鍋爐設計外文翻譯---邊緣冷卻的不穩(wěn)定效果

1、<p><b>　　附錄</b></p><p>　　附錄A 英文資料原文</p><p>　　Unsteady Effects on Trailing Edge Cooling</p><p>　　Mechanical Engineering Department, Stanford University, Stanford, C

2、A </p><p>　　(Received: January 21, 2003; revised: November 8, 2004)</p><p>　　It is shown how natural and forced unsteadiness play a major role in turbine blade trailing edge cooling flows. Reyno

3、lds averaged simulations are presented for a surface jet in coflow, resembling the geometry of the pressure side breakout on a turbine blade. Steady computations show very effective cooling; however, when natural—or even

4、 moreso, forced—unsteadiness is allowed, the adiabatic effectiveness decreases substantially. Streamwise vortices in the mean flow are found to be the cause of the</p><p>　　Introduction</p><p>　

5、　The trailing edges of high-pressure turbine blades are subjected to substantial heat loads. For this reason cooling air is blown from breakouts on the pressure side, jetting toward the trailing edge. A computational ana

6、lysis has tended to significantly overestimate the cooling effectiveness of these jets. Indeed, adiabatic effectiveness, , is found to be nearly 1 to the trailing edge; at least, that is so when the predictions in questi

7、on are steady, Reynolds-averaged (RANS) computations. Unfortu</p><p>　　In the present paper, we describe unsteady RANS computations of a flow that is representative of the pressure-side, trailing edge. Some

8、interesting phenomenology is observed. It is this, not applied prediction methods, that is the subject of this article. We find that natural unsteadiness does arise, due to three-dimensional vortex shedding from the uppe

9、r lip of the breakout (Fig. 2, later). This mean flow unsteadiness causes some extra mixing, and causes the time-averaged to decrease noticeably</p><p>　　Figure 2. </p><p>　　The rationale for un

10、steady RANS is sometimes a cause of confusion. There is no inconsistency between representing turbulent mixing by a statistical closure, while computing an unsteady mean flow. In the presence of coherent, periodic unstea

11、diness, the energy spectrum will look like Fig. 1. Mixing due to the broadband portion of the spectrum is represented by the closure model. The spike is due to mean flow unsteadiness. This must be computed by an unsteady

12、 simulation. It is a source of additiona</p><p>　　Figure 1. </p><p>　　Holloway et al. have previously suggested a role of unsteadiness in the pressure-side bleed problem. Indeed, the present is

13、a follow-on to their study, and is motivated by the same experiments. Those experiments are described in Holloway et al.. The papers by Holloway et al. appear be the only previous computational studies of coherent unstea

14、diness in external trailing edge film cooling. Computations addressing the passages internal to the trailing edge are discussed in Rigby and Bunker. A recen</p><p>　　Computations</p><p>　　The co

15、mmercial code, CFX, was used for the present simulations. Second-order time stepping must be used for this code to capture the coherent unsteadiness. With that switched on, we conducted a number of grid- and time-step re

16、finements to be convinced that the observed unsteadiness is not a numerical artifact. In fact, we ran a few simulations with a different code, Star-CD, with similar results. Hence, the numerical accuracy appears to be su

17、fficient for the task at hand. </p><p>　　The present computations invoke the SST model, as implemented in CFX. The broad features of these simulations are insensitive to the particulars of the turbulence clo

18、sure; similar results were seen with the two-layer RNG and Chen k– models. </p><p>　　Figure 2 shows the computational domain. It consists of an upstream plenum, a land that channels the flow into jets, and a

19、n external region of coflowing fluid. At the breakout the internal flow exits through a rectangular nozzle. The thickness of the upper lip of the nozzle is equal to the height of the jet. The land protrudes downstream fr

20、om the nozzle, in a wedge shape, to the trailing edge. The lower part of Fig. 2 shows a geometry constructed from four images of the domain. </p><p>　　The flow is from left to right, in two streams: the wall

