Modeling oxygen and hydrogen combustion under supercritical conditions using computational fluid dynamics

Cheng Huang, Purdue University

Abstract

Study of combustion between liquid oxygen and gas hydrogen under high temperature and high pressure that exceed the thermodynamic critical point of the injected fluid (supercritical condition) is an important topic in understanding the performance of many high pressure devices, including diesel, gas turbine, and liquid rocket engines. High pressure are motivated by increasing the power output, efficiency and specific impulse of combustion engine. This project, as an extension of the shear coaxial injector study using oxygen and hydrogen, is mainly aimed at access the capability of computational fluid dynamics (CFD) simulations in predicting the characteristic of liquid oxygen and gas hydrogen(LOX/GH2) combustion under supercritical condition and how well simulation methods can resolve the large density gradient at reaction interface. The Reynolds-average Navier Stokes (RANS) method is applied to investigate turbulent flow characteristic under stationary state to obtain an understanding of mixing and combustion behaviour of fluid under high pressure and high temperature. Two-dimensional opposing nozzle conguration is modeled to do the simulations to better investigate the mixing and reactions between oxygen and hydrogen by a counter flow diffusion flame. Gas oxygen and gas hydrogen(GOX/GH2) combustion is studied first just to get a basic idea of the kinematic model used in computation and also it is a good reference for understanding the flow characteristic under different operation conditions. During simulations, flow behaviour under different Reynolds number is of primary concern, including the properties of flames (locations, thickness), amplitude of pressure oscillation and computational performance. Simulations are performed both in laminar flow and turbulence flow category to compare effects of turbulent effects in physical phenomenon and also in influence of turbulence model in computation. In terms of LOX/GH2 combustion simulations, results of turbulent flow using RANS come with some issues of failure in mass conservation (100% error observed), unreasonable results in pressure calculation and up-rising tendency of variables' residuals in computational aspect, which indicate an inaccurate results given by the simulations while even in some high Reynolds number cases, a self-generated source flow has formed at combustion interface. Several new variables have been defined and constructed during simulations to detect the source of problems. Analysis of these variables shows that the primary issue comes from the ill-convergence of variables, especially density (ρ) and pressure (p) along combustion interface where large density gradient is located. This finding provides a conspicuous direction to move on to solve the problem. The detached-eddy simulation(DES) is also applied to do the simulation of LOX/GH2 combustion and it gives reasonable interpretation of turbulent flow characteristic. Beside, simulations using DES appear to maintain the mass conservation well inside the computational domain but it does not well predict the flow characteristic in stationary state. Several diagnostic methods has been tried to help solve the mass conservation problem and also to mitigate the sharp density gradient effects along combustion interface. Cold flow simulation result ts mass conservation quite well but bad convergence of pressure still exists, which means mass conservation issue comes from combustion while density gradient effects can spoil the computation of other variables. According to this, modifications both in computational aspects and operation condition aspects have been tried, including decreasing physical time step, introducing preconditioning algorithm, cutting inner iteration number, increasing back pressure and halving grid size. Detailed compare and analysis of these methods show that the first two does not solve the problems, the third one seems to help a little in help conserve mass and with back pressure increased, problems appear to be resolved; reasonable pressure contour is obtained and out-going mass flow rate overlaps with in-going one in stationary state of the flow, but with a halved grid size, even though the density gradient have been well-covered, results seem to be unreasonable due to lack of artificial dissipation and an accumulation of non-linearity of Navier Stokes equations.

Degree

M.S.M.E.

Advisors

Anderson, Purdue University.

Subject Area

Mechanical engineering

COinS