预混本生火焰中平面激光诱导荧光PLIF检测方案(气体流量计)

智能文字提取功能测试中

Fuel 242 (2019) 607-616Contents lists available at ScienceDirectjournal homepage： www.elsevier.com/locate/fuel Y. Nie et al.Fuel 242 (2019) 607-616 Fuel Full Length Article Flame brush thickness of lean turbulent premixed Bunsen flame and thememory effect on its development Yaohui Nie， Jinhua Wang*， Weijie Zhang， Min Chang， Meng Zhang， Zuohua Huang State Key Laboratory of Multiphase Flow in Power Engineering， Xi’an Jiaotong University， Xi’an 710049， China ARTICLEINFO ABSTRACT Keywords：Flame developmentFlame brush thicknessNon-local effectPDF distribution Flame brush thickness is a significant parameter that demonstrates the evolution of turbulent premixed flame. Inthis paper， the flame height and flame brush thickness of lean turbulent premixed Bunsen CH4/air and C3Hg/airflames were investigated under low turbulence intensities. The turbulence flow field is generated and controlledby changing the outlet velocity and perforated plates， and it is measured using a hot wire anemometry system.The turbulent flame fronts are detected by OH-PLIF technique. Results show that under fuel-lean conditions， thecharacteristic turbulent Bunsen flame height and centerline flame brush thickness decrease with increasingequivalence ratio， while they increase with the augment of outlet velocity. Turbulence intensity has marginaleffect on these two parameters. The variations of flame brush thickness indicate that there are three regionsduring the whole development of turbulent Bunsen flames under low turbulence intensity. The first is the tur-bulence dominating region， where the variations of horizontal flame brush thickness follow the turbulencediffusion theory. The second is the non-local effect manifesting region， where the equivalence ratio， whichcontrols the flame instability wavelength， influences the horizontal flame brush thickness. The third is the flametip region where the flamelets intersect and make the radial wrinkles offset while the axial wrinkles super-impose. This unique tip region of turbulent Bunsen flame can be demonstrated by the linear relationship be-tween its flame height and centerline flame brush thickness. Turbulent premixed Bunsen flames suffer differentnon-local effects during the flame development to downstream. To minimize the non-local effect on the statisticsof turbulent flame front structure， a new method based on the development time period is proposed. Comparedwith statistical analysis based on the whole development time， this method can better demonstrate the flame-turbulence interaction. 1. Introduction Turbulent premixed combustion has been widely studied because ofits significance in the development of low-emission combustion devicessuch as internal combustion engines [1]， gas turbines [2] and com-bustion furnaces [3]. Turbulent combustion usually operates in theflamelet regime [4-6]， even when the Karlovitz number exceeds ten. Inthis regime， turbulent flame speed is mainly controlled by flamewrinkles， especially when Lewis number of the fuel/air mixture is nearunity [7]. To get a better prediction of turbulent flame speed， a com-prehensive understanding of physical mechanisms which cause theflame wrinkling is essential [8]. However， for turbulent Bunsen and V-shaped flames， as the coming mixture is oblique to the mean flameposition， flame wrinkles generated upstream will be transferred todownstream. This is the non-local effect of turbulent flames， which isalso called“memory effect”[9]. It has been found that there is less flame wrinkles (thus a lower value of local turbulent consumptionspeed) near the base region of Bunsen flame than the tip region [10].Therefore， many studies have analyzed the middle part of turbulentBunsen flame front to investigate turbulence-flame interaction. Halter[11] used the PDF (Probability Density Function) of curvature of theturbulent Bunsen flame front calculated at the identical height to de-monstrate the effect of flame thickness and small scale vortex on theturbulent flame. For turbulent premixed Bunsen flame whose turbu-lence is generated by perforated plates， one effective way to increaseturbulence intensity is to enlarge the outlet velocity， which will result ina larger flame height. For the same perforated plates， when the outletvelocity is doubled， the flame height will be approximately doubled，too. Therefore， flame statistics calculated based on the whole Bunsenflame front or middle flame front fixed at the same height may notaccurately demonstrate the turbulent flame characteristics， becauselong turbulent Bunsen flames will have much time for wrinkling to ( * Corresponding author. ) ( E-mail address： jin h uawang@mai l .xjtu. e du. c n (J. Wang). ) develop compared with short flames. So a more reasonable method toexplore flame-vortex interaction by minimizing the non-local effect onthe statistics of turbulent premixed Bunsen flame front structure isneeded. In the turbulent premixed Bunsen flame [12]， V-shaped flame [13]and spherical expanding flame [14]， flame brush thickness， 8FBT， canrepresent the development of flame wrinkles. From a practical view-point， the turbulent flame brush represents the spatial region overwhich the instantaneous turbulent flame fronts are located. It controlsthe spatial distribution of the time averaged heat release in a combus-tion system [15]， which can influence combustion stability and emis-sions [16]. Therefore， the turbulent flame brush thickness not only canvalidate numerical models [17，18] but also may lead to a better un-derstanding of turbulent combustion [19]. Using tomography method，Goix [20] investigated the evolution of flame brush for several V-shaped premixed H2/air flames. It was