Jtam.dvi JOURNAL OF THEORETICAL AND APPLIED MECHANICS 45, 3, pp. 489-503, Warsaw 2007 PIV ANALYSIS OF TURBULENT FLOW IN A MICRO-CHANNEL Sławomir Błoński Tomasz A. Kowalewski Institute of Fundamental Technological Research, Polish Academy of Sciences, Warsaw, Poland e-mail: tkowale@ippt.gov.pl Turbulent flowofwater in a short 0.4mmhighmicro-channel of an emul- sifier is investigated experimentally using a micro-PIV technique and comparedwith numerical predictions. Themicro-flowmeasurements are based on epi-fluorescence illumination and high-speed imaging. Velocity fields obtained from themeasurements and direct numerical simulations indicate thatflowturbulization is delayedanddevelops only at the outlet region of the micro-channel. Key words: micro-channel, micro-PIV, turbulence 1. Introduction Turbulent flow is commonly used for the emulsification process in indu- strial applications. The process is very effective but produces relatively non- homogenous emulsions. It is usually assumed that the final droplet size is related to turbulent kinetic energy of the flow. Themain aimof the present study is to investigate the flow structureusing themicro-PIV experimental technique for the flow in a flat, continuous stream homogeniser (Błoński et al., 2007; Kowalewski et al., 2006). The micro-PIV technique allows for full-field analysis of the velocity fields in a narrow gap formed by rapid contraction of the emulsifier, offering a unique opportunity to verify assumptions on the development of the turbulent flow. In the last decade,micro particle image velocimetry (micro-PIV), as a non- intrusive technique formeasuring flowfields, has been successfully adapted for measuring flowfieldswithinmicro fluidic devices withmicron-scale resolution, e.g. Meinhart et al. (2000). For higher temporal resolution, Shinohara et al. (2004) used a high-speedmicro-PIV technique by combining a high-speed ca- mera and a continuous wave laser in order to investigate transient phenomena 490 S. Błoński, T.A. Kowalewski in micro-fluidic devices. The micro-PIV technique refers to the application of PIV to measure velocity fields of fluid motion with length scales of the order of 100 micrometers and with spatial resolution of individual velocity measu- rements of the order of 1-10 micrometers. Due to small dimensions, the light sheet technique, typical for macro-scale PIV, cannot be applied for flow illu- mination, and the whole investigated volume is flooded with light. Images of individual tracers, necessary forPIVevaluation, are obtainedby taking advan- tage of fluorescence, removing by appropriate filters diffused light originating from the bulk illumination. Fundamentals of the micro-PIV method were established by Santiago et al. (1998), who used a mercury arc lamp to continuously illuminate 300nm fluorescent tracers. Using amicroscope, an image intensifiedCCDcamera and narrow wavelength optical filter, they obtained discrete images of fluorescent tracers. Correlation analysis was then applied to the particle image field, pro- ducing a regularly spaced velocity field. The flow fields measured under the microscope is characterized by high relative velocity due to large spatial ma- gnification of recorded tracer displacements. It limits themagnitude of velocity to about 0.1mm/s for epi-fluorescent systems equippedwith amercury lamp. A double pulsed Nd:Yag laser applied for illumination allows for a very short illumination time (5ns), enormously extending the velocity range of evaluated under a microscope flow fields. Themajority of investigations performed in micro-channels are limited to steady, low Reynolds number flows. It is mainly due to experimental constra- ins, which demand high speed imaging systems and high pressures to enforce the flow. The present paper reports on micro-PIV measurements performed for flow velocities of several m/s in a channel of only 400µm height. To our knowledge, there are very few reports onmicro-flowmeasurements in this ran- ge of parameters. Recently Li and Olsen (2006) investigated the transition to turbulent flow in amicro-channel. Their experiment rejects earlier suggestions of possible early transition to turbulence in micro-channel flows. In fact, the main result of our paper claims even possible delayed transition to turbulence. At high Reynolds number (Re = 6770), the flow measured within our short micro-channel remains quasi-laminar. 