Electron beam melting® (EBM) is one of the fastest growing additive manufacturing processes capable of building parts with complex geometries, made predominantly of Ti-alloys. Providing an understanding of the effects of process parameters on the heat distribution in a specimen built by EBM®, could be the preliminary step toward the microstructural and consequently mechanical properties control. Numerical modeling is a useful tool for the optimization of processing parameters, because it decreases the level of required experimentation and significantly saves on time and cost. So far, a few numerical models are developed to investigate the effects of EBM® process parameters on the heat distribution and molten pool geometry. All of the numerical models have ignored the material convection inside the molten pool that affects the real presentation of the temperature distribution and the geometry of molten pool. In this study, a moving electron beam heat source and temperature dependent properties of Ti-6Al-4V were used in order to provide a 3D thermal-fluid flow model of EBM®. The influence of process parameters including electron beam scanning speed, electron beam current, and the powder bed density were studied. Also, the effects of flow convection in temperature distribution and molten pool geometry were investigated by comparing a pure-thermal with the developed thermal-fluid flow model. According to the results, the negative temperature coefficient of surface tension in Ti-6Al-4V was responsible for the formation of an outward flow in the molten pool. Also, results showed that ignoring the material convection inside the molten pool resulted in the formation of a molten pool with narrower width and shorter length, while it had a deeper penetration and higher maximum temperature in the molten pool. Increasing the powder bed density was accompanied with an increase in the thermal conductivity of the powder bed that resulted in a reduction in the molten pool width on the powder bed top surface. Experimental measurements of molten pool width and depth are performed to validate the numerical model.