Bir benzinli motorun türbülanslı akış alanlarının incelenmesi

Abstract

Günümüzde, enerji kaynaklarının yoğun bir şekilde kullanımı sonucu, hava kirliliği ortaya çıkmakta, bu da kaynakların verimli bir şekilde kullanılmasını gerekli kılmaktadır. Bu enerjinin büyük kısmı içten yanmalı motorlarda, mekanik güç üretmek için kullanılmaktadır. İçten yanmalı motorlar, başta karayolu taşıtlarında ve zirai araçlarda olmak üzere, deniz ile hava araçlarının birçoğunda, ayrıca enerji üretiminde yaygın olarak kullanılmaktadır. Bu kadar yaygın kullanım ise, motorların, hava kirliliği ve yakıt sarfiyatı açısından, araştırma ve geliştirmesini gerekli kılmaktadır. İçten yanmalı motorların hava kirliliğine olumsuz etkisi nedeniyle, emisyon değerlerini daha iyi hale getirmek için yoğun çalışmalar yapılmaktadır. Emisyon değerlerinin düzeltilmesi ise, yanma veriminin artırılmasına bağlı kalmaktadır. Yanma odası içindeki türbülanslı hız alanlarının yanmaya doğrudan etkisi olmaktadır. Türbülanslı hız alanlarının benzin ve dizel motorlarındaki görevi ilk olarak karışımın hazırlanması ve ikinci olarak ise büyük ve küçük ölçekli karışım içindeki yanmayı kontrol etmek olmaktadır. Türbülanslı akışlarda, transfer ve karışım miktarları, moleküler difüzyonla meydana gelen miktardan birkaç kat daha fazladır. Türbülans sonucu meydana gelen difüzyon, momentum, ısı ve kütle transferi miktarlarını artırmaya neden olmaktadır. Bu çalışmada, daha önce boru içinde daimi olmayan akışlara başarılı olarak uygulanan Türbülans Filtresi, benzinli motorun yanma odası içindeki hız alanlarına değişik emme sübapı, sıkıştırma oranı ve devir sayılarında uygulanmıştır. Türbülans Filtresi, yeni geliştirilen ve içten yanmalı motor yanma odası içi hız alanlarına ilk olarak uygulananan bir filtredir. Türbülans Filtresi, perdeli sübap için, toplam hız verilerinden organize ve türbülans bileşenlerini başarılı olarak ayrıklaştırmaktadır. Kullanılan diğer yöntem ise, alışılmış Faz Ortalama yöntemidir. Bu yöntem de, tüm motor konfıgürasyonlarında toplam hız verilerine uygulanmış ve Türbülans Filtresinin bulduğu sonuçlarla karşılaştırılmıştır. Başka bir yöntem ise, Lumley tarafından önerilen ve akışkanlar mekaniğinde yaygın olarak kullanılan POD yöntemidir. Bu yöntemin bulduğu en çok enerjik mod organize hız olmakta, diğer ortogonal ayrıklaştırılmış edilerin toplamı da türbülans bileşeninin tahmini olarak alınmaktadır. Toplam hız verilerinin ortogonal yapılarını incelememizi sağlayan eigen vektörleri POD yöntemi ile bulunmuş ve ilk on modda incelenmiştir. Türbülans filtresi ve POD metodu, içten yanmalı motor yanma odası içi hız alanlarına ilk olarak bu çalışmada uygulanmıştır. Ayrıca, içten yanmalı motorun yanma odası içinde değişik motor konfıgürasyonlarında ölçülen hızlar, birinci derece otoregresiv modeli kullanılarak modellenmiştir.