found that the flame brushthickness versus downstream distance was proportional to turbulenceintensity. Kwon [21] measured the instantaneous structure of turbulentspherical flames in a fan-stirred bomb and found that the dispersion offlame radius distribution appeared to be proportional to the mean brushthickness and increased with time. Moreau [22] and Chen [23] ob-served the similar phenomenon in confined oblique flames and Bunsenflames. All these experiments suggested that the development of tur-bulent flame followed the principle of turbulence diffusion law [24].For the above experiments where the turbulence is generated by per-forated plates，turbulence intensity， u， is proportional to the averageincoming velocity， U， then the dependence of8pBT~(y’)(H)~u't ~ Hu’'/U predicts that the flame brush thicknessincreases roughly linear with flame height， H， when outlet velocity isfixed [25]. However， Tamadonfar [26] found that the equivalence ratioaffected the turbulent brush thickness by investigating the separateeffects of turbulence intensity and outlet velocity under weak turbu-lence conditions. In turbulent premixed Bunsen flame， equivalenceratio controls laminar flame speed and thus can affect flame height. Inaddition， equivalence ratio also affects flame intrinsic instability.Fogla[27] demonstrated that effect of flame intrinsic instability on flamewrinkles can be hidden under strong turbulence. So the variations ofturbulent flame brush thickness under low turbulence intensity shouldbe studied in order to describe the underlying mechanism during thewhole development of turbulent Bunsen flame wrinkling. The objective of this paper is to investigate the separate effect ofturbulence intensity， outlet velocity and equivalence ratio on the tur-bulent Bunsen flame brush thickness under weak turbulence. The wholedevelopment of turbulent Bunsen flame brush is obtained. A newmethod is proposed to minimize the non-local effect on the statisticaldata of turbulent Bunsen flame front. The statistical data of turbulentBunsen flame was derived based on the same development time period.This paper is organized as follows. The experimental setup and mea-surement methodology are outlined in Section 2. Sections 3.1 and 3.2investigated these effects on development of turbulent flame， andSection 3.3 elaborates the memory effect on the turbulent Bunsen flameand the whole flame wrinkling process. In Section 3.4， a new method isproposed to minimize the non-local effect of turbulent Bunsen flamewhen we investigate the underlying mechanism of flame-turbulenceinteraction. Main conclusions are summarized in Section 4. 2. Experimental apparatus and methods 2.1. Experimental setup The experimental setup consists of turbulent premixed Bunsenburner， OH-PLIF system and gas supply system. Experiments are per-formed on a Bunsen type burner， as shown in Fig. 1(a)， whose outletdiameter is 20 mm. A circular slit with thickness of 0.5 mm is used aspilot flame. Turbulence is generated by perforated plates that can be placed at four locations in the burner， as shown A， B， C， D in Fig. 1(a)，which means that they are 20， 30， 50， 70 mm upstream the burner exit.The holes are arranged hexagonally with two kinds of blockage ratio togenerate controllable initial turbulence conditions. In this experiment，the perforated plates labeled as P1 and P2， as shown in Fig. 1(b)， arelocated at location D in the burner. Their geometrical properties， whichinclude the hole diameter d， mesh size M， and blockage ratio ω， arelisted in Table 1. The instantaneous flame front is detected by OH-PLIF system. Itincludes a pulsed Nd：YAG laser (Quanta-Ray Pro-190)， with power of300 mJ and wavelength is 532 nm， a Dye laser (Sirah PRSC-G-3000)，after which the pulse energy is about 15 mJ. The OH fluorescence iscaptured by an ICCD camera (LaVision Image Prox). The frequency ofthe ICCD camera is 10 Hz， which is synchronized with laser pulse. Moredetails of the experimental setup can be found elsewhere [28，29]. Turbulence is measured using a hot wire anemometry system(Dantec， Streamline 90 N) at sampling rate of 300 KHz for 5s along thecenterline of the burner exit under non-reacting conditions. The mea-surements are performed at 10 mm above the burner exit. Firstly， wemeasured 11 points evenly along the radial direction at the exit distanceof h=10 mm， then we measured 6 points along the centerline. It wasfound that the turbulence intensity was almost constant along the radialdirection except where near the brim of the burner. Along the axialdirection， the turbulence intensity decreased slightly even at the exitdistance of h = 30 mm. The turbulence parameters at the center loca-tion of h=10 mm was selected to represent the near isotropic flowfield. The air flow is controlled using a mass flow controller (MKS1179A， range： 0-50 SLM， Standard Liter per Minute) and the fuel ismetered by a mass flow controller (MKS 1159A， range： 0-5 SLM). All ofthem have an accuracy of 0.8% on its reading， and 0.1% on its fullscale. 