A theoretical approach to analyze robustness of feedback linearization is developed. Specifically, sufficient conditions for boundedncss and convergence of the system trajectories are found when feedback linearization based on a nominal mathematical model is applied to an uncertain real plant, which may have parametric and structural uncertainties. The developed approach does not require the restrictive matching condition, a global Lipschitz condition, nor the same equilibrium point for mathematical model and real plant for all possible model-plant mismatch. An example of an unstable chemical reaction in a CSTR demonstrates the utility of the developed approach in analyzing robustness of feedback linearization for nonlinear chemical processes.