The need to analyze and optimize the large and complex structures, locally and globally, and have knowledge of local and global dynamic, has induced the notion of Dynamic Substructuring. In the course of time several DS techniques are developed from which some are mainly based on experimental routine. They are known as experimental DS and assemble measured and analytically derived substructures in the frequency domain. In order to advance, compare and classify the experimental DS theory and technology a test bed (Ampair 600 wind turbine) is initiated by the substructuring focus group at the Society for Experimental Mechanics. Based on this idea the Engineering Dynamic workgroup at Delft University of Technology has started his program which runs through couple of iterations before the Ampair 600 is completely analyzed. This thesis concentrates on the first iteration in which attempt is done to combine three blades with the hub. The emphasis is on the Frequency Based Substructuring (FBS) in general and Lagrange Multiplier Frequency Based Substructuring (LM FBS) and experimental modeling techniques in specific. First DS is introduced followed by FBS and LM FBS. The LM FBS formulation for both coupling and decoupling are given and typical difficulties associated with experimental DS such as rigid body dynamic, coordinates incompleteness (the lack of Rotational Degrees of Freedom), noise and inconsistency are underlined and some methods are proposed to tackle them or minimize their influences. The notion of interface loading is adopted and applied using a fixture which mass loads and stiffens (rigidifies) the interface. The fixture is a chopped off version of the actual joint. As consequence the joint stiffness and damping are included in the models after removing the fixture by means of LM FBS decoupling technique. To model the Cyclic Symmetric Structures (CSS) two methods are proposed which seems to contribute to consistent and robust experimental modeling. One produces models from measurement data and the other one lumped models. Both construct models which are defined in the local coordinates of the interface nodes. The assembly results of four coupling variants according to LM FBS are given, validated and evaluated. It is shown that using regenerated models, particularly when the structures are too complex, results in wrong prediction of the assembled system. The results of the best two coupling variants are more discussed in details and their eigenmodes up to 185 Hz are indentified and visualized.