自动化 电气自动化毕业论文英文翻译.doc

华 北 电 力 大 学毕 业 设 计（论 文）附 件外 文 文 献 翻 译学 号： 姓 名： 所在院系： 专业班级： 指导教师： 原文标题：Numerical analysis of flame behavior in gas turbine combustors using LES 2008年 月 日外文文献翻译原文Numerical analysis of flame behavior in gas turbine combustors using LESJ. Shinjo, Y. Mizobuchi and S. Ogawa National Aerospace Laboratory of Japan 7-44-1 Jindaiji-higashimachi, Chofu, Tokyo 182-8522 JAPANInvestigation of lean premixed combustion dynamics in gas turbine combustors is numerically conducted based on Large Eddy Simulation (LES) methodology. The target combustors are a swirl-stabilized combustor and a flame holder model combustor. Unsteady flame behavior is captured inside these combustors. The unstable flame motions are mainly governed by acoustic resonance, velocity field fluctuations and heat release fluctuations. They are all coupled to sustain combustion instabilities. Velocity field analysis around the flame holder in the model combustor is also conducted to understand local flame shapes for further improvement of flame holder designing. It is demonstrated that numerical methods based on LES are effective in comprehending real-scale combustion dynamics in combustors.1. Introduction Reduction of NOx emissions is a key issue in modern gas turbine combustor systems due to environmental requirements. Lean premixed combustion is one promising way to reduce emissions because the flame temperature is low compared to conventional non-premixed combustion. However, lean premixed combustion is prone to combustion instabilities, and this sometimes may lead to detrimental damage to combustor systems. In order to attenuate unstable combustion behavior, combustion control is considered effective. Our final goal is to achieve stable combustion over a wide range of operation conditions by passive/active combustion control. Toward the final goal of combustion control, a detailed understanding of combustion dynamics in combustor systems is necessary because several factors are affecting combustion behavior in a complicated way. Mechanisms of combustion instabilities have been investigated worldwide both experimentally and numerically 1,2. In combustion experiments, especially for high-pressure tests, measurements are usually difficult to conduct. Visualization is in most cases limited to two-dimensional images and sensors can obtain data at several points. Recent progress in numerical simulation techniques has proven that CFD methods like DNS, RANS and LES can contribute to combustion research in various ways. To understand dynamics in combustor systems, it has been shown that the numerical approach based on LES techniques is effective, although several models are required in the formulation. This allows us to analyze data three-dimensionally. Here, we numerically simulate practical-scale gas turbine combustor systems to investigate dynamics and mechanisms of unstable combustion in combustor systems. The numerical methods are based on LES techniques and the flamelet model 3,4. The target combustors are a swirl-stabilized combustor and a flame holder type model combustor installed at National Aerospace Laboratory of Japan (NAL). Details of these combustors are described in the next section. 2. Combustors In this research we deal with two types of combustor systems. The first combustor is a swirl-stabilized gas turbine combustor. This type of configuration is generally used in modern gas turbine systems. Figure 1 shows the configuration of the swirl-stabilized combustor analyzed here. The combustion chamber length is L=0.3m and the diameter D=0.15m in this research. The combustor has an inlet section of inner diameter of Din=0.3D and outer diameter Dout=0.6D. The exit is contracted to 0.5D. Premixed methane/air at the equivalence ratio of 0.55-0.6 is swirled to the swirl number of about 0.8 in the inlet section. Swirl is numerically given to the flow by adding a hypothetical body force, not by solving swirler vanes. The unburned gas enters the combustion chamber with sudden expansion. The burned gas is accelerated in the contracted exit and goes out of the combustor. The pressure and temperature are set at 1 atm and 400-700K, respectively. We investigate several cases varying flow conditions as shown in Table 1.The second one is the model combustor installed at NAL, which has a cone-shaped flame holder to make the flame region larger. A