In the production process of low-density polyethylene (LDPE), an important step is the flash separation of monomers and other small molecules from the polymer produced. The process is carried out adiabatically in two stages. To improve the performance of thermodynamic models, it is very important to analyze the use of model binary interaction parameters (BIP) dependent on the phase characteristics for each phase (phase-dependent BIP). In this work the PC-SAFT (perturbed-chain statistical associating fluid theory) equation of state (EOS) is applied to the flash simulation of LDPE industrial separators using eight different resins. The main numerical aspects are examined with emphasis on the optimization strategy for the EOS BIP that explicitly characterizes each phase involved separately. The results demonstrate good predictive behavior. As a result of improved and more consistent modeling, a new strategy for optimized operation can be envisaged for the sequence of separators. (c) 2013 Wiley Periodicals, Inc. J. Appl. Polym. Sci. 130: 2106-2117, 2013