Topic: SARA randomization (fields)

I have some problems trying to use random fields in ATENA / SARA. I first wanted to do a simple test, using an example file task.cc2 (Example file in SARA) before I modeled my geometry (of course not half).

I made some modifications on task.cc2 in ATENA 2D.

1. Added a Material with Random Fields (no file specified)
2. Setting Concrete (already defined in the example file) as the base
3. Assigning the new material to the "concrete" Macro Element"

I saved the file

I started SARA studio (v 2.2.0.7)

1. New Project
2. It was impossible to write a directory like "c://folder/..." But I could do that in the menu "options" later on.
3. Imported the .cc2 file and selected some variables (E, ft and fc) (I used a normal distribution COV  = 0.1)
4. FREET -> LHS --> General data --> 10 samples --> run
5. started the Model Analysis

I get results and the graphs that are expected in that kind of problem. But when I read in a file (_r.cc2) and I want to see the values of the random field in post-process (scalar > .. > random field values) I get a blue color and no variation, where I expect to see three random fields for E, ft and fc. but I only get "Par (1)".

Some other questions

1. How do I have control on the spatial variation characteristics? (correlation-length and covariance function, e.g. exp(- |tau| / theta ))
2. Where do I find the values of the generated fields in ASCII files so that I can replace them with known fields?

Thank you.

Re: SARA randomization (fields)

Dear Pieterjan, in Selector, you have click in the column "Variable" to switch to "Field".

Re: SARA randomization (fields)

Thank you, I had overlooked that setting. I'm now able to generate random fields, but the control over spatial characteristics is still a "black-box".

What is the mathematical meaning of 'dx' and 'dy' (as seen in the picture https://dl.dropbox.com/u/2498198/RF.jpg) and in which manner do the relate to the correlation length theta; e.g.: exponential covariance function.

exp(-|tau|/theta)

Is it possible to user-define a covariance function?

Re: SARA randomization (fields)

Dear Pieterjan, we have forwarded your question to our colleagues in Brno, who are developing the stochastical part of SARA and FReET. I will post the answer as soon as we have it. Please be patient...

Re: SARA randomization (fields)

Parameters dx and dy are the (auto)correlation lengths in directions x and y, respectivelly.

The autocorrelation function has the following form (squared exponential fcion):

C(x,y) =  exp[ - (length_x / dx )^2  - (length_y / dy )^2  ]

Which is the same as formula 2 in paper founde here (pow=2)

http://www.fce.vutbr.cz/STM/vorechovsky.m/papers/8j.pdf

At the moment, there is no posibility to use a different autocorrelation function.