2. Experiment The flow of purewater was studied in a flatmodel of an emulsifier. It is a two dimensional substitute of the central cross-section of the axially-symmetric emulsifier thoroughly investigated in a separate study (Brenn, 2005; Steiner et al., 2006). The emulsifier consists of a small channel formed between two PIV analysis of turbulent flow in a micro-channel 491 glass plates anda triangular processing element creating rapidflowcontraction (Fig.1).Theflat geometry and twoglasswindowspermit application of optical methods for measuring flow velocity fields (PIVmethod). Fig. 1. Geometry of the processing element of the emulsifier model and location of the analysedmicro-channel. The micro-channel size: height 0.4mm, width 15mm and length 1mm. The flow inlet is indicated by an arrow: the inlet and outlet heights are 1.5mm, 7.5mm, respectively Dimensions of the gap between the glass plates and processing element (triangular obstacle) are 0.4mminheight, 1mmin length and15mminwidth. The processing element forms two inlets of a rectangular shape (upper and lo- wer); they are 15mm wide and 1.5mm high. After passing the gap, the flow geometryabruptlyexpands,filling15mmwideand7.5mmhighoutlet.The to- tal length of the inlet segment is 97.5mmand the length of the outlet segment is 78.5mm. These physical dimensions were implemented as a computational domain used to simulate 3Dflow in the emulsifier. However, experimental and numerical data presented in the following, deal with the central cross-section of the micro-channel and a short part of the outlet. The inlet and outlet of the channel are connected through 8mm tubes with the liquid supply system. Pressurized nitrogen was used to pump the liquid through the homogeniser to the collecting bottle. The flow rate was set by varying the reservoir pressure and using the system of valves. The exact value of the flow rate was obtained bymeasuring the time necessary to fill up a calibrated quantity in the collec- ting bottle. The flow rate used in the experimentwas QV =0.204dm 3/s. The same flow rate was used in the numerical models. The flowReynolds number is defined as Re= ρlV µ where ρ,µ indicate density and viscosity ofwater, l,V – height of the channel andmean flow velocity. For the given flow rate, the average flow velocity in the 8mm inlet tube is nearly 1m/s, and it increases to about 17m/s in the gap of 0.4mm height. 492 S. Błoński, T.A. Kowalewski Hence, the characteristic Reynolds number based on the inlet tube diameter and the micro-channel height varies from about 8000 to 6770, respectively. These values are high enough to expect a transition to the turbulent flow regime. Fig. 2. Scheme of the micro-PIV experimental set-up. The investigated channel is illuminated by Nd:Yag laser through the microscope lens. The same lens is used for imaging a selected flow plane, recorded by the CCD camera The main part of the experimental set-up consists of a microscope, an emulsifier with transparent windows, a laser light source and a digital camera (Fig.2). The flow was examined using an epi-fluorescence microscope (Nikon Eclipse 50i) equippedwith the 10× (NA0.3/WD17.30mm)microscopic lens. The flow observed under the microscope is characterized by large relative di- splacements; therefore application of high-speed recording techniques becomes essential. In our case, the image width of 1280 pixels corresponds, for the hi- ghestmagnification ratio, to the object dimension of 0.172mm.With the flow velocity of 1m/s, the illumination time necessary to freeze themotion is below 100ns. Such a short illumination time was achieved by using a pulsed light of Nd:YAG laser, delivering 30mJ energy at 532nm wavelength (New Wave Research, Inc.). For recording the images, a high-resolution (1280×1024 pi- xels) 12bit PCO SensiCam camera was used. When coupled with the double pulse laser it permits acquisition of two images at the minimum time interval of 200ns, exposition time of 5ns, and about 3.75Hz repetition rate. The PIV recording system installed on 3GHz Pentium 4 computer with 3GB RAM enabled us to acquire over 200 pairs of images during a single experimental run. Themicro-PIVmeasurements were performed for pure water seeded with fluorescent tracers, polystyrene spheres of 2µm in diameter (Duke Scientific PIV analysis of turbulent flow in a micro-channel 493 Inc.). The particle volumetric concentration was very low (< 0.0001%wt), hence they did not affect the flow structure. Particle ImageVelocimetry (PIV) based on correlation of pairs of images was used to evaluate instantaneous velocity fields in the channel. Unlike typical PIVmethods, themicro-PIVdoes not utilize a thin laser sheet to illuminate the seeding particles. The whole investigated volume was flooded with the laser light using a beam expander and themicroscopeobjective (Fig.2).Once theparticles are exposed to532nm light (green) from the laser, they emit red lightwith the emissionmaximumat 612nm, as specified by the