2.2. Measurement methodology and theory In this study， lean turbulent premixed CH4/air and C3Hg/air flamesare investigated at room temperature and atmospheric pressure. Theequivalence ratio， ， changes from 0.7 to 0.9 for CH4/air flames andvaries from 0.8 to 1.0 for C3Hg/air flames for comparisons with theexperimental study of Guilder [30]. The experimental conditions aregiven in Table 2. The turbulence intensity， u'， was determined as therms-value of axial velocity fluctuation. The integral length， lo， wasevaluated by integrating the longitudinal velocity correlation coeffi-cient of the velocity signal in the axial mean flow direction underTaylor’s hypothesis of frozen turbulence. The Taylor scales， A， andKolmogorov scales， n， are also listed in Table 2. The unstrained laminarburning velocity， Sio， was calculated by CHEMKIN Pro software withGRI-Mech 3.0 mechanism. The flame thickness， 8z， is defined as ratio ofthe thermal diffusivity to the unstretched laminar burning velocity， a/SLo. The experimental turbulent flames are plotted on the modifiedBorghi’s diagram， as shown in Fig. 2(a). It can be seen that most of theconditions are located in the wrinkled flamelet regime. Flame intrinsic instability is an important factor which can causeflame wrinkles. Sivashinsky’s linear dispersion formulation of growthrate， which combines the Darrieus-Landau instability and diffusive-thermal effects， is used to estimate the instability scales of the mixtures.The growth rate of the flame disturbances， o，is e(1+20)(Q0+E)(1-Le) 1/-1 ln(+1)di， a and B are thermal diffusivity and2(1-E)[+(1+E)20]JoZel’dovich number， Le represents the Lewis number of the mixtures[31，32]. The wave number， which can cause the greatest growth rate， isdefined as km， as shown in Fig. 2(b). The characteristics wavelength，lm =n/km， for all the experimental conditions is listed in Table 2. Pilot flame (a) Turbulent Bunsen burner Table 1 Summary of geometrical properties for perforated plates： d， hole diameter； M，mesh size； ω， blockage ratio. Perforated plates d(mm) M (mm) ω (%) Pl 3.5 3.6 63.5 P2 4 4.5 55.5 Table 2 Experimental conditions. Symbol： d：equivalence ratio； U： bulk flow velocity； u’：turbulence intensity； lo： integral length scale； A： Taylor scale； n kolmogorovscale； lm： instability wavelength. Set of Exp U u' lo SL，o l m/s m/s mm mm mm cm/s mm 0.7 2 0.093 7.431 1.129 0.417 19.02 1.186 P1， CH4 0.8 2 0.093 7.431 1.129 0.417 26.66 0.848 0.9 2 0.093 7.431 1.129 0.417 33.05 0.717 0.7 3 0.169 4.590 0.664 0.238 19.02 1.186 P1， CH4 0.8 3 0.169 4.590 0.664 0.238 26.66 0.848 0.9 3 0.169 4.590 0.664 0.238 33.05 0.717 0.7 4 0.201 3.784 0.547 0.193 19.02 1.186 P1， CH4 0.8 4 0.201 3.784 0.547 0.193 26.66 0.848 0.9 4 0.201 3.784 0.547 0.193 33.05 0.717 0.7 3 0.208 4.181 0.553 0.186 19.02 1.186 P2， CH4 0.8 3 0.208 4.181 0.553 0.186 26.66 0.848 0.9 3 0.208 4.181 0.553 0.186 33.05 0.717 0.8 2 0.093 7.431 1.129 0.417 36.87 0.059 P1， C3Hs 0.9 2 0.093 7.431 1.129 0.417 44.17 0.046 1.0 2 0.093 7.431 1.129 0.417 49.16 0.040 0.8 3 0.169 4.590 0.664 0.238 36.87 0.059 P1， C3Hg 0.9 3 0.169 4.590 0.664 0.238 44.17 0.046 1.0 3 0.169 4.590 0.664 0.238 49.16 0.040 0.8 4 0.201 3.784 0.547 0.193 36.87 0.059 P1， C3Hg 0.9 4 0.201 3.784 0.547 0.193 44.17 0.046 1.0 4 0.201 3.784 0.547 0.193 49.16 0.040 0.8 3 0.208 4.181 0.553 0.186 36.87 0.059 P2， C3Hg 0.9 3 0.208 4.181 0.553 0.186 44.17 0.046 1.0 3 0.208 4.181 0.553 0.186 49.16 0.040 2.3. OH-PLIF image processing For each condition， 500 OH-PLIF images (949*639 pixels) are re-corded by the ICCD camera for statistical analysis. The raw images arecut to 114*583 pixels and then the flame front is extracted by an auto-threshold binarization method proposed in our previous study [33]. Inthis process， the OH-PLIF image was binarized after being filtered toremove the pixel noise， and then a threshold value was selected basedon the local pixel intensity. After superimposing the 500 binarizedimages， the mean progress variable， ， indicating the temperature ormass fraction， with =0 is the unburned gas and =1 is theburned gas， can then be obtained by the gradient of the superimposedimage. The turbulent flame front can also be extracted by every bi-narized image. The image processing is shown in Fig. 3(a) to (d). Therehave been many definitions of the mean flame brush thickness in theliteratures by using different progress variables [34-36]. Although thequantitative values may vary in different definitions， the qualitativetrends are similar [30]. As shown in Fig. 3(e)， the horizontal distancebetween the flame front leading edge=0.1， and the half burningsurface=0.5， is defined as horizontal flame brush thickness，8Th.The centerline distance between the flame front leading edge= 0.1，and the half burning surface=0.5， is defined as centerline flamebrush thickness， 8r，o [26，36]. 3. Results and discussions 3.1. Development of horizontal flame brush thickness If a flame can be seen as a passive interface between the unburnedand burned mixtures， its brush thickness will follow the Taylor’s dif-fusion theory， where te is the flame development time， tt，z is the integral length timescale. When the two times are comparable， the flame brush thickness is (a) Regimes of experimental conditions (b) Flame instability wavelength Fig.2. Regimes of experimental conditions and flame instability wavelength. (a) (b) (c) Fig. 3. Schematic of image processing and the definitions of flame brush thickness. The flame height is needed to calculate the development time of theturbulent flame. It is defined as the distance from the centerline of theBunsen burner outlet to the half-burning surface=0.5. The flameheight normalized by flame diameter， He>=0.5/D， decreases when theequivalence ratio increases from lean to stoichiometric condition forboth CH4/air and C3Hg/air turbulent flames， as shown in Fig. 4. Similarobservations have been reported in [30，36]. In turbulent Bunsen flame，the normalized flame height increases with the outlet velocity of theburner exit at a constant equivalence ratio. It also indicates that theturbulence intensity has weak impact on turbulent Bunsen flame height，as the flame height of P1 and P2 keep almost constant when the outletvelocity is fixed at U =3 m/s for both fuels. The ratio between the flame height， Hc>=0.5， and the outlet velo-city， U， is the flame development time， tp. Those flame developmenttime of CH4/air and C3Hg/air flames are given in Table 3. For all theexperimental conditions， the flame development time based on thismethod is less than the integral time scale. The horizontal flame brushthickness can be predicted by Eq. (2) and Eq. (3). The horizontal flamebrush thickness of CH4/air flames normalized by the burner diameter，8rh/D， with respect to the normalized axial distance from the burnerexit， h/D， are shown in Fig. 5. The horizontal flame brush thicknesspredicted by Eq. (2) (blue short dotted line) and Eq. (3) (red shortdotted dashed line) are also presented. It can be seen that the turbu-lence diffusion predicted by Eq.