pilot burner, which injects richer unburned gas to stabilize the flame, is located at the center of the flame holder. The flame holder has eight arms to make recirculation zones behind them. Figure 2 shows the eight flameholder arms and pilot burner installed upstream of the combustion chamber. The combustion chamber has a square cross section, although the inlet section where the flame holder is installed is circular.The combustor is mainly tested by experiment, and this numerical simulation mainly investigates the flame shape and velocity field around the flame holder. In experiment, it is difficult to directly visualize the flame holder region and numerical investigation will contribute to the understanding of the flow around the flame holder region. In the numerical simulation, injection form the pilot burner is neglected for simplification. The flow conditions are pressure of 1atm, inlet temperature of 700K, inlet velocity of 30m/s and the total equivalence ratio of 0.55. 3. Numerical models LES techniques are used to capture unsteady turbulent phenomena. In LES, large-scale coherent structures are directly captured by mesh and small-scale (sub grid-scale) eddies are modeled based on the similarity law of turbulence. Spatial filtering operation decomposes the governing equations described below into the two scales. The governing equations of the system are the three-dimensional Navier-Stokes equations. Compressibility must be included because pressure wave propagation inside the combustor plays an important role in determining flame dynamics.In an LES grid system, grid resolution is usually not fine enough to resolve internal structures of flame, thus a flame model should be added to the above equations. The well-known G-equation approach is used in the present formulation. The flame front is assumed to be thin and treated as a discontinuity surface between unburned and burned gases. The propagation speed of the flame surface is called the laminar burning velocity SL。 Numerical methods are based on the finite-volume discretization. The convective terms are constructed by Roes upwind numerical flux formulation. The time integration method is two-stage Runge-Kutta integration. Walls are treated as non-slip adiabatic walls. Incoming and outgoing boundary conditions are basically given as non-reflecting conditions. Perfectly non-reflecting conditions, however, do not determine the global mass flow rate. Here, partially non-reflecting conditions are imposed based on Poinsots method 5 to set the time-averaged mass flow rate at a fixed value. The number of grid point for the swirl-stabilized combustor is about 1 million, and 0.3 million for the NAL model combustor. These grid systems are somewhat coarse to make integration time shorter. Calculations using finer grid systems are underway and will be compared.4. Results and discussion 4.1. Swirl-stabilized combustor (a) Flame behavior In this section, case #1 is mostly analyzed. Figure 3 shows the time-averaged flame shape defined by the G value for case #1. By the swirling motion given to the inlet unburned gas, the flame is stabilized behind the dump plane. 。Figure 4 shows the time-averaged velocity field around the dump plane. A central recirculation zone is formed along the centerline, which pushes the hot burned gas upstream to keep the flame stabilized. Recirculation zones are also created in corner regions due to sudden expansion. This type of flowfield is typically observed in experiments using swirl-stabilized combustors.The time-averaged flowfield shown above seems rather simple. But instantaneous flowfield structures are more complicated and change unsteadily. From here, these structures are analyzed. The temporal pressure trace of case #1 for a certain period is shown in figure 5. The Fast Fourier Transform of the pressure trace is also shown in this figure. The basic frequency is about 740 Hz. The quarter-wave mode frequency, which is usually observed in experiment as a basic mode, can be readily estimated from the combustor length and the speed of sound of burned gas. The numerical and analytical frequencies are close and this means the present simulation well produces acoustic resonance in the combustor. Higher modes are also observed as harmonics.Local flame shape is determined by the balance between convective and flame velocities. Inside the