21、 jet exits from the plenum and is channeled by the land; the external flow enters in the upper portion. Generally, the two streams have different bulk velocities; their ratio, Ujet/Ufree-stream is the blowing ratio, sinc

22、e we consider constant density, incompressible flow. A few compressible simulations showed the same vortical flow components and heat transfer that are described herein. </p><p>　　The geometry is subject to

23、a symmetry condition on the left and the right lateral sides of the domain. This emulates a series of jets, blowing toward the trailing edge. The computational domain contains only one-half of the jet exit; hence the sec

24、ond symmetry condition is at the center of the jet. Computations with a full jet cross section produced very similar results to those shown herein. In the presence of forcing, the flow becomes quite complex. That was the

25、 primary motive for testing the v</p><p>　　The final grid consisted of 0.75 million cells, in a block structured form. The solver treats it as fully unstructured, but block structured gridding produced a smo

26、oth grid, with good resolution near surfaces and in the wake of the upper nozzle lip. A grid refinement study was conducted, with a special focus on the grid blocks in the shedding region. The resolution in those blocks

27、was successively doubled in the streamwise and spanwise direction. The finest grid had 1.25 million cells. Coarsenin</p><p>　　For the time-accurate computations, the time step was adjusted to provide about 5

28、0t per period. In the natural case, no forcing was applied. The flow was allowed to develop a self-sustained unsteadiness. A large number of simulations, not reported herein, were conducted at various blowing ratios. It

29、was found that coherent unsteadiness developed spontaneously for all simulations with a blowing ratio larger than 0.35 (simulations for very low blowing ratios were not performed). For blowing ratios</p><p>

30、<b>　　Results</b></p><p>　　Observations will be summarized for steady, unsteady, and forced simulations. To an extent, we are using a RANS simulation to understand the averaged mixing processes of

31、 the pressure-side cooling jets. </p><p>　　Temperature contours in a vertical section through the mid plane of the nozzle shows how a layer of cool fluid lies next to the wall in the steady flow calculation:

32、 see Fig. 3. The same midplane section through an unsteady computation shows vortex shedding from the upper nozzle lip: see Fig. 4. </p><p>　　Figure 3. Figure 4. </p><p>　　Comparing the temperat

33、ure contours from the steady (Fig. 3) and unsteady solutions (Fig. 4) shows that the mean flow vortices cause substantial additional mixing. However, a layer of cool air persists next to the wall for a distance of about

34、eight jet heights. The cooling effectiveness, </p><p>　　depends only on the adiabatic surface temperature. Despite the enhanced mixing away from the wall, in the unsteady simulation remains near unity until

35、near the trailing edge. </p><p>　　In these incompressible computations, temperature is a passive scalar. The contour levels in the figures could be regarded as ranging from 0 in the coolant stream to 1 in th

36、e gas stream. Dark regions show where the temperature is low. </p><p>　　The plan form in Fig. 4 illustrates this more completely. Hot fluid is seen on top of the land. This is carried over the land, and is n

37、ot cooled by mixing with the jet; but the lower surface, between the lands, remains near the jet temperature to the trailing edge. Hot fluid begins to impinge near the vertical walls of the land. </p><p>　　T

38、ime histories of temperature near the lower wall show the strict periodicity. This demonstrates that the computation has converged to a limit cycle. Spectra contain a sharp peak at a Strouhal number of 0.2 based on the n

39、ozzle lip thickness (Fig. 5). In reference to Fig. 1, the spike in Fig. 5 is the coherent unsteadiness; it is resolved as part of the mean flow. The broadband is not simulated; it is represented by the Reynolds-averaged

40、turbulence model. </p><p>　　Figure 5. </p><p>　　The extra mixing due to unsteadiness motivated a further study in which the velocity at the inlet to the plenum was pulsated: </p><p>

41、;　　Although the forcing frequency, f, was varied, the largest and most interesting response was for fH/U=0.2; i.e., forcing with the natural shedding frequency. A was also varied. The value A=0.1 is selected as represent

