Computationally efficient models for the numerical simulation of thermodynamically complex flows

In this work we construct thermodynamic models for dense gases that can be efficiently implemented in Computational Fluid Dynamics (CFD) codes. Precisely, two strategies are considered. The first one is the development of an analytic EOS of the Span-Wagner type, characterized by an innovative choice of the independent thermodynamic variables, which greatly simplifies its implementation within numerical solvers and drastically reduces computational costs. The second strategy employs neural networks of the multi-layer-perceptron type to perform a regression from available thermodynamic data. Both approaches will be validated a priori, through detailed comparisons with well-known thermodynamic models and experimental data available in the literature, and a posteriori, by means of suitable dense-gas-flow test cases.