Determining the critical points and coexistence curve of multi-species equilibria using Gaussian Process Regression

We present a method for reconstructing the liquid-vapor coexistence curve and associated location of critical points of multi-species mixtures from experimental or simulation data, using Gaussian Process Regression. The distinguishing feature of the proposed method is that it does not require knowledge of the equation of state, as assumed by the currently prevalent methods for calculating the coexistence properties of mixtures. The proposed technique is validated using synthetic data generated from the Peng-Robinson equation of state, experimental data, as well as Gibbs Ensemble Monte Carlo simulation data. These validation studies show that the proposed method is robust to statistical noise and scarcity of data even in the vicinity of the critical point.