The turbulent flow field in an engine plays an important role in determining its combustion characteristics and thermal efficiency. Automotive engineers have learned that changes in the combustion chamber shape and inlet system geometry, both of which change the turbulent flow field, influence emissions, fuel economy and the lean operating limit of an engine. Most of this knowledge has been obtained on specific engines through direct experimentation or from global measurements. As a result there exist no general scaling laws to predict the combustion and emission characteristics of an engine. The importance of the turbulence structure in an engine has been recognized since the early experiments of Clark, in which the intake event was eliminated and the rate of flame propagation decreased. However, the lack of an adequate measuring instrument has made measurements of turbulence quantities in an engine difficult at best. As a result, relatively few investigators have been active in this field and few, if any, general scaling laws governing turbulence characteristics have been developed. The engine intake process governs many important aspects of the flow within the cylinder. In four-stroke cycle engines, the inlet valve is the minimum area for the flow so gas velocities at the valve are the highest velocities set up during the intake process. The gas issues from the valve opening into the cylinder as a conical jet and axial velocities in the jet are about 10 times the mean piston speed. The jet separates from the valve seat and lip, producing shear layers with large velocity gradients which generate turbulence. This separation of the jet sets up recirculation regions beneath the valve head and in the corner between the cylinder wall and cylinder head. The jet-like character of the intake flow, interacting with the cylinder walls and moving piston, creates large-scale rotating flow patterns within the cylinder. The details of these flows are strongly dependent on the inlet port, valve, and cylinder head geometry. These flows appear to become unstable, either during the intake or the compression process, and break down into three-dimensional turbulent motions. Recirculating flows of this type are usually sensitive to small variations in the flow: hence there are probably substantial cycle-by-cycle flow variations. In turbulent flows, the rates of transfer and mixing are several times greater than the rates due to molecular diffusion. This turbulent "diffusion" results from the local fluctuations in the flow field. It leads to increased rates of momentum and heat and mass transfer, and is essential to the satisfactory operation of spark ignition and diesel engines. Turbulent flows are always dissipative. Viscous shear stresses perform xvn deformation work on the fluid which increases its internal energy at the expense of its turbulence kinetic energy. So energy is required to generate turbulence: if no energy supplied, turbulence decays. A common source of energy for turbulent velocity fluctuations is shear in the mean flow. Turbulence is rotational and is characterized by high fluctuating vorticity: these vorticity fluctuations are three dimensional. The character of a turbulent flow depends on its environment. In the engine cylinder, the flow involves a complicated combination of turbulent shear layers, recirculating regions, and boundary layers. The flow is unsteady and may exhibit substantial cycle- to-cycle fluctuations. Both large-scale and small-scale turbulent motions are important factors governing the overall behavior of the flow. The induction process terminates with the closure of the inlet valve and establishes the initial conditions for the subsequent compression process. These initial conditions at IVC (Inlet Valve Closing) reflect port/valve geometry and the small effect of piston geometry on the flow development during induction. In contrast, as the piston approaches TDC (Top Dead Center) of compression, the effect of inlet geometry diminishes and the piston and cylinder-head geometry plays the dominant role in the structure of the mean and turbulent velocity fields. The closure of the inlet valve finds the turbulent velocity field in a state of continuing decay due to the termination of the turbulence production process. In the axial plane, the intake-generated vortex pattern breaks down, with no sign of inlet flow structures persisting through compression, except for the main vortex which sometimes survives although much weaker. The swirl-induced pressure field may also generate an axial motion rotating in the opposite direction to the main vortex during induction. In the tangential plane, the swirling flow is less influenced by the piston-induced pressure forces and by virtue of the cylindrical wall geometry is transformed into an organized strong structure whose velocity distribution resembles solid body rotation. This type of motion conserves energy since it has no internal shear. The low levels of tangential shear contribute little to turbulence production, except near the walls, which is contrary to the common belief that swirl enhances turbulence, an increase of induction swirl may even suppress turbulence with solid body type velocity distributions. An important characteristics of a turbulent flow is its irregularity or randomness. Statistical methods are therefore used to define such a flow field. In an unsteady turbulent flow, there are organized and turbulent motions. But, in engines, the flow pattern changes during the engine cycle. Also, while the overall features of the flow repeat each cycle, the details do not because the organized motions can vary significantly from one engine cycle to the next. There are both cycle-to-cycle variations in the organized motion at any point in the cycle, as well as turbulent fluctuations about that specific cycle's organized motion. The general characteristics of the in-cylinder flows in internal combustion engines which are common in both gasoline and diesel types can be summarized as: 1. unsteady or non-stationary flow as a result of the reciprocating piston motion, 2. turbulent flow at all engine speeds and for all inlet port/cylinder dimensions, 3. three-dimensional flow as a result of the engine geometry, xvm 4. bulk flow in phase with the engine cycle, 5. variations in cycle-to-cycle local flow properties, 6. time scales associated with the bulk flow variations of the same order as the turbulent time scales. It becomes clear from these features that the flow is complex and its interpretation is difficult. Separation of the mean flow from turbulence is a matter of definition and is important because the speed of flame propagation is related to small scale fluctuations. From