42、ative of cases where forcing has a pronounced effect. </p><p>　　Figure 6 contains a time history and spectrum for the flow produced by inlet pulsations. Rather curiously, the response contains a strong subha

43、rmonic of the forcing frequency, and even a sub-subharmonic. The period is four times that of the forcing. On close inspection, a very weak subharmonic is seen, even in the natural case of Fig. 5; it becomes quite pronou

44、nced with forcing. The flow structure responsible for the appearance of subharmonics will be discussed below. </p><p>　　Figure 6. </p><p>　　Again, the time history in Fig. 6 shows that we are si

45、mulating a periodic, ensemble-averaged flow. The chaotic, broadband component is represented by the closure model. There is no randomness in the time history; in particular, this is not a turbulent eddy simulation. That

46、point should be emphasized: there is no connection between the present unsteady RANS computations and large eddy simulation (LES). The latter simulates random fields and would have to be phase averaged to extract coheren

47、t unst</p><p>　　The influence of the plenum pulsations on mixing is protrayed in Fig. 7. Mixing now brings heated fluid to the wall a couple of nozzle diameters downstream. The pattern of wall temperature, i

48、n the lower part of Fig. 7, shows a distinct change in the distribution of mixing. The highest temperature now occurs on the midline, between the lands. The warmer fluid is swept down in the central region of the lower w

49、all, leaving a small region next to the lands at the cold temperature. </p><p>　　Figure 7. </p><p>　　The centerline effectivenesses for the three cases of steady, natural unsteadiness, and force

50、d unsteadiness are plotted in Fig. 8 versus the distance between the slot breakout and trailing edge. The forced case shows a significant decline in effectiveness, beginning shortly after the nozzle exit. It was the inte

51、ntion of this simulation to produce mixing that resembled lab tests. The data in Fig. 8 are from Holloway et al. [1]. </p><p>　　Figure 8. </p><p>　　While it is unlikely that simple, plane wave f

52、orcing occurred in the lab, both and the spatial pattern of heating in Fig. 7 very closely mirror those seen in experiments. To repeat a previous disclaimer, this simulation is not being presented as a prediction method.

53、 It illustrates the role that mean flow unsteadiness and three-dimensionality can play in trailing edge coolant flows. </p><p>　　The time-averaged, unsteady midspan temperature fields are compared to the ste

54、ady computation in Fig. 9. The contribution of coherent unsteadiness is enhanced mixing. The steady computation represents mixing by broadband turbulence alone (via the turbulence model). The natural vortex street wafts

55、the mixing layer, spreading the time-averaged temperature field. The evolution is similar to the steady case, but with faster spreading. Forcing produces stronger periodic, steamwise vortices. These cha</p><p&

56、gt;　　Figure 9. </p><p>　　The drastic change in mixing that accompanies unsteadiness warrants explanation. Its origin is in the vortical features that occur in the jet. The following observations are presente

57、d for the purpose of uncovering some of the physical mechanisms that are at work. </p><p>　　The unsteadiness is associated with quite complex flow patterns. The midspan sections (Figs. 4, 7) are misleading i

58、n their simplicity: the flow is highly three-dimensional. The midspan sections have the appearance of shedding from a blunt trailing edge. While it is obvious that a two-dimensional geometry will produce a von Kàrm&

59、#224;n vortex street, it is far from obvious in three dimensions; indeed, three-dimensionality can suppress coherent shedding. In fact, a modification to the present geometry </p><p>　　The question arose as

60、to whether the natural frequency is peculiar to the present geometry. To an extent it is. The present geometry was simplified to a series of rectangular wall jets in coflow by removing the protruding section of the lands

61、. Simulations then converged to steady flow, even though they were computed with time accuracy. Grid- and time-step refinements on the truncated land geometry always converged to steady flow. </p><p>　　These

62、 results are consistent with the observation by Martini and Shultz that coherent unsteadiness was not significant in their geometry—which did not have lands. This might not be surprising. The unsteadiness originates at

63、the upper wall of the breakout. The jets are not the cause of unsteadiness; rather, they break up the spanwise coherence of the flow leaving the upper surface, above the nozzles; i.e., three-dimensionality suppresses uns

