The reduction of fan noise contributions to the engine noise spectrum is an important area in overall engine design as noise restrictions on commercial aircraft are increasingly stringent from one year to the next. The ability to predict this noise furthers understanding of the noise-producing mechanisms of turbofan engines and may serve to develop noise control and suppression techniques. The current state-of-the-art noise prediction model is a two-dimensional linear cascade model with strip theory. A fully three-dimensional annular cascade model is presented, and comparisons are made against a linear cascade for various sound-generating interactions in the fan section of a high-bypass turbofan engine. First, the applicability of a linear cascade model with strip theory in determining the radiated noise from a realistic rotor wake with a stator cascade is presented. Results indicate that the two-dimensional model under-predicts the tonal noise of the annular cascade model. Second, the importance of flow turning, which is accounted for in the annular cascade model but not in the linear cascade model, is investigated. It is shown that the mean flow turning strongly affects the resulting intensities of the propagating acoustic modes, therefore indicating that a two-dimensional theory is insufficient. Third, as broadband noise has become an important component of the noise spectrum of turbofan engines, a broadband noise prediction model is presented for realistic engine cascade geometries. Comparisons between the linear and annular cascade predictions with experiments also indicate that a three-dimensional model is necessary to accurately predict the broadband sound spectrum. Finally, a rotor noise model is developed for the annular cascade theory. The significance of the Coriolis and centrifugal accelerations, which are not present in the linear cascade rotor formulation, are shown to be important for correctly calculating the aerodynamic and aeroacoustic response. Hence, it is determined that linear cascade theory cannot adequately predict the noise spectrum resulting from unsteady disturbances interacting with a rotating cascade.