supplier.The time-span inwhich particles continue to fluoresce after application of the laser pulse is of the order of nanoseconds, somotion induced blurring of the particles does not occur in the PIV images. Two low pass filters, mounted between the objective and the camera, only enable the fluorescent red light to pass, while preventing the green laser light to be detected by the camera. The micro-PIV images present well detectable bright spots of the seeding particles. Only particles being within the depth of focus are recorded. Particles that are out of focus add background noise, limiting the applicability of the technique to thin layers of the fluid (max. 10-15mm). The flow was illuminated and observed through the upper window of the channel. By traversing the field of observation in the horizontal and vertical direction, the position of the interrogated flowplanewas selected. The vertical resolution depends on the depth of field of the objective. For the micro-PIV experiments performed using the 10× objective, the vertical resolution was estimated to be 10µm and each PIV measurement covered an area of about 0.7mm×0.55mm.Thehorizontal resolution of the velocity fieldmeasurements for this objective was 0.5µm. The accuracy of the velocity measurement de- pends on several experimental factors (quality of the images, seeding concen- tration, particle displacement) as well as on the vector evaluation procedure. Using in-house developed software and by evaluating a uniform, predefined flow of water through the micro-channel, the error of velocity measurement was estimated to be below 5%. As results, five sets of velocity fields were obtained from the micro-PIV measurements – each of them at different locations within and behind the micro-channel (Fig.3). The area interrogated by the PIV method was in all cases located in the mid-plane between side walls. In the coordinate system described in Fig.3, themicro-channel extends from x=−1mm to 0.Measu- rements at positions P1 and P2 were obtained for the centre plane of the gap (y=−0.2mm) and used to evaluate velocity profiles along the x-direction in thevicinity of the gap entrance and the gap exit. It extends from x=−1.4mm to x=−0.7 for the location P1 and from x=−0.35mm to x=0.35mm for the location P2. Measurements at positions P3, P4 and P5 were performed for several y planes, changing the focal distance of the microscope from the 494 S. Błoński, T.A. Kowalewski vicinity of the upper glass wall (y=0) down to the centre plane of the outlet channel (y=−3.75mm). It permitted us to evaluate development of the flow structure along the y-direction at three investigated locations, i.e. 1mm,3mm and 8mm behind the gap. Fig. 3. Schematic drawing of the emulsifier with the coordinate system and locations where the velocity fields were measured by the micro-PIVmethod (P1-P5). The beginning of the coordinate system is placed at the exit edge of the processing element (x=0) attached to the top wall (y=0), and located in the symmetry-plane (z=0). P1, P2 indicates PIV interrogation planes located in the mid-height of the gap (y=−0.2mm) covering partly the gap inlet and outlet, respectively. P3-P5 indicate location of PIV interrogation planes used to measure velocity distribution from the top wall to the flow axis (y=0 to y=−3.75mm). They are located at distances x=1mm (P3), x=3mm (P4), and x=8mm (P5) from the gap exit In order to quantify the turbulence in the channel, the ensemble-averaged velocity fields were calculated at each analysed location. For locations P1 and P2 (within the gap), the micro-PIV measurements were repeated up to hundred times at the same position. For locations P3, P4 and P5 (behind the gap) measurements were repeated 28 times for each of about 20 planes of se- lected channel depths. The velocity was averaged over the velocity field and then over all 28 measurements taken at the interrogated location P3-P5 (over 100 measurements at P1 and P2). Velocity measurements performed at two positions P1 and P2 within the gap indicate that the flow through the gap is practically steady, no large tem- poral flow field fluctuations could be observed for these two locations. For position P1, the streamwise velocity of the flow rapidly increases from 8m/s to 16m/s (Fig.4 – P1) in the vicinity of the entrance to the gap, but it rema- ins steady in time. Despite high velocities flow seems to be laminar. Velocity measurements for location P2 (Fig.4 – P2) showflow development in the vici- nity of the gap exit. Just behind the gap, the streamwise flow velocity rapidly decreases from 18m/s down to 16m/s. Some initial spatial flow perturbation becomes visible here, indicating the beginning of transition