Exploration DOEs
Exploration DOEs are useful for getting information about the problem and about the design space. They can serve as the starting point for a subsequent optimization process, or as a database for response surface training, or for checking the response sensitivity of a candidate solution.
Random Sequence spreads points uniformily in design space. It is based on the mathematical theory of random number generation.
Sobol is a deterministic algorithm that mimic the behaviour of the Random Sequence: the aim is again the uniform sampling of the design space. But in this case the clustering effects of random sampling are reduced.
Latin Hypercube - Monte Carlo are two different schemes for design space sampling. Random samplings can conform to several multivariate statistical distributions. The difference between the two is that Latin Hypercube maps better the marginal probability functions, even with a small number of samples.
Cross Validation distributes the designs uniformly in the design space, on the basis of the Kriging algorithm used for the response surfaces. In fact this method estimates the error of the model and then chooses a good new set of points in order to make the response surface more reliable.