Scaling in the s5_energycurve.i file

[LiuPengPeter]

Hello,

I still couldn't understand the scaling factor and how it works on the residual.

In the s5_energycurve.i file, the scaling=1e+04 for variable c. And in the expression of material, kappa_c1e-27, M/1e-27 and f_loc1e-27.

Could anyone explain?

[jessecarterMOOSE]

1e-27 is for the value of the property, sort of like a unit conversion. It's the same everywhere so it doesn't affect the solution. Probably @dschwen can tell you more about why that value is chosen. The variable scaling affects the residual.

You might want to try the newer automatic scaling feature in moose. You can set automatic_scaling=true in your Executioner block and remove the scaling parameter for the variable.

[LiuPengPeter]

Than you!

[LiuPengPeter]

@dschwen , could you tell me more details about scaling.

1. Why and when does the scaling factor need to be added to variables and material parameters?

2)How the scaling factor influence the residual?
For example (maybe not right), there is a residual R1 whose variable is eta1 and coupled variable is eta2. M is the material parameter.
R1 = M*eta1^2 + eta2 + partial(f)/(eta1) + partial(eta1)/(x).
After add scaling=1e9 to eta1, what will R1 become?
The R1 may be not right, I just want to use it as an example to know hoe the factor influence residual.

3)In the case of s5_energycurve.i file from tutorial of phase filed, the scaling=1e+04 for variable c. And in the expression of material, kappa_c1e-27, M/1e-27 and f_loc1e-27. Could you explain it?

Thank you a lot!

[LiuPengPeter]

I still couldn't understand the scaling factor...Could anyone help?

[aeslaughter]

@lindsayad Just noticed this sitting around. The only information that I can find on our auto scaling is here: https://mooseframework.inl.gov/newsletter/2019_07.html. Do we have more information somewhere else?

[lindsayad]

We have this documentation page: https://mooseframework.inl.gov/source/systems/NonlinearSystemBase.html. @LiuPengPeter can you see whether this helps you in your understanding?

[LiuPengPeter]

Than you all of guys!