64、teadiness. The surprising observation is that the protr</p><p>　　A perspective on the flow complexity is provided by vortex visualization. Figures 10 and 11 show the three-dimensional vortex streets with nat

65、ural and forced unsteadiness. A surface Q>0 is plotted, where Q||2–|S|2, with S and being the rate of strain and rate of rotation tensors. Inside these surfaces, the rate of rotation is larger than the rate of strain:

66、 that is the sense in which Q detects vortices. Note that a full jet nozzle was created in these figures by reflecting the computational doma</p><p>　　figure 10 . figure 11</p><p>

67、;　　The natural unsteadiness (Fig. 10) takes the form of vortex tubes, with strong three-dimensionality only occurring near the vertical walls of the lands. The connection to a two-dimensional, von Kàrmàn street

68、 is apparent. The three-dimensionality due to the wall jets is not disruptive of shedding. </p><p>　　A horseshoe vortex wraps around the junctions between the upstream edge of the land and the upper and lowe

69、r walls in the plenum portion of Fig. 2. This vortex can be seen exiting the jet at the bottom of Fig. 10. These horseshoe vortices may contribute to the distortion of the shed vortices near the end walls in the case of

70、natural unsteadiness. However, they do not seem to make a major contribution to mixing beyond the nozzle exit. </p><p>　　The forced case, in Fig. 11, is more intriguing. The shedding now breaks into vortex l

71、oops. The subharmonic component in the spectrum seems to be due to the loops appearing alternatively at the sides and in the middle of the slot. This can be seen by comparing the figures at the left and the right; they a

72、re one natural period apart in time. Again, it must be emphasized that, as complex as the flow may seem to be, it repeats periodically; this is not a LES. </p><p>　　The mean flow vortices now have a stronger

73、 streamwise component than in the unforced case. Streamwise vorticity is known to greatly enhance mixing in shear layers. The surface temperature patterns in Figs. 4 and 7 reflect the role of streamwise vortices. In Fig.

74、 4, higher wall temperatures occur near the lands because that is where the streamwise vortices occur. In the forced case, Fig. 7, vortex loops in the middle of the flow result in the higher wall temperatures. </p>

75、<p>　　These simulations raise the intriguing possibility of reducing mixing and improving cooling via control of the unsteadiness. Because the mean flow unsteadiness is at issue, this does not require suppression

76、of turbulence; the broadband, turbulent component does not mix the hot stream to the wall. Passive devices might be able to break the coherence and suppress mixing. We have seen that modifications to the land geometry do

77、wnstream of the nozzle breakout can have this effect. </p><p>　　附錄B 英文資料翻譯</p><p>　　邊緣冷卻的不穩(wěn)定效果</p><p>　　機械工程部，斯坦福大學，史丹福,加州 9</p><p>　　(收到: 2003 年1月21日; 校訂: 2004 年11月8日

78、)</p><p>　　資訊科技顯示,自然不穩(wěn)定性和強制不穩(wěn)定性在渦輪機輪葉的邊緣冷卻流體中,起到了主要的作用。在表面噴射流上做雷諾數(shù)的平均模擬，就像在渦輪機輪葉上做幾何學上的壓力邊沿突破。穩(wěn)定性計算可以得到效果較好的冷卻，然而,當實驗允許自然的或更自然的,或強制的不穩(wěn)定性計算時，它的絕熱效果就會顯著減少。人們發(fā)現(xiàn)，中間流的旋渦是熱傳遞增加的影響因素。</p><p><b>　

79、　介紹</b></p><p>　　高壓渦輪機輪葉邊緣承載著主要的熱負荷。由于這個原因，冷空氣從邊沿上的突破口吹出,并一直在邊緣噴射。計算分析容易對這些噴射流的冷卻效果作夸大的評價。實際上,熱交換率η在邊緣處接近1;至少在用穩(wěn)定性計算平均雷諾數(shù)時是這樣的。遺憾的是，實驗室測試表明：熱交換率在大約四個噴嘴直徑之后開始降低,而且可能在邊緣的附近大約落到0.5。相關資料表明，在預測值和觀測值之間的誤差可能是