(2) is slightly higher than that of Eq.(3). For all the four series of conditions， the normalized turbulent flamebrush thicknesses increase with the normalized axial distance. At thefirst development stage of Bunsen flame， the variations of normalizedflame brush thickness are in good agreement with the prediction byturbulence diffusion. When the flame goes to the downstream furtherwhich means the flame develops longer time， the flame brush thicknessat different equivalence ratios begin to diverge. It increases when theequivalence ratio increases from lean to stoichiometric condition. Thevariation of horizontal flame brush thickness then deviates from theprediction which means that at this stage the turbulent flame is not onlydominated by turbulence diffusion. The normalized horizontal flamebrush thickness of C3Hg/air turbulent Bunsen flame in function ofnormalized axial distance is plotted in Fig. 6. We can see that theoverall variations of flame brush thickness in C3Hg/air flames are si-milar to that of CH4/air flames. From the similar variations of flame brush thickness of CH4/air and (d) (e) (a) Flame height of CH4/air turbulent Bunsen flames. (b) Flame height of C3Hg/air turbulent Bunsen flames. Fig. 4. Flame heights of CH4/air and C3H8/air Bunsen flame under variousconditions. C3Hg/air turbulent Bunsen flames， it can be concluded that there are atleast two stages during the turbulent Bunsen flame development at lowturbulence intensity. At the first stage， its development follows theprinciple of turbulence diffusion theory. At the second stage， the flamebrush thickness increases when the equivalence ratio becomes stoi-chiometric. This phenomenon suggests that the Taylor diffusion theory，which shows that the flame brush thickness increases roughly linearwhen the outlet velocity is kept constant， may can’t predict the turbu-lent Bunsen flames development in low turbulence intensity. Thisphenomenon will be discussed in detail in Section 3.3. The variations of the centerline flame brush thicknesses of CH4/airand C3Hg/air turbulent Bunsen flames are shown in Fig. 7. It can beseen that the centerline flame brush thickness is mainly controlled byoutlet velocity and laminar flame speed. It increases with the increaseof outlet velocity while it decreases as the equivalence ratio increasetowards stoichiometric condition. Similar observations can also befound in [26，36]. It also shows that turbulence intensity has marginaleffect on centerline flame brush thicknesses when the outlet velocitykept constant. Variations of the normalized centerline flame brush thickness withrespect to normalized flame height are presented in Fig. 8. It shows thatthe normalized centerline flame brush thickness varies linearly with thenormalized flame height and it decreases with increasing equivalenceratio. This linear relationship between the centerline flame brush andnormalized flame height will be discussed further in Section 3.3. 3.3. Whole development of turbulent Bunsen flame and its uniquecharacteristics In turbulent flames， such as V-type flame and Bunsen flame， as thecoming velocity is not normal to the mean flame brush， it will have amean tangential velocity along the mean flame position. This will resultin the flame wrinkles generated at a given spatial location to transferdownstream， as shown schematically in Fig. 9. Hemchandra [9] derivedthe wrinkles of turbulent V-shaped flame using G-equation. It should benoted that the derivation is based on flamelet assumption， in which ourexperimental conditions are located. By rewriting the G-equation on theflame coordinate (s-y-n)， the flame shape can be From Eq. (4)to Eq. (7)，is the flame surface wrinkle， fare randomshape functions of their arguments with zero mean and unity variance.Eq. (6) and (7) are the first and second order of flame surface shape.The left hand side of Eq. (6) and (7) have the form of a 1-D advectionequation which was defined along the nominal-flame surface. The ad-vection velocity is caused by the mean tangential flow velocity. Thusthe total flame surface wrinkles at any given point can be seen as asuperposition of flame surface perturbations generated by local flowperturbations and that from upstream points at previous time. This iscalled the non-local effect of turbulent flames， which is also calledmemory effect. CH4 C3Hg TtL F，中=0.7 F， 中=0.8 F， 中=0.9 F， 中=0.8 F，d=0.9 4F， 中=1.0 P1，U=2m/s 7.99 2.53 1.73 1.38 2.16 1.57 1.23 P1， U=3m/s 2.72 1.83 1.35 1.12 1.79 1.33 1.08 P1， U= 4m/s 1.88 1.61 1.20 1.12 1.62 1.16 0.97 P2，U=3m/s 2.01 1.86 1.37 1.15 1.69 1.24 0.99 0.25 h/D h/D h/D h/D Fig. 5. Variation of the normalized horizontal flame brush thickness of CH4/air turbulent Bunsen flames. This non-local effect is significant in the development of turbulentpremixed Bunsen flame. It can be seen in Table 2 that the wavelength offlame instability is smaller than the integral length scale， which is thelargest vortex in turbulence， under all experimental conditions.Therefore， the flame wrinkles at a given point caused by turbulence arelarger than that caused by flame instability. At the first stage of