combustor, pressure waves are propagating and this causes local velocity field fluctuations. The inlet unburned gas velocity at the dump plane is changing temporally leading to periodic vortex shedding from the dump plane. Figure 6 shows an example sequence of coherent vortical structures and flame shape. Blue structures are vortices identified by the vorticity magnitude. Red surfaces are flame surfaces identified the G value. The vortex shedding frequency in this case is determined mostly by acoustic resonance in he combustor. Figure 6 corresponds to nearly one acoustic period of the basic quarter-wave mode. Shear layer instability may also play a role in vortex forming. The relationship between acoustic shedding and shear layer instability is still not clear and should be analyzed in the future. The flame shape is locally deformed when vortical structures are present in the vicinity. This vortex/flame interaction changes the total flame surface area and flame position. Because heat release occurs at the flame front, fluctuations in flame shape and location are directly connected to heat release fluctuations. The Rayleigh index is a criterion of combustion system stability. The global Rayleigh index is given by integrating the product of pressure and heat release fluctuations over time and space dimensions. In the present model, local heat release rate is estimated byIf the global Rayleigh index is positive, the system is unstable by combustion. For case #1, contours of the local Rayleigh index integrated over time around the dump plane are shown in figure 7. In some regions the local Rayleigh index is negative, but in most areas, the index is positive. The global Rayleigh index integrated over space is 2.498(non-dimensional) and positive. Thus heat release and pressure fluctuations are coupled to make the system unstable. From these results, it can be said that combustion instabilities are driven by interactions between acoustic pressure oscillations, velocity fluctuations leading to vortex shedding and heat release fluctuations. Attenuation of combustion instabilities depends on how to weaken or change these links. This requires a detailed understanding of effects of various parameters, and we need much more calculation cases for this purpose. Below, a preliminary comparison is made between other cases in different conditions. Some experimental results indicate that preheating of unburned gas makes combustion more stable 6 to some extent. Comparing pressure traces of case #1 (preheated) and #2 (less preheated), the mean amplitude of case #2 becomes larger. Furthermore, if the global equivalence ratio is increased slightly in case #3, the amplitude becomes again larger. Figure 8 shows the relative pressure amplitude of the three cases. Heat release and thermal expansion ratio may be part of the reason, but further investigation is needed for quantitative evaluation. And also as a future plan, we need to examine the effect of equivalence ratio fluctuations. In real combustor systems,perfectly constant equivalence ratio premixing is almost impossible because fuel and air injection rates can be affectedby pressure waves from the combustor. This causes fluctuations in local equivalence ratio, thus local heat release and is expected to make the system unstable. We will soon conduct simulations including this effect. 4.2 NAL model combustor (a) Flame shape The model combustor investigated at NAL is simulated using the same methodology. Figure 9 shows an experimental image of the combustor in operation. The burned gas illuminates brightly, but the flame shape cannot be directly seen. The same LES methodology is applied to this combustor to understand the flame shape around the flame holder.Direct comparison of experimental and numerical flame shapes may be difficult, but some experimental gas-sampling results can help us to understand the averaged flame shape. One example is illustrated here. Shown in figure 10 (a) is the experimental NOx distribution result measured by gas sampling at the combustion chamber entrance. Dashed lines indicate the flame holder arm positions. Because high NOx regions correspond to high temperature regions, combustion seems occurring mainly in areas near the center axis and the arm skirts. Figure 10 (b) shows the numerical result of temperature distribution at the same axial position. The