Learning Forum

[Gisel]

Hello,

I am new to ANSYS. I know how to set up and run a parametric study by manually typing in the values in the Table of Design Points inside ANSYS Workbench, that I want it to calculate.

Now, I want to run a parametric study for a panel with different dimensions for thickness (t), length (l), and width (w), to compare the results.

For example, let's say I have the following dimensions for w, l, and t:

t = 0.04", 0.4", 0.5", 0.1", 0.25"

l = 10", 20", 30", 40"

w = 5", 10", 15"

I can run a parametric study with the following combinations for t, l, and w:

1. t = 0.04, l = 10, w = 5
2. t = 0.04, l = 10, w = 10
3. t = 0.04, l = 10, w = 15
4. t = 0.04, l = 20, w = 5
5. t = 0.04, l = 20, w = 10 ......... and so on.

Using the Combinations Formula (from Statistics) where order doesn't matter, there would be 3 X 4 X 5 = 60 different combinations that I can run in the parametric study. This would be too many to input by hand in the Table of Design Points in ANSYS Workbench. Is there a way to make ANSYS automate the 60 combinations using the given values for each parameter (i.e., for t, l, and w)?

I am grateful for any help you can offer.

Thank you,

Gisel

[peteroznewman]

I have a statistics program called JMP that will create those 60 rows for me. Then I can copy and paste the rows into the Table of Design Points. Minitab also has this capability. Unfortunately, JMP and Minitab are expensive programs. If you are skilled at Python, there is a free DOE generator. https://tirthajyoti.github.io/DOE/DOE.html#PythonOpenSource A free statistics program is R and there is a DOE generator for that as well. Without those, you could make the 60 rows in Excel using copy paste a few times.
The 60 rows is called a Full Factorial Design of Experiments (DOE). If the model runs quickly, then 60 rows is not prohibitive. Some models might take 1 hour to run a single row. In that case, 60 hours becomes prohibitive. This is where ANSYS provides the Design Exploration tools. One is called Response Surface. This module selects points in a DOE to cover the ranges for t (0.04-0.5), l (10-40) and w (5-15) then computes a statistical fit to the responses but only evaluates 15 points (or less) to cover the entire space.
When you click on Design of Experiments, you define the parameter limits:
Then you click the Preview and a DOE table is automatically generated. When you click Update, the solver will run through these 15 rows.
In the end, it will plot the response, in my case Frequency, vs any two variables to show the surface.

[Gisel]

Thank you, Peter. This is very helpful.
I may have to run simulations with many more than just 60 combinations and for other parameters as well, not just for the dimensions. Can APDL be used to accomplish this? That is, can APDL generate the combinations and evaluate all the design points? I don't know much about APDL; I've been reading up on it, but my question is: is this the route I should investigate further, that is, APDL, to accomplish this?
Thank you again.

[peteroznewman]

APDL is a programming language, so you can do anything you want if you know the language and have "learned to code". I haven't learned the language and generally avoid programming, so I personally would go the Design Exploration route in Workbench.

[Gisel]

Thank you so much, Peter. This clarifies everything.

[Gary Wang]

Hi Gisel and Peter,

Have you successed to use parametric by APDL? I am having the same issue. I have realized multiple runs using Ansys Scripting, but its suquential calculating can't get advantage of Ansys parallel simulation of Parametric(I suppose). So I am trying to 1) run my APDL codes within Ansys parametric or 2) leverage parallel simulation in my codes. Can you provide any help?

link to my posts: /forum/forums/topic/how-to-do-parametric-simulation-with-scripting/

Thanks

[peteroznewman]

Hi Gary, I don't have anything to add to the replies you recieved in your post.