FE models for composite structure are generated just like any other FE model, except that material orientations have to be defined for the elements and appropriate material/property input defined. Typically the fibers and matrix are not modelled separately, but are smeared into effective orthotropic properties at the ply or lamiante level.  Details of material orientation and property input depend on the particular FE code that you are using.
 

There ar emany instances where FEA has benused to model a structure and the most significant thing is when a staement is made along the lines of "The results should eb compared against actual test data"!

Inexperience in modelling complex structures such as composites is a recipe for disaster. FEA is a fabulous tool but result needs to be tempered depending upon what you are using it for in your work.

I would suggest the first port of call would be the FEA software houses and compare how each product models composites. Also ask them what books they would recommend to more fully understand the subject.

Before you use an FEA software, you may want to do some hand calculation. One of the better books which may help you is by Jack Vinson from Univ of Delaware.

Otherwise as SWC has mentioned you do the modelling like aany other problem but need to be careful while defining the orthotropic material. Also one common mistake is that the material is defined properly but the co-ordinate system of the part does not match what you had in mind while defining the material. YOu land up with either a very high or low deflection.

Another quick and dirty way to do an analysis ( for 45 degree plies) is to take the lowest E values and treat the material as isotropic. I have seen it done when we run out of licenses for the software.