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.
Time-scaling coefficient vs time step
Posted Mar 25, 2010, 7:40 a.m. EDT 2 Replies
Please login with a confirmed email address before reporting spam
Hi everybody,
I am trying to understand the relationship between the Time-scaling coefficient and the time step in transient analysis.
In the manual, the only thing written for the Time-scaling coefficient: " This coefficient is normally 1. If desired, you can change the time scale, for example to minutes by setting it to 1/60"
This means for an analysis in the ms or the micro-second, this coefficient has to be equal to 1000 or 1e6 respectively.
I am wondering if there is any relation between this coefficient and the time step. Logically it should be in the same scale as the chosen time step. But what about if it is smaller than the time step?
Does anybody have an experience to share in this point?
Your comments are welcome,
Cheers,
I am trying to understand the relationship between the Time-scaling coefficient and the time step in transient analysis.
In the manual, the only thing written for the Time-scaling coefficient: " This coefficient is normally 1. If desired, you can change the time scale, for example to minutes by setting it to 1/60"
This means for an analysis in the ms or the micro-second, this coefficient has to be equal to 1000 or 1e6 respectively.
I am wondering if there is any relation between this coefficient and the time step. Logically it should be in the same scale as the chosen time step. But what about if it is smaller than the time step?
Does anybody have an experience to share in this point?
Your comments are welcome,
Cheers,
2 Replies Last Post Apr 7, 2010, 11:57 a.m. EDT
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hi
I have seen it appear here and there in the equations in the documentation, for me its just a "scale" change issue as by default COMSOL works in seconds I beleive and oten you want to express items in rpm or 1/min, including in the settings to avoid using too large numbers. I must admitt I have never used it.
But I agree, one have to know where it is used to not correct for it twice. The easiest is probably to check the physics equations variables of you application mode
Good luck, and pls keep us informed if you progress, always usefull to know ;)
Ivar
I have seen it appear here and there in the equations in the documentation, for me its just a "scale" change issue as by default COMSOL works in seconds I beleive and oten you want to express items in rpm or 1/min, including in the settings to avoid using too large numbers. I must admitt I have never used it.
But I agree, one have to know where it is used to not correct for it twice. The easiest is probably to check the physics equations variables of you application mode
Good luck, and pls keep us informed if you progress, always usefull to know ;)
Ivar
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
As Ivar points out, this is just a factor that can be mutiplied to the PDE. You can thereby change the unit of the internal time t.
Example:
Suppose you have simple Fickean diffusion
dc/ct + div(-D*grad(c)) = R
You want to compute the problem with days as the internal time. Substitute t'=t/SF where t' is the new time in days, and SF is a scale factor 1/(60*60*24). The time differentiation operator d/dt will then be SF*d/dt. The scaled equation will be
SF*dc/ct + div(-D*grad(c)) = R
You have to be careful if t appears in D or R, in which case you need to compensate those terms too, maybe with other scale factors. Also, if you have more than one time depdendent equation, all need to be scaled the same way.
In every application mode that uses this feature it should be noted in user interface subdomain settings where the scale factors appear. One example is the Nernst-Planck convection diffusion in the Chemical Engineering Module, another is Richard's equation in the Earth Science module.
Niklas Rom, COMSOL
Example:
Suppose you have simple Fickean diffusion
dc/ct + div(-D*grad(c)) = R
You want to compute the problem with days as the internal time. Substitute t'=t/SF where t' is the new time in days, and SF is a scale factor 1/(60*60*24). The time differentiation operator d/dt will then be SF*d/dt. The scaled equation will be
SF*dc/ct + div(-D*grad(c)) = R
You have to be careful if t appears in D or R, in which case you need to compensate those terms too, maybe with other scale factors. Also, if you have more than one time depdendent equation, all need to be scaled the same way.
In every application mode that uses this feature it should be noted in user interface subdomain settings where the scale factors appear. One example is the Nernst-Planck convection diffusion in the Chemical Engineering Module, another is Richard's equation in the Earth Science module.
Niklas Rom, COMSOL