Experimental and numerical study of heat transfer on turbine blades.

Abstract

An experimental and numerical study of he aerodynamics and the associated heat transfer on turbine blades, has been carried out as part of the ongoing Armscor Denel aircraft engine maintenance program. The experimental tests were performed using an existing continuous flow cascade test facility at the University of KwaZulu-Natal, Durban. These experimental results were used to validate the two-dimensional numerical results, generated usmg a commercially available Computational Fluid Dynamics (CFD) package, FLUENT. The existing experimental turbine test facility utilises a continuous flow cascade technique where a cooled, instrumented blade is rapidly introduced to the hot-air stream exposing it to the cascade flow. This creates the heat transient required for measurement of the isothermal heat transfer coefficients, using thin-film heat flux gauges. A static pressure test blade is used in conjunction with a scanivalve system, to determine the blade mid-span pressure distribution. This latest research effort requires validation of de Villiers' [2002] results, whilst improving the error discrepancies between the experimental and numerical analyses. Maintenance on the test rig has been performed, including the addition of a new pressure control system to ensure the correct cascade flow conditions and boundary conditions are obtained. Experimental pressure distribution measurements were performed, to validate previous work by de Villiers [2002] and to ensure the correct operation of the test rig. Experimental error was identified in de Villiers' [2002] suction surface pressure distribution, and new experimental pressure results were acquired. Following the essential overhaul of critical rig components, experimental heat transfer tests were performed. The newly restored equipment produced new isothermal heat transfer coefficient results that validated the results of de Villiers' [2002]. Numerous CFD meshing techniques were investigated and implemented in FLUENT, to produce the numerical solution. The pressure correlation proved to be excellent with an average error of 3%. The varying cascade inlet turbulence intensity was identified as a major source of heat transfer error. Implementing this variance into FLUENT, a significant reduction in error was seen. The resulting average heat transfer error measured 12%, a major improvement from 29% error in 2002.