Surrogate model and Latin Hypercube Sampling

[L3munoz]

Hi everybody,

this is my first post and I hope someone could help me...I'm interested in developing a surrogate model or metamodel for a computer experiment planning. The idea is to use a Response Surface Model because of its simplicity, and in future to use a Kriging model since it seems it ¡s better for computer experiments (as far as I understood...) The point is that to get the metamodel a sampling plan is required. At first I thought on applying a classical DOE but then i read a modern one is preferable in these cases. So a Latin Hypercube Sampling Design is proposed as a solution.

My question is if both (Latin Hypercube and RSM) are combinable or Latin Hypercube should be used with Kriging...I am a mechanical engineer and my statistics background is not so rich, so sorry if its a thumb question but...

Finally, if it is possible to use both, the sampling size of the LHS is determined by the order and number of parameters of the RSM?

Thanks in advance, any help is welcome!!

L3munoz.

[L3munoz]

Anyone can help me please!!

[Link]

I'd like to help, but I am not familiar with the methods you've asked about. I cannot remember if we have any statisticians on here in the engineering field either. Anyone out there???

[L3munoz]

Thanks Link for at least trying to help me. Let's hope someone has more idea and can help us!

L3munoz.