Real rotor with multiple disks, ball bearings and flexible bearing supports has been considered in the present work. Reduced order model required for simulations of the dynamic system is developed using multi-objective genetic algorithms by matching its dynamic response with full order Finite Element Method based model. This enables us the possibility of including ball bearing dynamics as well as experimental data in model updating. In a rotating machinery, multiple types of faults can be expected during its service life, such as transverse breathing crack, looseness of bearing supports and unbalance. Rotor crack can be quantified by using the dynamic response of the rotor under coast up conditions in open loop situation using Pattern Search optimization. Wavelet transformations are used to denoise the signal and enhance the estimation accuracy.