the modeling point of view, the distinction is also important since it is necessary that definitions of experimentally measured turbulence quantities have a modeling counterpart. The conventional statistical approach for stationary turbulent flows is to decompose the instantaneous flow properties 0(x,y,z,t) into a time-averaged and a fluctuating component. Due to the non- stationary character of the in-cylinder flow, an equivalent definition is required and is based on ensemble averaging over a statistically acceptable number of engine cycles. The most frequently used approach in engine velocity measurements, which are more readily obtained than those of scalar properties, is ensemble averaging which effectively becomes a time/ensemble averaging procedure when the output signal from the measurement system is intermittent and random. A second data-reduction procedure is the cycle-by-cycle analysis whereby a separate mean velocity is estimated for each engine cycle. This is achieved by some form of the filtering which causes the instantaneous velocities to pass through a filter. Depending of the filter, information is obtained about the mean flow (low pass) or turbulence (high pass). This approach is based on selection of a cut-off frequency for the part of the engine cycle under investigation and involves some arbitrariness, although physical justification can be provided. There are also decomposition procedures which have been applied to stationary turbulent flows, in which the relevance of the decomposition procedure to turbulent motions is less direct. For example, the Proper Orthogonal Decomposition has gained popularity as an efficient way of organizing data into orthogonal modes, which decrease in order of their contribution to the energy of the flow. Lumley proposed describing random fields such as turbulence as deterministic functions which are as nearly parallel as possible to u, ensemble vector of the field. In the statistical sense, the modulus squared of the product of the vector and the deterministic function are maximized. The significance of this decomposition is that its modes constitute a special basis which is an optimal set in the sense that when truncated at any arbitrary number of modes, no other decomposition truncated at the same order can capture a greater fraction of the energy of the data series. The proper orthogonal decomposition has been widely used in many aspects of turbulence research. For low Reynolds number flows, a large portion of the energy is often contained in relatively few modes, so that coherent structures are easily detectable and may be identified by this decomposition. However, at high Reynolds numbers, the energy of turbulent flows is often distributed over many modes, with the result that detection of coherent structure becomes difficult and they are less visible. If there exists, in advance, XIX information that one kind of motion will be very much more energetic than others, a proper orthogonal decomposition will isolate that motion so that it is represented as orthogonal to all others. In this study, Turbulent Filter proposed by Kodal in his Ph.D. thesis is used to decompose the turbulent data series of the S.I. engine. The Turbulent Filter is for general classes of unsteady turbulent motion and does not rely on a precise knowledge of the period of organized motions which are identical from one cycle to another. Instead, when the periodicity of the organized motion is imprecisely known, or might vary from cycle to cycle, periodicity information must be replaced by sufficient information of other kinds to determine the appropriate decomposition. The information used is a model of the shape of the power spectrum of turbulent component. Its goal is to decompose turbulent data series into spatially/temporally organized and turbulent motions. The Turbulent Filter can be related to transfer functions in classical control theory, though it does not appears to have been applied to decomposition problems in fluid mechanics. For simplicity, the formulation is developed for one-dimensional time series, though it may easily be extended to more dimensions. The measured data series are modeled by using high order autoregressive models. The turbulent power spectrum is modeled by first order autoregressive model. The energy spectra of the organized and turbulent components are permitted to overlap and so take different non-zero values at discrete frequencies. This assumption distinguishes the decomposition from other optimal filtering approaches and recognizes the broadband nature of turbulence and organized unsteadiness in fluid flow. It is assumed that a good estimate of the energy spectrum of turbulent contribution, Suv, can be made from an energy spectrum of a measured data series, Suu. Then, the turbulent power spectrum is iterated towards the correlation between the organized and turbulent components are zero. Also, in this study, Phase Average and POD (Proper Orthogonal Decomposition) method together with the Turbulent Filter method are applied and compared with each other for the measured velocity data series in the motored S.I. engine. The used data series are 30 cycles, and each cycle has 1306 data points. In figure 1, measured and decomposed organized motions by these techniques are showed. Organized motions deduced by the phase average and POD methods are very close to each other and the organized motion deduced by the Turbulent Filter is approximately close to them. The general trend of three methods' organized motions are decreased with crank angle. But, at the end of TDC, organized motions are increased as shown in figure 1. Moreover, first ten modes of the eigen vectors of the measured velocity data series are determined. The eigen modes are uncorrelated with each other. The total energy fraction of the decomposed eigen modes are computed. The computed eigen modes showed that most of the activity is at the first four modes. Decomposed eigen modes are shown in figure 2. As shown in figure 2, first mode of the measured circumferential velocity are decreased with crank angle, and the others fluctuate around zero. Finally, the measured data series are modeled by means of first order autoregresiwe method. xx 15 -15 -- Total - Phase Average - Turbulent Filter ?POD ^ \ /"%, /*" ~?$hî''\ ?'?<* *(!*'". 180 210 240 270 300 330 360 Crank Angle (degree) 390 420 Figure 1. Circumferential total and organized motions for the shrouded valve (n=1500 rpm and 8=8:1, ü=16.05 m/s, Rei=3889.36) - 1 th Mode 2th Mode 3th Mode 4th Mode -+- 180 210 240 270 300 330 360 Crank Angle (degree) 390 420 Figure 2. Circumferential first four modes for the shrouded valve (n=2000 rpm and s=8:l, u=17.39m/s, Re,=4189.15)