80、由于不穩(wěn)定性引起的。 </p><p>　　在目前的論文中,我們描述流體的不穩(wěn)定性雷諾數(shù)計算的說的是邊沿壓力。觀察一些有趣的現(xiàn)象發(fā)現(xiàn)，它不能應用于預測方法,而只是這篇文章的主題。我們發(fā)現(xiàn)自然的不穩(wěn)定性確實存在，那是因為從上沿口中流出的三維旋渦。 (如下所示，圖2) 這就意味，中間流的不穩(wěn)定性引起一些額外的混合，而且導致平均時間的η顯著減少在到1以下;然而，它似乎并不像實驗室中所期待的一樣降低。上游送氣通風脈動的增

81、加會引起更大的η的降低。然后，絕熱熱交換率模仿實驗室觀察。然而，至今還不是很清楚，脈沖和實驗室中所做的實驗有什么相似之處;因此，他們在這里只是簡單地做強制方面的研究。已經(jīng)發(fā)現(xiàn)，中間的旋渦結構在周期性的強制力下有較大的改變?？臻g旋渦變得更加立體,形成回流,回流就是混合增強的原因。 </p><p><b>　　圖2 </b></p><p>　　不穩(wěn)定平均雷諾數(shù)的根本原

82、因很多時候就是造成混亂的因素。當計算一個不穩(wěn)定中間流時，在表述紊流混合時時一致的。在互相密合時，周期性的不穩(wěn)定性能量波譜看起來如圖1所示。由于波譜的寬頻部分的混合由封閉模型表現(xiàn)。釘狀是由于中間流的不穩(wěn)定性引起的。這必須用不穩(wěn)定模擬計算。高度混合的源點，不是紊流，而是中間流的旋渦。</p><p><b>　　. 圖1</b></p><p>　　Holloway et

83、 al 先前已經(jīng)表明不穩(wěn)定性在壓力邊沿排氣問題所起的作用。的確，現(xiàn)在他們的研究很受追從,而且被相同的實驗所證明。那些實驗在 Holloway et al的論文中都有描述。Holloway et al 的論文，是外邊緣薄膜冷卻不穩(wěn)定性的唯一早先計算研究。在 Rigby 和Bunker的文章中，對邊緣的內部通道計算進行討論 。Martini和 Shultz 的一篇最近文章中[描述一個邊緣幾何學的實驗和計算問題,由一排噴射流進行冷卻,沒有平臺

84、。他們發(fā)現(xiàn)不穩(wěn)定性是由在于在噴射流之間的任意混合。然而，不穩(wěn)定性在他們的 CFD 分析中并不是很重要。由于平臺不存在的原因，他們的幾何學實質上和現(xiàn)在的不同。</p><p><b>　　2 計算</b></p><p>　　商業(yè)代碼，CFX,現(xiàn)在已經(jīng)用于模擬。這個代碼必須用二級時間階來尋找互相結合的不穩(wěn)定性。隨著它的打開,我們確認了若干的晶格和時間的安排，使觀察的

85、不穩(wěn)定性不是一個數(shù)字的人工品。事實上，我們用不同的代碼運行一些模擬,如激光唱碟,都是相似的結果。因此，數(shù)字的精度似乎對現(xiàn)在的操作是足夠的。 </p><p>　　現(xiàn)在的計算可以喚起 SST 模型,如 CFX 所實現(xiàn)的相似。這些模擬的廣泛特征對特殊的紊流是不敏感的;相似的結果在雙層RNG和Chen k –模型中可以見到。 </p><p>　　圖 2顯示了計算的定義域。它包含了一個向上游的送

