Thermo-hydraulic analysis of wall bounded flows with supercritical carbon dioxide using direct numerical simulation

Abstract

The power cycle based on supercritical carbon dioxide technologies promises a higher thermal efficiency and a compact plant layout. However, heat transfer and hydraulic characteristics are peculiar in the near-critical region due to the sharp variation of thermophysical properties in a narrow temperature and pressure range. Therefore, this works presents the results of several direct numerical simulations (DNS) of turbulent wall-bounded flow at supercritical pressure. The spatially developing pipe flows are simulated with the low Mach number approximation to characterize the cooling process of supercritical carbon dioxide. The upward and downward flow of carbon dioxide in vertical orientation has been considered. Heat transfer deterioration followed by recovery is observed in the downward flow while enhancement occurs in the upward flow as compared to forced convection. During the heat transfer deterioration, sweep and ejection events are decreased greatly, triggering the reduction in turbulence. The recovery in turbulence is brought by the Q1 and Q3 (also known as outward and inward interaction) events, contrary to the conventional belief about turbulence generation. The turbulence anisotropy of the Reynolds stress tensor showed that the turbulence structure becomes rod-like during the deteriorated heat transfer regime in the downward flow and disc-like for the upward flow. In addition to low Mach number DNS, a framework for using fully-compressible discontinuous Galerkin spectral element method for DNS of supercritical carbon dioxide is presented. A turbulent channel flow is considered to demonstrate the ability of this framework and to observe the effects of Mach number in the supercritical fluid regime. The increase in the Mach number increases the turbulence in the flow for a given Reynolds number. Finally, a computationally light data-driven approach for heat transfer and hydraulic characteristics modeling of supercritical fluids is presented based on the deep neural network. This innovative approach has shown remarkable prediction capabilities.