1、- 1 - 附录 附录 A 英文资料原文 Unsteady Effects on Trailing Edge Cooling Mechanical Engineering Department, StanfordUniversity, Stanford, CA (Received: January 21, 2003； revised: November 8, 2004) It is shown how natural and forced unsteadiness playa major role in turbine blade trailing edge cooling flows.Rey
2、nolds averaged simulations are presented for a surface jet incoflow, resembling the geometry of the pressure side breakout ona turbine blade. Steady computations show very effective cooling； however,when naturalor even moreso, forcedunsteadiness is allowed, the adiabatic effectivenessdecreases subst
3、antially. Streamwise vortices in the mean flow are foundto be the cause of the increased heat transfer. Introduction The trailing edges ofhigh-pressure turbine blades are subjected to substantial heat loads.For this reason cooling air is blown from breakouts onthe pressure side, jetting toward the t
4、railing edge. A computationalanalysis has tended to significantly overestimate the cooling effectiveness ofthese jets. Indeed, adiabatic effectiveness, , is found to benearly 1 to the trailing edge； at least, that isso when the predictions in question are steady, Reynolds-averaged (RANS)computations
5、. Unfortunately, lab tests show that the effectiveness starts todrop after about four jet nozzle diameters, and might fallto about 0.5 near the trailing edge, at typical blowingratios. It has been suggested that the discrepancy between predictionand observation might be due to coherent unsteadiness
6、. - 2 - Inthe present paper, we describe unsteady RANS computations of aflow that is representative of the pressure-side, trailing edge. Someinteresting phenomenology is observed. It is this, not applied predictionmethods, that is the subject of this article. We findthat natural unsteadiness does ar
7、ise, due to three-dimensional vortex sheddingfrom the upper lip of the breakout (Fig. 2, later).This mean flow unsteadiness causes some extra mixing, and causesthe time-averaged to decrease noticeably below 1； however, itdoes not seem to drop as much as lab testslead one to expect. Pulsations added
8、to the upstream plenumcause a more substantial drop in . Adiabatic effectiveness thenmimics that observed experimentally. It is unclear whether the pulsationshave any analogy to conditions that occur in lab tests；so they are presented here simply as a study inthe effect of forcing. A fascinating cha
9、nge in the meanvortical structure is seen under the imposition of periodic forcing.The shed vortices become more three-dimensional, forming into loops, whichare the cause of greatly enhanced mixing. Figure 2. The rationale forunsteady RANS is sometimes a cause of confusion. There isno inconsistency
10、between representing turbulent mixing by a statistical closure,while computing an unsteady mean flow. In the presenceof coherent, periodic unsteadiness, the energy spectrum will look likeFig. 1. Mixing due to the broadband portion of thespectrum is represented by the closure model. The spike isdue t
11、o mean flow unsteadiness. This must be computed byan unsteady simulation. It is a source of additional mixingmixingthat is not due to turbulence, but rather, to vorticesin the mean flow. - 3 - Figure 1. Holloway et al. have previously suggesteda role of unsteadiness in the pressure-side bleed proble
12、m. Indeed,the present is a follow-on to their study, and ismotivated by the same experiments. Those experiments are described inHolloway et al. The papers by Holloway et al. appear bethe only previous computational studies of coherent unsteadiness in externaltrailing edge film cooling. Computations
13、addressing the passages internal tothe trailing edge are discussed in Rigby and Bunker.A recent article by Martini and Shultz describes experimentsand computations of a trailing edge geometry, cooled by arow of jets, without lands. They found unsteadiness due torandom coalescence between the jets. H
14、owever, unsteadiness was not importantin their CFD analysis. Their geometry differs substantially from thepresent, because of the lands. Computations The commercial code, CFX, was usedfor the present simulations. Second-order time stepping must be usedfor this code to capture the coherent unsteadine
15、ss. With thatswitched on, we conducted a number of grid- and time-steprefinements to be convinced that the observed unsteadiness is nota numerical artifact. In fact, we ran a few simulationswith a different code, Star-CD, with similar results. Hence, thenumerical accuracy appears to be sufficient for the task athand. The present computations invoke the SST model, as implementedin CFX. The broad features of these simulations are insensitiveto the particulars of the
