KR391: What are different types of response surfaces available in ANSYS 12.0?


General rules:

`Standard 2nd order polynomial are efficient when the variation of the output is smooth wrt the input parameters

`Kriging proves to be efficient in a large number of cases, even with highly nonlinear responses (such as emag)

`Non-parametric regression is suited for nonlinear responses. It is usually slow to compute

`Neuronal network seem to be successful with highly nonlinear responses. Control over the algorithm is very limited

In practice:

`Start with the 2nd order polynomial. If goodness of fit is weak, switch to Kriging.

`The two other models should be used very carefully





Show Form
No comments yet. Be the first to add a comment!