Abstract

Numerical simulation of two-phase multicomponent flows requires solving continuity, momentum, energy, and transport equations. Typically, these conservation equations are solved for computing the main variables of pressure, enthalpy, velocity, and composition. Variation of thermophysical properties (e.g., density, viscosity, etc.) as functions of the main variables necessitates introducing equations of state (EOS) to the modeling scheme, equating the number of unknowns and equations. The problem arises here as almost all the available EOSs in the literature receive temperature as an input, which is not a main variable. Guessing temperature, as an unknown input, imposes more iterations on the already iterative algorithm of the EOS and increases the computational cost. The primary focus of this study is to provide highly-precise, but fast EOS scheme for calculating two-phase fluid properties using artificial intelligence algorithms. In the first step, a Fugacity-Activity model is implemented to supply a supervised learning algorithm with a large dataset. The provided data are fed into a machine-learning (ML) model called gene expression programming (GEP). The outputs of this GEP model are high-preciseness explicit formulas for non-iterative computing of temperature and equilibrium constants. Testing the proposed GEP equations for 1,000,000 arbitrary sets of inputs revealed high accuracy in predicting desired outputs (e.g., < 0.6% error in calculating temperature). Implementing GEP equations in modeling platforms can result in ∼90% reduction in EOS-related computational cost. This ML-based EOS is a transparent box for computing thermophysical properties of two-phase mixtures containing H2O, CO2, CH4, N2, H2S, NaCl, KCl, CaCl2, and MgCl2.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call