Thank you for your comment.

When you do a regular parametric sweep, COMSOL solves the problem for each parameter from scratch. However, if you need to use the solution from a previous parameter value as a trial solution for the current parameter in a nonlinear problem then you should use an Auxiliary Sweep.

If you click on a study step, you will see on its settings window a section “Study Extensions”. If you expand that section, you will find “Auxiliary sweep”. That is what you use if you are, for example, load ramping and want to use the solution from the previous load step as a trial solution in the current nonlinear iteration.

Hopefully that clears things up. Let us know if you have further questions.

]]>what i understand about parametric sweep is that, comsol evaluate in every single parameter or its combination as if no further evaluation has been made.

example: parameter called ‘p1′ has value 1 2 3 (no geometry, no mesh parameter) using stationary solver. The evaluation starts using p1=1. Afterwards p1=2 is evaluated and comsol could not ‘remember’ that some of the results are already caculated for p1=1. Or the result from p1=1 may only need to be ‘adapted’ to have result of p1=2.

i’m sure that comsol uses same equation, so that compiling equation is not necessary; and same dependent variables.

or i misunderstood about what comsol does with parametric sweep.

best regards

akmal

Thanks for the comment.

You can join solutions from two different time series. If you use a one-all combination then you can join a solution from one time step of one study with all time steps of another study. If you use all-all, you make a combination of solutions having the same indices (first time step from one with first time step from the other study and so on). This is something like “specified combinations” when you are doing multiparameter sweeps. The analog to “all combinations” which makes a Cartesian product is not available. I am guessing you expected that with an all-all joining.

Hopefully that sheds some light.

Thanks again.

Thanks for detailed explanations, there are many useful tricks coming up in the Blog (something I have been missing since that chapter disappeared from the COMSOL journal, at least half a decade ago …

Now, concerning “Joining” sequentially “time series”, is it now possible to join two time series, i.e.

Data Set 1: time 0[s] to 1[s] by step1 and another

Data Set 2: time 2[s] to 10[s] with a step 2

(or even with overlapping time values, being thus sorted out nicely ?

I believe this could be of interest for long runs, that one had interrupted for diverse reasons.

The same for egenfrequency analysis, one with the first N modes, the next with N+1 to M modes ?

Yours sincerely

Ivar