Differences in molecular size and shape have long been known to cause difficulties the modeling and simulation of fluid mixture behavior and generally manifest themselves as poor predictions of densities and phase equilibrium, often resulting in the need to regress model parameters to experimental data. A predictive approach to molecular geometry within the Gibbs-Helmholtz Constrained (GHC) framework is proposed. The novel aspects of this work include (1) the use of NTP Monte Carlo simulations coupled with center of mass concepts to determine effective molecular diameters for non-spherical molecules, and (2) the use of effective molecular diameters in the GHC equation to predict phase behavior of mixtures with components that have distinct differences in molecular size and shape. Numerical results for a CO2-alkane, alkane-water and CO2-alkane-water mixtures show that the proposed approach of combining molecular geometry with the GHC equation provides accurate predictions of liquid densities and two- and three-phase equilibrium. (C) 2011 Elsevier Ltd. All rights reserved.