In general, the fewer your fuzzy sets are, the easier for you to practically implement it, however, more fuzzy sets tend to capture and segregate the inputs to different classifications, thus, will be able to give you a response with higher resolution.  

The more membership functions you have, the more rules.  If you want to implement this in a dsPIC30f, there will be a longer if-else implementation.  You may also consider if you will be using linear or non-linear membership functions for this. case.

If you are working with Takagi Sugeno fuzzy models for the objective of stabilizing a complex system based on say, Lyapunov stability principles, it is generally advisable to have very few rules because the more the number of fuzzy rules, the more difficult it could be to get a common Lyapunov function whose derivatives could guarantee asymptotic stability. On the other hand, working with fuzzy sets for non-control objectives could require more number of fuzzy rules depending on the number of fuzzy sets. This is especially true when you are using the Mamdani fuzzy models. Also, arbitrary increase in fuzzy rules will lead to more implementation cost, less efficiency and punishing tuning routines.

I agree with all answers posted. However, you could use a higher number of fuzzy sets using a table based controller, a two dimensional look-up table. You can read about it in: http://www.iaster.com/uploadfolder/9DevelopmentofFuzzyLogicBasedV-FSpeedControlSystemforInductionMotorGM25Sept13Copy/9Development%20of%20Fuzzy%20Logic%20Based%20V-F%20Speed%20Control%20System%20for%20Induction%20Motor%20GM25Sept13%20Copy.pdf

Hi all 

please help as you can 

i want some articles relate to stress in air traffic controller   

Increasing Memberships causes to have an accurate and high quality fuzzy system

but it depends on an expert person's knowledge. maybe you need to have 1 membership or more...