Prediction of natural convection heat transfer in gas turbines

Abstract

The aim of this work is to improve numerical predictions of natural convection heat transfer inside of gas turbines. It is desirable for engineers to use numerical simulations be able to accurately predict the heat transfer inside of a gas turbine casing during engine shutdown. This will help them to understand and alleviate problems such as engine casing distortion.A natural convection test rig was constructed to provide simulation validation data. The rig represented a section of a large gas turbine casing and operated at the same Rayleigh number as a real engine. The casing rig was able to provide a unique dataset that could not be obtained through real engine measurements.Analysis of the casing rig was performed using a large eddy simulation and RANS simulations. It was found that the baseline RANS simulation did not predict the heat transfer accurately enough for engineering use. The simple gradient diffusion hypothesis (SGDH) turbulent heat flux model used in the baseline simulations was replaced by more advanced generalised gradient diffusion hypothesis (GGDH). A modification to the GGDH model, called GGDH+, was developed to account for buoyancy effects on the turbulent heat flux. The GGDH+ model was then able to give heat transfer predictions comparable to the LES results.