to turbulence. PIV analysis of turbulent flow in a micro-channel 495 Fig. 4. Velocity vector field and velocity magnitude |V | contours measured 0.2mm below the top wall at the gap entrance (location P1), at the gap exit (P2), and 8mm behind the gap (P5); image displays the velocity field for the area of 0.7mm×0.55mm Themain reason of the observed flow acceleration and deceleration is the rapid change of the channel height in the vicinity of the processing element. The channel height converges from 1.5mm before the gap down to 0.4mm in the gap, and expands to 7.5mmbehind the gap. Due to the fluid acceleration anddeceleration, strong shear stresses areproduced in theflow.Shortlybehind theprocessingelement, theflowvelocity apparently changes its character, both flowdirection andvelocity amplitude stronglyvary in time (comp.Fig.4 –P5). The flow structure behind the processing element was analysed at three locations (P3, P4, and P5) along the x-axis, and for about 20 positions along the y-axis (comp. Fig.3). At each position, the PIV velocity field was acqu- ired for a small interrogation area of 0.7mm×0.55mm, delivering nearly 106 individual vectors. In addition, measurements were repeated 28 times for the same location. Assuming homogeneity of the velocity field within the small area covered by the PIV measurement, the ensemble-averaged velocity and velocity fluctuationswere evaluated at each interrogated position using a huge set of measured velocity vectors. These micro-PIV measurements were used to evaluate profiles of the time and space averaged x-component of velocity along the y-direction, i.e. channel height (Fig.5). The coordinate value y =0 corresponds to the surface of the upper glass wall (comp. Fig.3), and y=−3.75mm to the plane of symmetry of the emul- sifier. Symbols 〈Vx〉 indicate themean values of the velocity x-component Vx. It is worth to note the high velocity ”jet flow” present in the vicinity of the topwall. The reversal flow can be found closer to the channel symmetry plane indicated in the profiles as negative values of the flow velocity (comp. Fig.5 – P5). For evaluating turbulent characteristics of the flow field behind the pro- cessing element, the velocity fluctuations V ′ x and V ′ z were evaluated from the experimental data. The sum of their squared values was used to represent the turbulent kinetic energy (tkexz) of the flow. It was obtained from both ava- 496 S. Błoński, T.A. Kowalewski Fig. 5. The time-averaged x-component of the flow velocity 〈V x 〉 (solid line), and fluctuations tke xz (dashed line) representing turbulent kinetic energy. Data obtained frommicro-PIVmeasurements performed along the y-direction for locations P3, P4 and P5 ilable horizontal components of the velocity, however streamwise fluctuations were dominating in all investigated cases. Figure 5 shows variation of tkexz (dashed line), representing turbulent kinetic energy calculated from measurements performed at selected locations P3, P4 and P5. Profiles of tkexz indicate that turbulent fluctuations emerge just behind the gap and persist along the x-direction. The tkexz reaches its maximumabout 0.3mmbelow the glass wall. This is the region, where intense mixing occurs between high speed fluid arriving from the gap and low velocity recirculating flow behind the processing element. The peak value of the tkexz is observed for the last interrogated positionP5, i.e. about 8mm from the gap. 3. Numerical simulations Numerical simulations of the flow of water were done using CFD codeFluent (Fluent Inc.) (Błoński et al., 2007). Two types of numerical simulations were performed. To evaluate flow fluctuations Direct Numerical Simulation (DNS) was performed by solving the Navier-Stokes equations without any turbu- lence model. In the separate computational runs, the averaged Navier-Stokes equations were solved using the k-ε turbulence model. These computations delivered the averaged flow structure and the turbulent dissipation rate in the PIV analysis of turbulent flow in a micro-channel 497 vicinity of the processing element (Brenn, 2005; Błoński et al., 2007; Steiner et al., 2006). The DNS model allows one to obtain an accurate, unsteady solution to unmodified Navier-Stokes equations by resolving the whole range of spatial and temporal scales of the turbulence fromthe smallest dissipative scales (Kol- mogorov scales) up to the integral scale associated with motions containing most of the kinetic energy. The numerical domain used in the simulations had to cover full 3D geometry of the physical channel. All the spatial scales of the turbulence had to be resolved in the computational mesh. Hence, a very fine mesh and small time steps were