A Prandtl-type eddy viscosity model including the first-order effect of turbulence structure has been developed to deal with curved free-shear flows. The model is generalized to account for the effect of arbitrary cross-section of the jets injected from a various nozzle configurations into a uniform crossflow. The model is implemented as a module of a general purpose finite-element computer code. The finite-element procedures used here follow from a Galerkin type variational formulation with the penalty approximation for pressure in a consistent manner, with which a significant savings in computational time and storage are achieved. In order to simulate complicated 3-D turbulent flow with a restricted computational space and modest mesh, a slip condition is employed to model the wall flow and stress-free conditions are used for the farfield and outflow boundaries. Numerical predictions are performed for three problems: a single circular jet in a crossflow, a single streamwise aligned rectangular (aspect ratio 4) jet in a crossflow, and dual side-by-side rectangular jets in a crossflow, all at a jet-to-crossflow velocity ratio 4, which is important for V/STOL and other applications. The prediction of the mean velocity components of the circular jet case is in excellent agreement with the measured data except for the near wall region. The surface pressure comparison is very good except for the viscous wake region right behind the nozzle due to flow separation. The current pressure prediction is as good as any inviscid solution given by singularity or panel method with empirically tuned jet properties. No mean flowfield comparison is made for the single rectangular jet case due to the lack of available measured data. Surface pressure comparison is consistently very good, especially for the region near the front corners of the nozzle where the large negative peaks appear. The agreement for this case seems to be even better than the circular jet case, and the reason is, as observed in the surface velocity vector plot, the different vortex structure and mixing in the vicinity of the nozzle. For the dual jets case, the surface pressure prediction is still in a very good agreement, and the mean velocity comparison shows better agreement as the mesh is refined. The flowfield is found to be more complicated than the circular jet case due to the jet interaction, and further mesh refinement is needed for the complete resolution of the jet/wake flowfield. However, if the surface pressure prediction is the major concern, as in the V/STOL applications, the current size of computational space along with numerical strategies adopted here can serve that purpose effectively. Finally, the mean velocity and the pressure prediction obtained here for rectangular jet(s) are the first known to this author, and will provide useful information for the 3-D, complex, turbulent, free shear flow computations. / Ph. D.