flamedevelopment， flame wrinkles at a given point are controlled by the localflame wrinkles， as the accumulated wrinkles generated upstream arevery small because of the short development time. This can be verifiedby that the flame brush thickness of turbulent C3Hg/air Bunsen flamesat P2 plate (U = 3m/s) is higher than that of P1 plate， as the turbulenceintensity of P2 is larger. Therefore， during this stage， the horizontalturbulent flame brush thickness follows the turbulence diffusion theoryin both strong and weak turbulence intensity. This prediction has beendemonstrated by the horizontal flame brush thickness in Figs. 5 and 6.This stage is labelled as I in Fig. 10， which can be called turbulencedominating stage. Here， it should be mentioned that there is no obviousdifference among the flame brush thickness in turbulent CH4/airBunsen flames at U= 3 m/s in Fig.10(a). This can be explained by thefollowing. As shown in Table 2， the instability wavelength of CH4/airflames is double than that of C3Hg/air flames. This means that CH4/airturbulent flames tend to be more wrinkling. When the velocity per-turbation is the same， the tendency of more wrinkling will offset thevelocity perturbation. By contrast， for C3Hg/air flames， the offset isrelatively small. Therefore， there is no obvious difference among theCH4/air Bunsen flames at U =3 m/s. It should also be mentioned that although the flame wrinkles based on Eq. (4)-(7) is referred to wrinklesnormal to mean flame position. The horizontal flame brush thicknesscan represent it because the horizontal flame wrinkleisf (s， y， 0，t)cosy in Bunsen flame. When y is small， the two wrinklescan be seen as the same， which is the case in our experiments. As the flame develops downstream to the second stage， the accu-mulated wrinkles generated upstream will become large. For strongturbulence intensities， although the upstream non-local effects may belarge， its effect on a given point at downstream will be hidden by locallarge velocity perturbations. This is why the experimental data of[26，37] follow the Taylor diffusion theory. However， in low turbulenceintensities where the local flame wrinkles caused by velocity pertur-bations are relatively small， the non-local effect will manifest atdownstream positions. At these positions， the accumulated wrinklescaused by upstream flame instability and turbulence probably will becomparable to local velocity perturbations. Therefore， the equivalenceratio will have an impact on the flame brush thickness. This stage islabelled as II and is non-local effect manifesting stage， as can be seen inFig. 10 for both CH4/air and CsHg/air flames. Figs. 5-7 show that the variations of horizontal flame brush thick-ness and centerline flame brush thickness experience the oppositetrends. Under all conditions， for constant outlet velocity， horizontalflame brush thickness increases with increasing equivalence ratio fromlean to stoichiometric condition， while centerline flame brush thicknessincreases by decreasing the equivalence ratio. This is because the flamewrinkles enter another stage at the tip where the flamelets begin to 0.25 0.25 h/D h/D 0.25 h/D h/D Fig. 6. Variation of the normalized horizontal flame brush thickness of C3H8/air turbulent Bunsen flames. intersect. The flame wrinkles， ftip， as shown in Fig. 9， can be divided intotwo parts： the horizontal wrinkles， fuip，h， and the vertical wrinkles， ftp，o.As the Bunsen flame is axisymmetric， the horizontal wrinkles will offset，while the vertical wrinkles will superimpose， thus making the center-line flame brush thickness depends mainly on the flame developmentdistance. This is also demonstrated by the nearly linear relationship ofthese two parameters in Fig. 8. As shown in Fig. 9， at the flame tip， theflame wrinkles can be expressed as.Fip=2fap (s，…)siny=2frip((d...siny，where， Frip， is the super-imposed flame wrinkles， ftip， is the flame wrinkles， s， is the flame de-velopment distance. It can be seen that the superimposed flame wrin-kles is influenced by burner diameter. That’s why many flameparameters， such as flame height and flame brush thickness， are nor-malized by the burner diameter in this paper. In turbulent Bunsenflames， horizontal flame brush thickness is dominated by both the localand upstream flame wrinkles， while centerline flame brush thicknessdepends mainly on flame development distance. The turbulent Bunsenflame tip is labeled as III： flame tip region， as shown in Fig. 10. The above analysis suggests that there may be three stages in thewhole development of turbulent Bunsen flame. At the first stage， theBunsen flame is dominated by turbulence diffusion. At the second stage，it is dominated by the coupled effect of turbulence and wrinkles gen-erated upstream due to non-local effect. These two processes are similarto that of turbulent V-shaped flames. At the tip of turbulent Bunsenflame， it is determined by flame development time. This is the uniquecharacteristics of turbulent Bunsen flame. 