachment)
But when I activate the ALE node, the following error occurs (also as the Fig.2 shown).
Attempt to evaluate non-integral power of negative number. - Function: ^ Failed to evaluate temporary symbolic derivative variable. - Variable: mod1.rolB_r@VDN${mod1.genext1@13} - Defined as: ((1/(mod1.dB*((unit_m_cf^2)*unit_s_cf)))*((((((((log((mod1.TB/T_m))*(unit_s_cf*(mod1.DB*2)))/(r_0*unit_m_cf))^(-0.5))*0.5)*((1/(r_0*unit_m_cf))*((((((1/(mod1.dB*mod1.hcB))*(((1/(unit_K_cf*unit_m_cf))*unit_W_cf))))*2)*unit_s_cf)*log((mod1.TB/T_m)))))*unit_kg_cf)*((mod1.Pr_sB/unit_Pa_cf)^0.25))*((((unit_m_cf^3)*mod1.dB)/unit_kg_cf)^0.75))) Failed to evaluate temporary symbolic derivative variable. - Variable: mod1.ale.vn@VDN${mod1.genext1@13} - Defined as: (((1/(mod1.dB*((unit_m_cf^2)*unit_s_cf)))*((((((((log((mod1.TB/T_m))*(unit_s_cf*(mod1.DB*2)))/(r_0*unit_m_cf))^(-0.5))*0.5)*((1/(r_0*unit_m_cf))*((((((1/(mod1.dB*mod1.hcB))*(((1/(unit_K_cf*unit_m_cf))*unit_W_cf))))*2)*unit_s_cf)*log((mod1.TB/T_m)))))*unit_kg_cf)*((mod1.Pr_sB/unit_Pa_cf)^0.25))*((((unit_m_cf^3)*mod1.dB)/unit_kg_cf)^0.75)))) Failed to evaluate expression. - Expression: d(d(((mod1.ale.xt_free*mod1.ale.nx+mod1.ale.yt_free*mod1.ale.ny-mod1.ale.vn)*test(-x_lm))*(dvol),{test@0}),{mod1.genext1@13})
I’m quite puzzled about that, I can solve the model and evaluate all the relative variables without the ALE. This means there is no any wrong with the parts except the ALE section. While the above-mentioned error “fail to evaluate the relative variable” puzzles me a lot.
I have attached my .mph model, is anyone kind to check and give me some suggest?
If you wish, we can cooperate together and co-author, plz contact me.
Thanks sincerely in advance for your response.
Best wishes to you and your family.
Yours
FM Huang
Dec.22th.2011
Attachments:
7 Replies Last Post 27 oct. 2016, 22:07 UTC−4
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
for me 2 have at least two issues
1) you coordinate transform is on the spatial frame (dependent on the solution), I expect it to be on the Reference frame (fixed in space), then it's strangely orientated, but as you seem to use only "r" its probably OK)
2) then your Gaussian beam has an amplitude of >1:1E11 that is a lot, do not forget that the binary representation of number has a limited resolution hence a limited dynamic check the value of "eps" and compare to 1, and do not forget that you need sqrt() and ^2 to resolve your physics
3) I fear that you extrusion coupling is not working as expected, but I'm not sure, neither why not ...
4) you remain with a sqrt(-1) issue, either because some of your formulas are complex, or because during the solution something goes negative. This often happens with the diffusion solution in time stepping when the ratio between the time steps, and the mesh density are not sufficient to resolve steep diffusive thermal gradients, have you checked the diffusivity of your material nad the related mesh size and thermal properties.
FEM as any "sampling" solver scheme needs to respect the traditional sampling criteria: at least 3 samples across a gradient to resolve it (à la Nyquist for time series)
So for me "nothing is wrong with ALE" ;)
--
Good luck
Ivar
for me 2 have at least two issues
1) you coordinate transform is on the spatial frame (dependent on the solution), I expect it to be on the Reference frame (fixed in space), then it's strangely orientated, but as you seem to use only "r" its probably OK)
2) then your Gaussian beam has an amplitude of >1:1E11 that is a lot, do not forget that the binary representation of number has a limited resolution hence a limited dynamic check the value of "eps" and compare to 1, and do not forget that you need sqrt() and ^2 to resolve your physics
3) I fear that you extrusion coupling is not working as expected, but I'm not sure, neither why not ...
4) you remain with a sqrt(-1) issue, either because some of your formulas are complex, or because during the solution something goes negative. This often happens with the diffusion solution in time stepping when the ratio between the time steps, and the mesh density are not sufficient to resolve steep diffusive thermal gradients, have you checked the diffusivity of your material nad the related mesh size and thermal properties.
FEM as any "sampling" solver scheme needs to respect the traditional sampling criteria: at least 3 samples across a gradient to resolve it (à la Nyquist for time series)
So for me "nothing is wrong with ALE" ;)
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hello
I´m trying to simulate shrinkage of a sphere.
I´m using 2D axisymmetric, heat transfer, mass transfer and ALE.
In Moving mesh I´m using "prescribe mesh displacement".
I set up dr and dz as follow:
dr=dz=(c-c0)/c0*R
where c is the concentration that changes during the time because of the mass transfer, c0 is the initial concentration and it is constant and R is the initial radius of my sphere.
the coordinate system that I consider are the global coordinate system.