used. A direct numerical simulation (DNS) performed with the classical finite volume code implemented in Fluent is time consuming and vulnerable. Ne- vertheless, it appeared that for the investigated geometry it was possible to obtain reasonable solutions reproducing typical for the turbulenceflow charac- teristics. These resultswere comparedwith the outcome of the steady, average turbulencemodel, where the flow symmetry was assumed, and only one quar- ter of the model geometry was used for the computational domain. The turbulent flow model assumes a priori that the flow is turbulent and properly described by the averaged variables using Reynolds hypothesis. The reliability of the simulation results depends on several empirical parameters assumed to be applicable to a specific flow configuration. The main advan- tage of using the average turbulence models is their computational efficiency. The computations were performed using the standard k-εmodel with default parameters and an enhanced wall treatment. These parameters and appro- priate computational meshwere selected after performing several test calcula- tions. The averagemodel enables one to assumeflow symmetry, therefore only one-quarter of the physical domain was implemented into the computational scheme, extremely improving computational efficiency. In both computational models, the structural hexahedron mesh with a boundary layer was generated in the gap and in the vicinity of the proces- sing element. The tetrahedron mesh was used in the remaining parts of the computational domain, reproducing all geometrical details of the emulsifier. Several mesh resolution tests were performed to identify the optimal mesh, being a compromise of the computational time and numerical accuracy. The most critical flow region is within the gap. The optimal mesh sought using the convergence grid index and applying mesh adaptation technique for the gap region was found for the turbulentmodel to consist of nearly 0.5mln grid cells. For the DNS model, it was found that sufficient accuracy was obtained for the gapmesh of 50×60×50 nodes and the total number of computational cells for the whole domain amounting over 1.7mln. A direct numerical simulation allows for an unbiased analysis of the turbu- lent flow to extract physical fluctuations and toperformuser-definedaveraging 498 S. Błoński, T.A. Kowalewski of the flow. The optimal time step of the simulation was estimated by perfor- ming several computational tests. It was found to be of the order of 10−7 s. Using this time step, itwas possible to obtain a sequence of solutions for about 10ms flow time only. Analysing the obtained results, it appeared that velocity fluctuations within the gap are relatively small. Their initial amplitude at the gap entrance was below 1% and increased to about 8% at the gap exit. The DNS solution confirmed the experimental findings for the quasi laminar flow within the gap and the early transition at the gap exit. Figure 6 shows velocity fluctuations extracted from the DNS simulations for three selected points behind the processing element. It can be found that theflow in thevicinity of theupperwall is characterized bynearly regular high amplitude turbulentfluctuationswith aperiodof about0.2ms.They represent small scale vortices created in the turbulent boundary layer separating from the gap exit. The reversal flow present closer to the channel symmetry plane exhibits low frequency variation of its amplitude. Here, the flow is mainly concentrated in a large vortex with slowly varying in time and space location. Such behaviour can be well visualized by constructing a shortmovie from the sequence of DNS solutions. Fig. 6. Direct numerical simulation. Temporal variation of the streamwise velocity componentmonitored for the central plane at three selected locations behind the gap (x=1mm, 3mm and 8mm); (a) y=−0.2mm; (b) y=−2.0mm Theaveraged flowcharacteristics close to the topwallwerewell reproduced using about 200 time steps, i.e. by averaging turbulent fluctuations over 2ms time interval. This averaging time was selected as a compromise necessary to limit the total computational time. Themain features of the averaged velocity fields obtained from the DNS flow simulation were in a good agreement with the experimental findings. Itwas confirmed that upstream the gap the average flow velocity was about 5m/s, and in the gap a strong acceleration of the flow occurs with velocity peak values reaching about 23m/s. PIV analysis of turbulent flow in a micro-channel 499 Numerical simulation allows for evaluation of all three components of the velocity fluctuations, however for comparison of the obtained results with the