when data aren't real and precise we can use fuzzy data. when we we are going deal with quality data (such, bad, good, excellent and ...) fuzzy data are very useful.
Normal Fuzzy set or type-1 fuzzy set can only incorporate the vagueness not the uncertainty in the model. type-n sets are the effort to incorporate the vagueness as well as the uncertainty. It handles the uncertainty involved in assigning membership values to any fuzzy variable in the form of n-dimensional structure.
The degree of membership is an interesting focus here. With this property, we can formulate Rule Base systems that can be useful for classification, prediction of a complex system that is quite difficult for classical sets.
Now a days people are using this for prediction purposes. For example, what is the future value of share market, etc.
If you like the ides of fuzzy control, you might also like the idea of multiple model adaptive control. We use nonparametric models of drug behavior. They have multiple discrete support points, which permit multiple model dosage design for maximally precise dosage regimens. Each discrete support point is like a fuzzy point, but is quantitatively better. I would call multiple model control, which is widely used in the aerospace community, something like fuzzy but not fuzzy. That is the advantage, I think. You might go to www.lapk.org and look around.