A vibration based computational framework for damage identification of composite cylindrical parts, produced on a spinning axis by winded carbon fibers, cascaded on specified number of plies, in various angles and directions, was presented in this work. First, a discrete FE model of the examined structure is developed, by consecutive shell and solid elements, simulating each carbon fiber ply and resin matrix. Focusing on the updating methodology, coupled with robust, accurate and efficient finite element analysis software, the linear and non-linear behavior of the composite parts was examined under various load conditions followed by equivalent experimental trials, in order to classify the material properties (isotropic, orthotropic, anisotropic) and develop a high-fidelity FE model. This is achieved through combining modal residuals, that include the lowest identified modal frequencies and mode shapes, with response residuals, that include shape and amplitude correlation coefficients considering measured and analytical frequency response functions and time-histories of strains and accelerations. Single objective structural identification strategies without the need of sub-structuring methods, are used for estimating the parameters (material properties in each deformation plane) of the finite element model, based on minimizing the deviations between the experimental and analytical dynamic characteristics. A stochastic optimization evolution strategy is applied in parallel computing, to solve the single-objective optimization problem, arising from combining the above residuals. The effect of model error, finite element model parameterization, number of measured modes and number of mode shape components on the optimal models along with and their variability, are examined.