86、氣通風，一個引導流體進噴射流之內的平臺和一個流體的外面區(qū)域。在內部流體的突破口存在一個矩形噴嘴。噴嘴的上噴嘴的厚和噴射流的高度相等。楔形狀的平臺從噴嘴下游凸出,一直到邊緣。圖2表示的是四個假想定義域的幾何尺寸。 </p><p>　　流體是從左往右流的,分成兩股:層噴射流來自送氣通風的出口而且通過平臺引導;外面的流體在上部進入。通常，二股流體有不同的體積流速，它們比值為Ujet/Ufree，即吹制比,這是因為我們

87、考慮流體的密度是持續(xù)的,并且不可壓縮。 在此描述的一些可壓縮的模擬流體表現(xiàn)了相同的旋渦流度分力和熱傳遞。 </p><p>　　幾何學受限于在定義域的左邊和右邊的橫邊沿上的一個對稱狀態(tài)。這模擬一系列的噴射流,向邊緣流動。計算的定義域只含有一半的噴射流出口，因此第二個對稱狀態(tài)在噴射流的中心。一個完全的噴射流的計算斷面和在此顯示的有著相似的結果。 在強制力存在之前，流體變得相當復雜。那是測定對稱假定的有效性的最初原因

88、。 現(xiàn)在的強制力是平面波浪形; 因此它和對稱相符合，但是流體是否像表現(xiàn)出來那樣存在著對稱性，還是難以確定的。因此,我們只能得出圖2所示的幾何結果。根據(jù)自由流速度和噴嘴孔厚度，可得雷諾數(shù)是5×104。 </p><p>　　在一塊結構化模型中，最后的晶格有七十五萬個細胞室。實驗者視它為完全無結構,但是塊結構能產(chǎn)生一個光滑結構化格子,在表面附近能很好地分解，并在上面的噴嘴鉆孔唇緣之后。曾做過一項關于晶格的精

89、密研究, 在研磨臺的隱蔽區(qū)域有個特別的焦點。這些區(qū)域在細流和平面方向中被增加兩倍。最好的晶格有一百二十五萬個細胞室。 以同樣的劣等方式把晶格研磨成大約二十五萬個細胞室的格子將會導致一個精度的損失。從這些研究中設計七十五萬個細胞室的晶格,用來提供晶格的精度。 </p><p>　　對于正確時刻的計算，時間頻率調整成提供大約每段時間 50個 t標識。在自然狀態(tài)下，沒有應用強制力。這樣流體就可以發(fā)展自我持續(xù)的不穩(wěn)定性。

90、在這里沒有提到的大量的模擬,運用各種不同的吹制比。資訊科技發(fā)現(xiàn)，互相結合的不穩(wěn)定性自然地為所有模擬實驗提供了一個大于 0.35 的吹制比。(模擬對于非常低的吹制比不能運行) 。由于吹制比超過 1.5,不穩(wěn)定性降下來，流體變成穩(wěn)定。為了要保證計算結果的正確性,在 Holloway et al 描述的打井機上做過該實驗，只是表明了互相結合的不穩(wěn)定性是否會發(fā)生。結果是正確的:測試出頻率符合到Strouhal 數(shù)，大約為0.2。 </p&

91、gt;<p><b>　　3 結果</b></p><p>　　科學家將會描述出關于定態(tài)，不穩(wěn)定和強制式的模擬情況。從某種意義上來說，我們正在使用平均雷諾數(shù)模擬，以了解壓力邊沿冷卻噴射流的平均混合過程。 </p><p>　　噴嘴的中間平面垂直截面的溫度表明了靠邊界的層流冷流體是進行恒流計算的方法，見圖3。相同中間截面的不穩(wěn)定計算表示從旋渦上面的噴嘴

92、孔流出的情況，見圖 4 。 </p><p>　　. 圖3 圖4 </p><p>　　比較溫度從穩(wěn)定狀態(tài)到 (如圖 3)不穩(wěn)定狀態(tài)(如圖 4)，我們可以看到中間流旋渦會引起本質的混合。然而，冷空氣層一直存在直到大約八個噴射流高度。 冷卻的熱交換率</p><p>　　它只取決于絕熱層表面溫度。 盡管從層開始一直在加強混合,在不

