UC Irvine

In recent decades, the escalating global energy consumption has increased interest in thermodynamic power cycles, such as the Organic Rankine Cycle (ORC), which are known for their high waste heat recovery efficacy. The fluids in such power cycles have complicated molecular structures and are typically operated near the saturation vapor line and critical point. These gases exhibit similar behaviors to those investigated by Bethe (1942), who established a thermodynamic property known as the fundamental derivative (Γ), which measures the variation of the sound speed of a gas with pressure change during an isentropic process. Gases with Γ values less than 1 are referred to as Bethe-Zel’dovich-Thompson (BZT) gasesand are commonly used as working fluids in the ORC. BZT gases demonstrate a range of qualitatively distinct phenomena compared to conventional gases, especially when Γ < 0. These include expansion shock waves, double sonic shock waves, compression fans, and expansion shock fans. Despite the lack of experimental evidence to confirm such non-classical gas flow behaviors, the increasing interest in using supercritical heavy gases and academic curiosity warrant further in-depth study. A comprehensive literature review of past studies on the fundamental derivative and corresponding non-classical gas behaviors is conducted.

In order to enhance the prediction accuracy of the thermodynamic properties of dense gases and address the existing gap in the availability of precise and accurate equation of state within the Bethe-Zel’dovich-Thompson (BZT) region, this research investigates the potential of machine learning algorithms’ application in the thermodynamics field. By meticulouslyanalyzing the influence of various hyperparameters, an optimized artificial neural network (ANN) model has been developed and subsequently incorporated into our in-house code, which solves the Navier-Stokes equations using a finite-volume method. The successful implementation of the optimized ANN model to calculate the thermodynamics properties demonstrates the potential to significantly decrease the reliance on traditional real gas equations of state for simulation results. This advancement offers a more robust and reliable approach to predicting thermodynamic properties, thereby contributing to the broader field of thermodynamics research.

Besides the state-of-art technology in the thermodynamics part, this study also employs the Van der Waals model to illustrate the existence of negative fundamental derivative regions for dense gases and numerically investigates the unique phenomena of dense gas flow. A numerical solver based on the Jameson-Schmidt-Turkel scheme for dense gas flow is developed, and various counter-classical gas dynamics flow behaviors of real gases in different Γ regimes are emonstrated through selected cases. The simulation results of dense gas flow across various geometries are presented and analyzed, revealing intricate wave fields.

In conjunction with the traditional density-based solvers, this study also proposes an innovative pressure-based solution methodology for compressible flow, specifically designed to address the stiffness issue encountered near the critical point due to the heightened sensitivity of pressure with respect to density. A comprehensive derivation of the novel approach is presented, emphasizing its theoretical underpinnings and practical applications.