3.4. Flame front characteristics based on flame development time The local flame front curvature， K， and its PDF (probability densityfunction) distribution are usually used to quantitatively describe theflame front wrinkling. After the flame front is extracted， its cartesian co-ordinates can be stored. Then variations of the two parameters， x(s) andy(s)， with the path length variable s， can be obtained by a cubic splineinterpolation. Subsequently， the local front curvature can be computed， The flame curvature is defined positive if the flame element isconvex toward the reactants. In the curve-fitting scheme， appropriatechoice of interval length is of significance [38].Oversampling will shiftthe PDF toward large curvatures， as it will induce small-scale digitiza-tion noise. By contrast， relevant scales of the flame front will be ne-glected if the sampling rate is too low. In this paper， the interval lengthis set as the order of the laminar flame thickness defined by the steepesttemperature gradient. The interval length scale is chosen based on theinner cutoff scale. The inner cutoff scales is the smallest scales of flamefront wrinkles and invariant with respect to the turbulence conditions.Cintosun [39] reported that the inner cutoff scales obtained for theseflames from fractal analysis is within the range of 0.2-0.4 mm and thisis of the order of the flame thickness calculated by the steepest tem-perature gradient. For each condition， 250 single images are recordedat 200 equally spaced flame elements. These 50，000 data points aresufficient for statistical stability. 0 (a) Centerline flame brush thicknesses of CH4/airturbulent Bunsen flames. 0 (b) Centerline flame brush thicknesses of C3Hg/airturbulent Bunsen flames. Fig.7.The centerline flame brush thicknesses of turbulent Bunsen flames undervarious conditions. In turbulent premixed flames， integral length scale is the largestvortex which can interact with flamelets. Therefore， turbulent Bunsenflames with the largest integral length are predicted to have the mostwrinkled flame front when other parameters are constant. However， inFig. 11(a)， it can be seen that although the integral length scale of thefirst serial experimental conditions (P1， U= 2 m/s) is almost twice thanthat of other three serial experimental conditions， their PDF distribu-tion of turbulent flame front curvature are almost the same. This maybe explained as follows. In Section 3.3， we have analyzed that the non-local effect of turbulent Bunsen flame will transfer flame wrinkles todownstream and this effect is controlled by flame development time. Inthis way， the turbulent Bunsen flames (P1， U= 2 m/s) will suffer lessnon-local effect because its flame height is smaller compared with otherthree serial experimental conditions. This means that statistical de-scription based on the whole development time for different turbulentBunsen flames may can not demonstrate the flame-vortex interaction inturbulent flame. In this paper， a new method is proposed to minimizethe non-local effect on the statistical data of the turbulent Bunsen flame.Since the non-local effect is controlled by flame development time， we (a) Normalized centerline flame brush thickness with respectto normalized flame height of CH4/air turbulent Bunsen flames. (b) Normalized centerline flame brush thickness with respectto normalized flame height of C3Hg/air turbulent Bunsen flames. Fig. 8. Variations of the normalized centerline flame brush thickness with re-spect to normalized flame height. can calculate the statistical data based on the same development timeperiod instead of the whole development time. Fig. 11(b) shows thePDF distribution of turbulent Bunsen flame front calculated at the samedevelopment period (tp =4-8 ms). This time period is decided becausethe turbulent flame has entered the non-local effect manifesting regionfor all of these four conditions. It can be seen that turbulent flames withthe largest integral length (P1，U= 2 m/s) manifest more wrinkles thanother three turbulent Bunsen flames. This means that statistical data ofturbulent Bunsen flame calculated by this new method may be better atdescribing flame-vortex interaction. This method can be used in tur-bulent Bunsen flames and V-shaped flames to get a better statisticalanalysis of vortex-flame interaction in turbulent flames. 4. Conclusions A systematic study about the effects of turbulence characteristics，include outlet velocity， turbulence intensity， and laminar flame char-acteristics， include laminar flame speed， flame intrinsic instability， onthe evolution of flame brush thickness of turbulent premixed CH4/airand C3Hg/air Bunsen flame were investigated under low turbulenceintensity. The turbulent flow field and instantaneous flame front weremeasured using the hot wire anemometry and OH-PLIF technique. Theturbulence intensity was controlled by changing outlet velocity andperforated plates. All experimental conditions are located in the Fig. 9. The non-local effect of flame-turbulence interaction on the turbulentBunsen flame. (a) The whole development of CH4/air turbulent Bunsen flames (b) The whole development of C3Hg/air turbulent Bunsen flames. Fig. 10. The whole development of CH4/air and C3H8/air turbulent Bunsenflames. .(a) Curvature PDF distribution calculated by whole development period. (b) Curvature PDF distribution calculated by the same development period. Fig.11. Statistical analysis of turbulent Bunsen flames with two methods. wrinkled and corrugated flamelet regimes. The main conclusions aresummarized as follows： 1. The normalized flame height of turbulent Bunsen flame increaseswith the augment of outlet velocity while decreases with increasingequivalence ratio from lean to stoichiometric， it has a relativemarginal effect from turbulence intensity. 2. The non-local effects result in three regions for the whole develop-ment of turbulent Bunsen flame， the turbulence dominating region，non-local effect manifesting region and flame tip region. 3. At the turbulent Bunsen flame tip， as the Bunsen flame is axisym-metric， the horizontal flame wrinkles offset and the vertical flamewrinkles superimpose， which results in the centerline flame brushthickness is dominated by flame development time. It increaseslinearly with flame height， which is a unique characteristic of tur-bulent Bunsen flame. 