I would like to couple Persalys with my own code in order to do sensitivity analysis. The outputs i have to deal with are 1D profils. I don’t know how to define 1D outputs in the definition of the coupling model.
Hi,
The answer depends whether you are interested in the dependency between the components of the outputs.
- If the vector output has no particular structure, you can define a “basic” physical model, as is presented in e.g. the documentation of the Python physical model.
- If the vector output has the structure of a 1D field (i.e. a time series or a spatially discretized output), you can define a Field model. This requires to define the physical domain (time or space) and its discretisation (i.e. the location of the nodes). In this case, Persalys can compute statistics which are relevant to the field representation : pointwise intervals, Karhunen-Loève dimension reduction, HDR plot, etc.
In which category does you model fall?
Regards,
Michaël
Thank you for your answer. My model falls on category 2.
Should I indicate the true discretization or is the number of points sufficient?
Sincerely
Anne
Hello,
Welcome to the forum !
I think it is better to provide the real discretization. You can define a regular mesh directly in Persalys but you can also import a text file with the mesh values, it might be easier to use this option if it is not regular.
Regards,
Antoine