UQLink and random fields

[bsanderse (Benjamin Sanderse)]

Hi UQLab team,

I have a question that is related probably to an earlier question raised here.
I have an external piece of software for which I want to use the UQLink functionality. Suppose I have a couple of random variables that are used to define a random field (e.g. through a KL-expansion). This random field needs to be written as an input file for the software. How can I do this?

Best regards,

Benjamin Sanderse

[moustapha (Maliki Moustapha)]

Hi @bsanderse,

I am not sure I understood your question correctly. By “how can I do this?” Do you mean writing the input file for your software or how to link UQLab to your software through an intermediate input file ?

In any case the way to go is as Bruno suggested in the first reply to the post:

Creation of an input for Random Fields

Hello Thanks for your question about random fields. We have a not-yet-published module for random field discretization that works with UQLab… but it is not yet ready for delivery. Too bad ! In any case, your INPUT object should contain your input standard normal variables that are used to generate realizations of your random field (BTW: don’t use a Gaussian random field for Young’s modulus, but a lognormal one!). Then you should embed the evaluation of the RF realization, followed by the fini…

Can you please have a look and may be from there tell me what is the missing piece to solve your problem?

Cheers,
Moustapha

[bsanderse (Benjamin Sanderse)]

Hi @moustapha,

Thanks for your reply. I read the post from Bruno indeed, and the piece of pseudocode that he gives is exactly what I want to do, but then within the UQLink framework. Is that possible, or should I have some patience for this functionality to become available? Or do you know any workarounds?

For me the key would not be to replace a single random variable in an input file (e.g. ), but instead to write a vector or field into an input file, where this field or vector can be obtained (via a known function) from this single random variable.
I hope this clarifies?

Best,

Benjamin