4. A new method calculates statistical parameter of turbulent Bunsenflame based on the same development time is proposed to minimizethe non-local effect on the statistics of turbulent Bunsen flame frontstructure. This new method is better at describing flame-vortex in-teraction in turbulent flame compared with the conventionalmethod. Acknowledgements This study is supported by National Natural Science Foundation ofChina (Nos. 51776164， 91441203). The support from The State KeyLaboratory of Engines， Tianjin University (K2017-03) is also appre-ciated. ( [ 1 ] S ak a i S ， R o t ha me r D . E f f e c t o f e thanol bl e nding o n p a r t ic ul a t e f o r mat i on f r o m p remixed c o m bustion in spark-i g ni tio n e n g i ne s. F u e l 2 017；1 9 6：154-68 . ) ( [2] O rbay R C， N o g enmy r K J， K l i n g ma nn J ， B ai XS. S wi r li n g t urbu l ent f lows i n a c o m - b u s t ion c ha m be r w i t h an d w i thout h e at r e le a se . F u el 2 013 ；1 0 4 ：133-46. [3] B ecker L G， K osaka H ， Bo hm B ， D oo st S， K n appstei n R ， Ha b e r me h l M ， e t a l. ) ( E x peri m e n t a l in v e st iga t ion o f f l a me s t a b i l iza t io n in s id e t h e q u ar l of an o xyf ue l s w i r l b ur n er . Fuel 2017 ； 2 01 ： 124- 3 5 . ) ( [4] Y uen FT C ， G ul d er OL. T urb u lent p remi xe d fl a me f r ont d y n a m ic s a nd i mpl i cati o ns f or limits of flam e l et hy pothesi s . Pro c Co m b us t I nst 2013 ；34 ( 1) ： 1 393- 4 0 0 . ) ( [5] 1 C hen Y C ， B i l g er R W . E x p e r im e n t al in v estig at io n o f t h re e - d ime n s i on a l f l a m e-fro n t s tr u c t ur e in pre m i xed t urbule n t c o m b u s ti on ： I I . L e a n hyd ro ge n /air B u nsen flames . Combust F l a m e 2 004； 1 38( 1 - 2 ) ： 15 5 - 74. ) ( [6] S jo ho l m J， R o s e l l J ， L i B， Richt e r M ， L i Z， B a i X S， e t al. S i mul t an eo u s vi s u a l i z at i o n o f O H， C H ， C H2 0 and tol u ene P L IF i n a met h ane j e t f l a m e w i t h va r y in g d eg r ees of t u r b u l e nc e. P roc Com b us t Inst 2 013 ； 34 ( 1 ) ： 147 5 - 8 2 . ) ( [7] D ri scol l J. T ur bule n t p re m ixed c o m bust i o n ： fl a m el e t str u c t u r e a nd i t s e f f e c t on t u r b u l e n t b urni n g v elocities . Pr o g E nerg Combust 2 0 08；34 ( 1 ) ： 9 1 -134 . ) ( [8] L i p at n i kovJC A N. T u rb ule n t f lame s pe ed a n d t h i c kne s s ： phen n o menology， e v a lua- t i o n ， a nd a p p li ca t i on i n mu l t i -di m ensi on al s im ul a ti o n s. P r og E n e r g Com b us t 2 002；28：1-74. ) ( [9] L i eu wen T . L o c a l c o n sump ti on speed o f turbulent premix ed fl a m e s - a n an a lysi s o f “ me m or y e f f e c t ” . C o m b u s t F l a me 2 0 10； 1 5 7： 9 55 - 65. ) ( [10] .A s hurs t WT. Da rri e us - L a ndau in s t a b i l ity， g r owi n g c y c l o i d s a n d expa n din g f l ame a c cel er at ion . Combust Th e o r Model 1997 ； 1 ( 4 )： 405 - 28 . ) ( [11F ] r a g n e r R ， H a l t er F ， M az e llier N ， C h auvea u C ， G okalp I . I nves t igat i o n o f p r e ssu r e e f f e c t s o n the sma l l sc a l e wr i nk l i ng o f t u rb u le nt t 9 p r e m ixed B unsen f l a m es. Pr ocCombust I n st 2 015； 3 5(2)：1 5 27- 3 5 . ) ( [12] L e e TW， S h an k l an d R ， F e n t on M . F l a me c u r v a tur e st at is t ic s in a x is ym m etr i c tu r- b u le n t j e t f la mes . Co mbu st S ci T e c hnol 1 995；1 0 8 (1 -3) ： 3 1-4 6. ) ( [13] V ey na nte D， D u clos J M ， P i a na J . E x p eri m ental analys i s o f f l a me l et m o dels f o r p remi xe d tu r bu l e n t co m b u s t io n . Pr o c C o m bust In st 1 9 9 4； 25 (1) ：12 4 9 - 5 6. ) ( [14] B i e l ert U ， K l ug M ， A d o m ei t G. Ap p l ica t io n o f f r o n t tr a ck i n g t ech niq ue s to t he t u r - b ul e nt c o mb u stion p r ocess e s i n a si n g le strok e de vi ce . Comb u st F l ame 1 9 9 6； 1 06( 1)：1 1 -2 8. ) ( [1 5 ] L i u K ， B urluka A A ， S hep p ard CGW. Tu rb ul e n t fla m e and mass b urn i n g ra t e i n a s pa rk ig nitio n en gi n e. F u el 20 13 ； 10 7 ：2 0 2 - 8 . ) ( [16] D i ne s h K K J R ， Jen k in s K W ， K i rkp a tri ck M P， M alal a sekera W . M odelli n g of i n- ) ( s t a bi li tie s i n tu r b ulen t s wir l ing fl a m e s . F ue l 2 0 1 0 ； 8 9 ( 1 ) ： 10 -8. ) ( [17 l ] Zi mo n t V L . G as p r em i xed com b u st i on a t hi g h tu rbu l ence . Tu r b ul e nt fla m e c l o s ur e comb u stion m odel . E xp The r m F l u i d S ci 2000； 21 ( 1- 3 )： 1 79-86. ) ( [18] L i p a t n ikov AN . D ev e lop i ng pr e mixe d tu rbu l ent f l a m e s Pa r t I A self si milar reg i me of ) ( f l a me p r o pa g atio n . C o m b u s t S c i T e c h no l 2 001 ； 1 62 ： 8 5-1 1 2. ) ( [19] G o u l d i n F C ， M i l e s PC. C h e mic a l c l o s ure a n d b u rn in g ra t es i n p r emix e d t ur b ulen t f la m es. Combus t F l a me 1 9 9 5；1 00(1 - 2 ) ： 2 02-10 . ) ( [20] G o i x P ， P a r an t hoe n P， T r i ni t e M . A t omog raph ic s t udy o f m e asur e m e nt s in a V - s h ap e d H/ a i r f l a m e a n d a l a gr an g i a n in t e r pretation of t h e t u r b u len t f lame br u sh e v o l uti o n . C omb u s t Fl a me 1 9 9 0 ；81( 3 -4) ：229 - 41 . ) ( [21 ] K w o n S ， Wu MS ， D r i scoll JF， F a e th G M . F la m e surf a ce p r ope r t i es of p remix e d f l a mes ) ( i n isotr o pi c t urbu l enc e ： mea s u re ment s a nd n um e ric al s i m u la t ions . Co m bust Fl a m e 1 992 ；8 8(2 )：221 - 38. ) ( [22] 1 M o reau P . T u r b u l e n t f l a m e d e v el o p m en t in a hi g h v e lo ci t y pr e m i xed f low . A erospace Sc i Me et 1977：7 7- 9 . ) ( [23] Chen YC ， P e t e r s N， S c h nee mann G A ， W ruck N ， R en z U ， M a n s our M S. T h e d et ai le d f lame st r u c t ure of hi ghl y s tr e t c he d turbulen t premi x e d meth a n e- ai r f la m es. Combust F la m e 1996； 1 07( 3 )：223 - 4 4 . ) ( [24] P o pe SB. Tu r b ul e n t flows . M ea s Sci T ech nol 20 01；1 2 (11) ：20 20-1. ) ( [25] G oix P ， P a r an thoen P . A t o mo g ra ph ic s t udy o f m e as u rem e nt in a V -s h a p ed H 2 / ai r f lame an d a La g r a n g i a n i n terpre ta tion o f t he turb u lent fla m e b r u sh e vo l ut ion . C o mbust F la m e 1 9 90； 8 1： 2 29-41. ) ( [26] T a m adon f a r P ， G i i l d er O L . F l ame bru s h c ha r acte r i stics an d b u r ni ng v e lo citi e s o f pre mi x e d t urb u len t me t han e /a ir B u nse n f l a m es . Combust Flame 2014；16 1 ( 12 )：3 1 54 - 65. ) ( [27] F o gl a N， C r e t a F ， M a ta lo n M . T h e t u r b u l en t fl a m e s p e e d f o r l o w - to-m o der a te tur - b ul e nc e i n t e nsi ti es ： h y drodyna m ic t heo r y v s. ex p e r i men ts . Co mb u st F l a m e 2 017； 1 75 ： 1 5 5 -6 9. ) ( [28]J i n W ， W an g J， N ie Y ， Yu S ， H u ang Z. Ex p e r im e n ta l s tu d y on f la m e i n stab i l i ti e s o f l am i nar p r emix e d CH4 /H2 / air n on-adi a bati c fl a t f l a me s . F uel 2015； 1 5 9 ：599-60 6 . ) ( [29] Z han g M ， W a ng J， W u J， W e i Z， H u a n g Z， Ko b a y a shi H . F l am e fron t s tru c tur e o f t ur b u lent p r e m ix e d f l a m es of s y ng as o x yf uel m i x tur e s. I n t J H y d roge n E n e r gy 2014； 3 9( 1 0) ：5 176-85. ) ( [30] T a m a do nfar P ， G i i lder O L . E ff ect s of m i xt u re c om pos i tio n and t u rbulence in t e n sit y o n fl a m e f ront s tructure a n d b ur ning vel oc i t ie s o f prem i xed tu rb ul e n t hy dr o c a r b o n / a i r Bunse n f lame s. Combu s t F lam e 201 5； 16 2 ( 1 2 ) ： 441 7- 41. ) ( [31] B ec htold J K ， M atalon M . T h e d e pe n denc e o f t he M a r k st ein l engt h on st oic hi omet r y. C o mbust F la m e 2 001； 1 27( 1 - 2 )：1906- 1 3 . ) ( [32] X i e Y ， W ang X ， B i H， Yua n Y ， W a ng J ， Hu ang Z ， et al. A c o mpreh e nsi v e re v ie w o n l am i nar s p h er i ca l ly pr em i x ed f l am e p r o p aga ti o n o f s yngas .F u el P r o cess T e chnol 2 018； 1 81： 9 7- 11 4 . ) ( [33 ] Z h ang M ， W a ng J， Xi e Y ， W e i Z ， Jin W ， H u an g Z ， e t al . M e asu re ment on i n - s tantaneous f l am e f ront s t r u c ture of tu r b ul e nt pr e mixed CH4 /H 2/ a i r f lames . E xp T he rm Flu i d S c i 2 01 4； 5 2 (1) ： 2 88-9 6 . ) ( [34] L e e G G ， K a n g YH， K oba y a s h i H . M e asureme nt an d an a lys i s of f l ame s u rf a ce d e n s i t y f or turbul e n t prem i xed c ombustion on a n ozzle- t y p e b urner. C ombust F l ame 2 000； 1 22( 1 ) ： 4 3 -57 . ) ( [35] T a lbot M NG S . Cha r a c t e ri z a t ion o f t h e d e n sity f l uc t ua t io n s i n tu r b ulent V- s h a pe d pr e m i xed f la m e s . Comb u s t Flame 1 9 86； 6 4(3) ： 299- 3 08. ) ( [36] G r i e b e l P ， S ie w ert P， J a n s oh n P. Fl am e c h a r acter i s t ics o f turbule n t le a n p r e m i xe d m et h a n e / a ir f l a m es a t h i g h pr e s s u r e ： t ur b ulent f l a me sp e e d a n d fla m e b r u shthic k nes s . P r o c Combust I nst 200 7 ；3 1 (2) ：30 83 -90 . ) ( [ 3 7] R enou B ， M u ra A， Samson E，Boukhalfa A. 1 8th international Colloquium on theDynamic of Explosions and Reactive system. Seattle， Washington， USA； 2 001. ) [38]Soika A， Dinkelacker F， Leipertz A. Pressure influence on the flame front curvature ( o f turbul e n t premi x ed f l am e s： c o mp a ris o n b e tween e x p e ri m ent a n d th e o r y. C ombust F la m e 2 0 0 3 ；1 32(3)：4 5 1-6 2 . ) ( [39] C intosun E ， Sma l l wood G J ， Gi i lder L O m er . F l a me sur f a ce f r a c tal c h a r a c t e ristic s i n pr em i xed tu r bulent c om b ustio n a t h i gh turb u lence in te n siti e s . AIA A J 2 0 07； 4 5 (11 ) ： 2 785-9 . ) https：//doi.org/j.fuel. Flame brush thickness is a significant parameter that demonstrates the evolution of turbulent premixed flame. In this paper, the flame height and flame brush thickness of lean turbulent premixed Bunsen CH4/air and C3H8/air flames were investigated under low turbulence intensities. The turbulence flow field is generated and controlled by changing the outlet velocity and perforated plates, and it is measured using a hot wire anemometry system.The turbulent flame fronts are detected by OH-PLIF technique. Results show that under fuel-lean conditions, the characteristic turbulent Bunsen flame height and centerline flame brush thickness decrease with increasing equivalence ratio, while they increase with the augment of outlet velocity. Turbulence intensity has marginal effect on these two parameters. The variations of flame brush thickness indicate that there are three regions during the whole development of turbulent Bunsen flames under low turbulence intensity. The first is the turbulence dominating region, where the variations of horizontal flame brush thickness follow the turbulence diffusion theory. The second is the non-local effect manifesting region, where the equivalence ratio, which controls the flame instability wavelength, influences the horizontal flame brush thickness. The third is the flame tip region where the flamelets intersect and make the radial wrinkles offset while the axial wrinkles superimpose.This unique tip region of turbulent Bunsen flame can be demonstrated by the linear relationship betweenits flame height and centerline flame brush thickness. Turbulent premixed Bunsen flames suffer different non-local effects during the flame development to downstream. To minimize the non-local effect on the statistics of turbulent flame front structure, a new method based on the development time period is proposed. Compared with statistical analysis based on the whole development time, this method can better demonstrate the flameturbulence interaction.