Experimental Observations of Thermal Bow due to Natural Convection on a Gas Turbine Compressor Rotor Shaft Analogue

Abstract

This study describes the development of an experimental apparatus designed to provide initial validation of numerical simulations of a gas turbine compressor rotor shaft under thermal bow due to natural convection. The experimental analogue represents a simplified model developed by the authors in previous works, designed to represent the basic elements of a compressor rotor shaft, as part of an ongoing parametric study into the influence of compressor design and integration on the onset time, duration, and severity of thermal bow, and the propensity of an engine to suffer from the Newkirk Effect. The semi-enclosed steel shaft, mounted on a support frame, was heated to approximately 600K in a convective oven before being removed and allowed to cool under ambient conditions. Observations of the shaft under natural cooling were made using several experimental techniques, including temperature profiling using thermographic imaging and a thermocouple array, and physical distortion measurement using linear variable displacement transducer (LVDT) probes. The behaviour of the buoyant plume was also observed using background oriented schlieren (BOS). Using these techniques, the natural convection-driven thermal gradient and resultant physical deformation were measured and recorded over a period of 60 minutes. The observed thermal gradients and resultant thermal bow distortions were then compared to a one-way fluid thermal structural interaction (FTSI) 3D conjugate heat transfer (CHT) computational fluid dynamics (CFD) and finite element analysis (FEA) model developed by the authors, in order to validate the numerical model. The behaviour of the thermal plume from the BOS imagery was also used as a qualitative validation method. The temperature measurements and overall cooling rate measured on the experimental model showed good agreement with the numerical predictions (within 1%); however, uncertainties in the initial phase of the experiment led to error in the numerical prediction of the thermal gradient and resultant thermal bow measurements (error of up to 63%). Noting the uncertainties in the experiment, the agreement between numerical and experimental results with respect to the overall cooling rate indicates that the numerical approach being employed as part of the larger parametric study into gas turbine compressor rotor shaft thermal bow is appropriate and valid.