A New Physics Based Unsteady Transition Model Using the Universal Intermittency Function

Abstract

Abstract The flow inside a gas turbine engine has unique complexities. One of the important characteristics of such flow field is the existence of periodic unsteady wakes, originating from stator–rotor interaction. The unsteady wakes, with their highly vortical core, impinge on the downstream blade surfaces and cause an intermittent transition of the boundary layer from laminar to turbulent. The relative intermittency value corresponding to the wake vortical core and the calm region outside the wake, irrespective of freestream turbulence intensity and wake frequency, exhibits a universal behavior which is best described by the universal intermittency function of Chakka and Schobeiri [1, 2]. This study aims at introducing a new physics-based universal intermittency function that in conjunction with the current turbulence models accurately predicts the unsteady behavior of an intermittent flow. For that reason, a transport equation for turbulence intermittency was proposed based on this function and was integrated into a RANS based solver with k-ω turbulence model. The model was tested for reliability. Experimental aerodynamics and heat transfer measurements conducted at Turbomachinery Performance and Flow research Lab (TPFL) at Texas A&M University, were used as benchmark tests. For experimental measurements, an unsteady linear cascade facility in TPFL was used to produce the periodic unsteady flow condition. Moving wakes, originating from upstream blades, were simulated in this facility by rods attached to two parallel timing belts in front of the turbine blades. Heat transfer measurements along the suction surface were conducted utilizing a specially manufactured blade with an internal heater core, instrumented with liquid crystal. All Measurements and calculations were conducted at Reynolds number of 264,000. The computational results, obtained from implementing the new enhanced intermittency transport equation into the solver, are compared with (a) experimental measurements and (b) with the computational results from RANS that incorporates Langtry-Menter [3, 4] method.