Material property heterogeneity can be introduced into composite structures during the manufacturing process. The term heterogeneity is used in this dissertation to refer to structures whose material properties are a function of position. This investigation examines different analysis techniques for predicting the effect of property gradients on the structural performance of curved composite beams. Microstructural changes during two different manufacturing processes are evaluated. Diaphragm sheet forming is used to produce curved channel beams and a two step stretch forming process is used to produce curved J-beams. A long discontinuous fiber thermoplastic material system is used for both processes. The material heterogeneity is introduced in the form of fiber realignment and spreading which can cause property gradients in both the radial and tangential directions. Two methods are used to determine the effect of heterogeneity on the state of stress in curved beams. The first method uses a separate two-dimensional elasticity solution for the stresses in the flanges and web sections of the beam. Superposition is used to combine these solutions to determine the state of stress in a curved beam of any general cross-section. This analysis shows that radial heterogeneity only has a significant effect on beams with a small average radius to depth ratio. The second stress analysis method uses the Principle of Minimum Potential Energy to determine the effect of a more general state of heterogeneity. Results show that tangential heterogeneity can have a significant effect on the state of stress in curved beams as well as other structures. The tensile failure of notched curved beams is predicted with a Damage Zone Model which uses an iterative finite element analysis. This model incorporates the accumulation of damage at the area of stress concentration into the failure prediction. Curved J-beams axe tested in a combination of axial load and bending. Load versus strain curves are generated with strain gages located at several positions on the beams. This data is correlated with predictions from the stress analysis models.