Hello, 

I am not sure about the purpose of your study.

For instance, Kriging could be used as a  meta-modeling technique. In the context of  computer simulation, for each set of parameters values x (geometric parameters, material properties,..etc), an output s y= f(x) is computed via an expensive simulation.

To use kriging, one should solve first  several  design points   X  ( with different values) to compute  the outputs Y.  Kriging is then used to compute a meta-model ( called also surrogate-model) to replace the expensive simulation by an approximation  \hat{f}.

 I think that in your case, you should define the parameters  that are likely the most influential on the output (geometric parameters that define the shape).   Then, you have to  generate a design of experiments and  simulate the outputs.  

finally you can use any kriging implementation to compute the surrogate model. 

I hope that this would help you.  

Best regards,

In terms of searching the literature it is mis-leading to say that "kriging is a meta modeling technique", that is not what it is. However there are examples of kriging being used to construct surrogate models.Do a search on Google and you will find a number of relevant papers. In particular see

cdn.intechopen.com/pdfs/30305.pdf

This is not my paper but my work has been cited in this literature

I have edited my  previous answer.  

 I  had said  roughly  that  "kriging is a meta modeling technique" because I have assumed that the question concerns an engineering context where the purpose is the use kriging as a meta-modeling technique.  So I am sorry for the possible confusion and the lack of caution.

Edit:  "Kriging could be used as a meta-modeling technique" 

Thank you for the remark.

Wow.. Thanks for all your answers. It is indeed very enlightening.