In the mesh I´m using a boundary layer.
My simulation stops before reaching 1% of progres and I get this error:
Recoverable error in residual, but solver could not recover.
Time : 0.1259350191973199
Last time step is not converged.
and a lot of warnings like:
Inverted mesh element near coordinates (1.05095e-005, -0.00498949).
can you help me to figure out how to use ALE?
thank you really much
Loredana
I´m trying to simulate shrinkage of a sphere.
I´m using 2D axisymmetric, heat transfer, mass transfer and ALE.
In Moving mesh I´m using "prescribe mesh displacement".
I set up dr and dz as follow:
dr=dz=(c-c0)/c0*R
where c is the concentration that changes during the time because of the mass transfer, c0 is the initial concentration and it is constant and R is the initial radius of my sphere.
the coordinate system that I consider are the global coordinate system.
In the mesh I´m using a boundary layer.
My simulation stops before reaching 1% of progres and I get this error:
Recoverable error in residual, but solver could not recover.
Time : 0.1259350191973199
Last time step is not converged.
and a lot of warnings like:
Inverted mesh element near coordinates (1.05095e-005, -0.00498949).
can you help me to figure out how to use ALE?
thank you really much
Loredana
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hello
I´m trying to simulate shrinkage of a sphere.
I´m using 2D axisymmetric, heat transfer, mass transfer and ALE.
In Moving mesh I´m using "prescribe mesh displacement".
I set up dr and dz as follow:
dr=dz=(c-c0)/c0*R
where c is the concentration that changes during the time because of the mass transfer, c0 is the initial concentration and it is constant and R is the initial radius of my sphere.
the coordinate system that I consider are the global coordinate system.
In the mesh I´m using a boundary layer.
My simulation stops before reaching 1% of progres and I get this error:
Recoverable error in residual, but solver could not recover.
Time : 0.1259350191973199
Last time step is not converged.
and a lot of warnings like:
Inverted mesh element near coordinates (1.05095e-005, -0.00498949).
can you help me to figure out how to use ALE?
thank you really much
Loredana
Hello, Loredana.
I read about your problem of ALE, I encountered some similar problem rencently.
Have you solved it yet?
Best Regards!
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
check the doc on ALE and the different "spring methods" it cn help, also if you have large deformations, you might need to consider remeshing while solving
--
Good luck
Ivar
check the doc on ALE and the different "spring methods" it cn help, also if you have large deformations, you might need to consider remeshing while solving
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
check the doc on ALE and the different "spring methods" it cn help, also if you have large deformations, you might need to consider remeshing while solving
--
Good luck
Ivar
I appreciate your reply, Ivar!
I'm simulating the dissolving process of a sphere like droplet in another liquid. The droplet it composed of S and P.
S can dissove into the out phase liquid W, while P cannot and stay in the droplet. The droplet surface(S&P-W) is shrinking with time. The problem is a transient one.
The governing equation is Fick's second law. Because of the symmetry, the problem can be describled by a 1 D transient equation.
I set a 1D model with ALE to solve it.The surface moves at a speed of dR/dt. In initial, I encountered minus concentration problem because the sharp concentration gradient near the boundary. Then I found that the Heaviside Function can be apllied to conquer this. However , another problem comes. The concentration should be in the range[0,1], I got jumping profile, wich shows that Cs sometimes higher than 1, which is not the case obviously.
Would please give me some suggestions?
Appreciation!
Best Regards!
Tsong Gau
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
if you use the heaviside function there are diferent ones, some with others without"overshoot", Use rather the "step" function in V4 its (I believe) based on the Heaviside functions, and has no overshoots
--
Good luck
Ivar
if you use the heaviside function there are diferent ones, some with others without"overshoot", Use rather the "step" function in V4 its (I believe) based on the Heaviside functions, and has no overshoots
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
8 years ago
Hi
check the doc on ALE and the different "spring methods" it cn help, also if you have large deformations, you might need to consider remeshing while solving
--
Good luck
Ivar
I appreciate your reply, Ivar!
I'm simulating the dissolving process of a sphere like droplet in another liquid. The droplet it composed of S and P.
S can dissove into the out phase liquid W, while P cannot and stay in the droplet. The droplet surface(S&P-W) is shrinking with time. The problem is a transient one.
The governing equation is Fick's second law. Because of the symmetry, the problem can be describled by a 1 D transient equation.
I set a 1D model with ALE to solve it.The surface moves at a speed of dR/dt. In initial, I encountered minus concentration problem because the sharp concentration gradient near the boundary. Then I found that the Heaviside Function can be apllied to conquer this. However , another problem comes. The concentration should be in the range[0,1], I got jumping profile, wich shows that Cs sometimes higher than 1, which is not the case obviously.
Would please give me some suggestions?
Appreciation!
Best Regards!
Tsong Gau
Hi Tsong Gau,
Recently I have encountered the similar problem as yours. Could you please tell how you solved this problem use the heaviside function? Also, i am building an expanding droplet with 2 components in ALE. Have you met the mass non-conservation? if so, how do you solve it? or we can discuss about it together? thanks in advance !
best regards!
yuhong