dashed square indicates the measurement area of figure 10 (a). High temperature regions are located at the center and the arm skirt areas. Both results are in good agreement and the present method is able to capture the flame shape.Figure 11 shows the three-dimensional flame shape viewed from upstream and the temperature distribution behind the flame holder arm. The unburned gas from the inlet is accelerated when passing through flame holder slits. Behind the flame holder arms, small recirculation and flame regions are created. These recirculation zones are a little smaller than expected. In the pilot burner wake along the center axis, a recirculation zone is also formed. However, these results may change a little if different wall boundary conditions are imposed because wall heat transfer may have some influence. More experimental data set will be available in the near future and further analysis will be conducted.5. Conclusions Numerical investigation of combustion behavior in gas turbine combustors is presented. Two types of combustors are investigated: a swirl-stabilized combustor and a flame holder combustor. LES-based numerical methods employed here are useful in reproducing combustion instabilities in practical-scale combustors. Unstable combustion behavior is driven by the coupling effects between acoustic resonance, velocity fluctuations and heat fluctuations. If equivalence ratio fluctuations are present, this will also affect the behavior. The coupling between these factors changes depending on flow conditions, and detailed parametric studies are underway. For the NAL model combustor, the flowfield and flame shape analysis has been conducted and revealed the flowfield structures. Further investigation will be conducted in cooperation with the experimental team. 6. References 外文文献翻译译文燃气轮机燃烧室火焰大涡模拟数值分析 原文出处及作者：Numerical analysis of flame behavior in gas turbine combustors using LESChofu, Tokyo 182-8522 JAPAN J. Shinjo, Y. Mizobuchi and S. Ogawa 对燃气轮机燃烧室稀预混料燃烧状态的数值研究是基于Large Eddy Simulation（大涡模拟）方法进行的。作为研究对象的燃烧室是漩涡稳定和带有火焰稳定器的。在这些燃烧室中，可以捕获不稳定的火焰燃烧状态。引起不稳定燃烧的主要因素有音响共鸣，流速脉动以及热释放脉动。这些因素共同作用，导致火焰燃烧不稳定。本文同时研究了目标燃烧室火焰稳定器周围的速度场，以及了解该处燃烧形式，对于将来改进火焰稳定器的设计提供帮助。研究表明，基于大涡模拟的数值分析方法对于全面了解燃烧室实时动态是十分有效的。1.绪论根据环境保护的需要，在现代燃气轮机燃烧室系统中，减少NOx的排放是极为重要的。稀预混料燃烧室的方法对于减少NOx排放有广阔前景。与传统的非预混料燃烧室相比，它的火焰温度要低的多。但是，稀预混料燃烧室有其固有的弱点，即燃烧不稳定，这可能损害燃烧室系统。为削弱不稳定燃烧的影响，必须采取有效的燃烧控制方法。该项研究的最终目标是在多种运行工况下，通过能动/非能动的燃烧控制，实现稳定燃烧。为实现最终的燃烧控制目标，必须对燃烧室系统的燃烧动态有详细的了解，因为影响燃烧状态的因素是多样的，机制是复杂的。世界范围内，对于燃烧不稳定的问题，学者们已通过实验和数值计算的方法进行过深入的研究1，2。在进行燃烧实验，尤其是高压状态下的实验中，很难进行具体测量。大多数情况下，可以实现的是在二维平面下，使用传感器获得有限点的数据。数值仿真技术最新的进展表明，CFD方法，例如DNS，RANS和LES对于燃烧的研究，从各个方面以不同的方式发挥着作用。已经证明，基于LES技术的数值方法对于燃烧室系统燃烧动态的研究是有效的。这使得我们可以在三维状态下分析数据。在这里，为了研究燃烧室系统不稳定燃烧的实时动态和机制，我们数值模拟了实际的燃气轮机燃烧室系统。计算方法基于LES技术以及flamelet模型3,4。研究对象是一个漩涡稳定的燃烧室和装在日本国家航空和宇宙实验室（National Aerospace Laboratory of Japan (NAL)中一个带有火焰稳定器的模型。在下一部分，将详细的介绍燃烧室。2.燃烧室在这一部分，讲述两种类型的燃烧室系统。第一种燃烧室是漩涡稳定的燃气轮机燃烧室。它广泛应用于当今燃气轮机系统中。作为研究对象的漩涡稳定燃烧室详细配置见图1.燃烧室的长度L=0.3m，直径 D=0.15m。燃烧室进口的内径是 Din=0.3D，外径 Dout=0.6D。出口收缩到0.5D。入口部分预混合的天然气/空气比率是0.55-0.6，入口的漩涡系数为0.8.漩涡的形成数值上看做是通过向混合气流施加假想的质量力，而不是通过涡旋式离心喷嘴。未燃烧的气体进入燃烧室时突然膨胀。燃烬的余气通过压缩喷嘴出口加速离开燃烧室。压力和温度分别设定为1 atm和400-700k。在各种不同条件下的流动状态如表1所示。第二种燃烧是装在NAL的燃烧室。它装有锥形的火焰稳定器以扩大火焰燃烧区域。火焰稳定器的中心装有注入富气以稳定燃烧的实验燃烧器。火焰稳定器有8臂，在其后形成了流通区域。图2 展示了8个火焰稳定器的支臂以及装在燃烧室上方的实验燃烧器。实验燃烧器有一个方形的截面，装有火焰稳定器的入口部分为环形。该燃烧室主要通过试验测试，数字仿真模型研究对象主要是火焰的形状和火焰稳定器周围的流速场。在试验中，很难直接观察火焰稳定器周围区域，数值研究主要集中在火焰稳定器周围的流动状态。在数字仿真中，为简化起见，自先导燃烧室注入的富气忽略不计。流动条件是1 atm（标准大气压），入口温度700K，入口流速为30m/s,入口气体的配料比例是0.55.3.数学模型LES技术被用来捕获不稳定紊流现象。在LES中，大范围的连续的燃烧状态被分解为栅格结构的小范围的漩涡流，建立栅格模型的是基于紊流规律。空间滤波运作分解为基本方程，如下所述：系统的基本方程是三维的粘性流体方程。由于燃烧室中的压力对于火焰动态起着重要作用，所以必须考虑可压缩性。在一个LES栅格系统中，栅格方案并不足以描述火焰燃烧的内状态，因此，必须在上面方程的基础上，添加火焰燃烧模型。在目前应用的方程中，使用了著名的G方程。火焰锋作为未燃气体和已燃气体的间断面被假定为稀薄的，火焰表面的传播速度定义为层流燃烧速度SL。在基本方程中，使用了LES立体过滤处理。例如，经处理的动量方程是拔（上划线）表示平均值，以及Favre平均。最后的一个条件是未解出的子栅格条件。动态过程的模型如下所示：插入公式（4）C表示动力系数，过滤比例，应力当量。 数值方法基于有限离散化。对流条件约束由Roe上风数值通量方程确定。时间积分方法引用二阶的Runge-Kutta积分。假定燃烧室壁为理想的非滑动绝热。流入和流出的边界约束条件在无反射条件下给定。全无反射条件可能不足以决定全部的质量流量率。在这里，部分无反射条件基于Poinsot方法5给定，以设置固定数值的单位时间平均质量流量率。稳漩涡稳定的燃烧室栅格计算点数目为100万，对于NAL的模型燃烧室，取30万个计算点。这些数目的栅格系统可使的积分时间更短一些。针对更好的栅格系统的计算方法正在研究中，将来会加以比较。 4.结论和讨论 4.1漩涡稳定燃烧室（a）火焰状态在这一部分，案例#1研究最深入。图3展示了平均时间内由G评价决定的火焰形状。通过给定入口未燃气体的漩涡运动，在气体倾入燃烧室的截面，火焰燃烧稳定。图4展示了平均时间内在倾入截面的流速场。在轴线附近形成了中心再循环区域，使得燃烬的气体远离，以保持火焰燃烧稳定。由于突然的膨胀，在转角区域也形成了再循环区域。在漩涡稳定的燃烧室中，可以观测到典型的该形式的流动场。上图所示的平均时间的流动场看起来似乎很简单，但是瞬时的流动场结构更加复杂，变化不稳定。从这开始，分析瞬时