93、穩(wěn)定模擬中的η直到邊緣附近仍然保持一致。</p><p>　　在這些不能壓縮的計算中，溫度是一個應變量。圖中的曲線可以認為是從0的冷卻流到1的氣流。較暗的區(qū)域表示溫度低的地方。 </p><p>　　圖4的平面圖模型更加充分地舉例說明了這個原理。 熱流體可以在平臺頂上看到。它來自于平臺，但不能于噴射流混合。但是在平臺的低表面，一直到邊緣仍然維持在噴射溫度。熱流開始在平臺的垂直面附近沖擊。&

94、lt;/p><p>　　在較低層的溫度顯示出嚴格的周期性。這就說明計算就在一個有限的循環(huán)之內。 </p><p>　　頻譜在以噴嘴為基礎鉆孔唇緣厚 (圖 5) 的一個 0.2 的 Strouhal 數(shù)目含有一個最高值。參照圖1，圖 5 的道釘是互相密合著的不穩(wěn)定性;它作為中間流的一部分處理。寬頻不能模擬; 它用平均雷諾數(shù)的紊流模型表現(xiàn)。</p><p><b>

95、;　　圖5</b></p><p>　　別的混合是因為不穩(wěn)定性引發(fā)了更深的研究，在這項研究中，從入口到送氣通風口的速度在變動：</p><p>　　雖然力壓頻率ωf會不斷改變，但是最顯著的和最有趣的反應是公式ωfH/U=0.2;也就是自然的頻率。A也被會改變。數(shù)值A=0.1選擇為代表強制發(fā)生的情況。</p><p>　　圖6表示進水口脈動產(chǎn)生的流體的時間

96、周期和波譜。有趣的是，在強制力下含有力壓周率的一個強亞諧波,甚至是子亞諧波。 這個周期比強制力存在時的四倍。在接近的檢驗，可以觀察到一個非常弱的亞諧波,甚至在圖5所示的自然情況下，它會因為強制力變得更加顯著。流體的結構就代表了下面將要討論得流度構成有責任的亞諧波的形狀。 </p><p><b>　　. 圖6 </b></p><p>　　另外，圖6所示的時間

97、周期表明了我們正在模擬周期性的、整體平均的流體。紊流, 寬頻分力用封閉模型表現(xiàn)。在時間周期中沒有任意性;尤其，這不是一個紊流的漩渦模擬。應該強調一點: 在不穩(wěn)定平均雷諾數(shù)計算和大的漩渦模擬 (LES)之間不存在聯(lián)系。 后者模擬任意的區(qū)域而且必須是平均吸取互相結合的不穩(wěn)定性的相。這是一個昂貴的計算，包括計算幾百個周期永來獲得相關數(shù)據(jù)。</p><p>　　和送氣通風脈動的混合的影響如圖7所示?，F(xiàn)在的混合帶著熱流體到

98、層的噴嘴直徑下游。層溫的類型，用圖7較低部分表示，表明在混合分配上有明顯的改變。最高的溫度現(xiàn)在在中線上發(fā)生,在平臺之間。較熱的流體在較低的層中央?yún)^(qū)域中撤除下來，只留下在低溫平臺附近的一塊小區(qū)域。</p><p><b>　　. 圖 7</b></p><p>　　圖8中心線上畫出穩(wěn)定態(tài)，自然的不穩(wěn)定性和強制強制的情況在熱效率上有明顯的的減少，從離開噴嘴附近就開

99、始。用這些模擬混合產(chǎn)生實驗數(shù)據(jù)正是原因所在。圖8的數(shù)據(jù)來自 Holloway et al的實驗。 </p><p><b>　　圖 8</b></p><p>　　當它不太可能那么簡單時，平面波浪力就會在實驗室中發(fā)生,兩者都如圖 7 中所示的的 和熱學空間套式非常接近地反映在實驗中見到的。為了要重復事先聲明,這一個模擬沒有呈現(xiàn)成一個預測方法。 資訊科技舉例說明均流