experimental data, a two-dimensional projection of the flow field was used. Hence, themean square value of the velocity fluctuations V ′ x and V ′ z obtained from the direct numerical simulation were used to evaluate tkexz representing the turbulent kinetic energy in the similar way as in the experimental part. Figure 7 displays tkexz obtained for a relatively short averaging time. It indicates that the flow turbulence, i.e. the averaged velocity fluctuations reach their maximum about 6mm behind the gap. It is in relatively good agreement with the experimental data obtained from the PIV measurements (comp. Fig.5 – P5). Fig. 7. Direct numerical simulation. Contours of the mean square value of the velocity fluctuations tke xz calculated from 200 time steps. Visible asymmetry of the contours is due to short averaging time (2ms) Direct numerical simulations utilize large amount of computer resources and computational time. The analysis performedusing the averaged turbulent model applied to one-quarter of the cavity, confirmsmain features of the flow field obtained by direct numerical simulation. The average velocity vector field attains themaximumvalue of about 20m/s in the gap between the processing element and the top wall, and a well developed recirculation region is present behind the gap. However, assuming in the numerical model turbulent flow at the channel inlet implies its fast development within the gap. Hence, the averaged turbulent model predicts highly developed turbulent flow already within the gap, contrary to the experimental observations. A comparison of the numerical and experimental results for the avera- ged velocity profiles was done to validate the numerical simulations. Figure 8 combines the profiles of the x-component of the velocity obtained from the numerical simulations using the turbulent model (gray line), the direct nu- merical simulation (solid line), and micro-PIV measurements (circles). Each experimental point represents the space and time averaged velocity value eva- luated as it was described above. The DNS numerical data are single point averages over 2.5ms flow time. The numerical and experimental profiles are 500 S. Błoński, T.A. Kowalewski compared at three locations behind the processing element: 1mm (Fig.8 – P3), 3mm (Fig.8 – P4), and 8mm (Fig.8 – P5). Qualitative agreement of Fig. 8. Comparison of the numerical and experimental longitudinal velocity profiles evaluated along the channel depth from the top wall to the channel axis. Three locations interrogated experimentally are displayed: (P3) – 1mm, (P4) – 3mm, and (P5) – 8mm behind the processing element. Black circles indicate averagedPIV measured data; gray line indicates CFD k-e data from the k-ε turbulent numerical model; solid line indicates CFD-DNS averaged data from direct numerical simulations the numerical and experimental results in the vicinity of the processing ele- ment is fairly good. The presence of strong velocity gradients close to the top wall is probably responsible for diminishing of the resolution of the PIV evaluation, which in turn decreases in all three cases the peak magnitude of the measured velocity profile. The negative velocities close to the flow axis indicate the reversal flow, confirmed in both numerical profiles. However, a more detailed analysis of these results shows significant differences – the local flow rates calculated from the experimental data along profiles P3-P4 is much lower than the numerical prediction. For the profiles located 8mmbehind the processing element (Fig.8 – P5), the difference between numerical and expe- rimental results becomes even more significant, although the location of the velocity maximum is still relatively well preserved. There are several factors PIV analysis of turbulent flow in a micro-channel 501 responsible for the obvious discrepancies between the experimental and nume- rical profiles. First of all, it must be noted that the analysed turbulent flow is strongly three-dimensional and the flow measurements performed at few po- ints cannot properly represent the average flow field across the channel. It is particularly true for the fluctuating recirculation flow present behind the pro- cessing element. Hence, fluxes calculated using only one-component velocity profile are not representative, and as its is seen in Fig.8 it leads to serious deviations. The DNS simulation indicates strong temporal flow field fluctu- ations with a broad range of characteristic frequencies (Fig.6). Therefore, it was difficult to obtain a representative averaging time for theDNS simulation, and the time interval of 2.5ms used to obtain the averages is most probably insufficient, as it is already indicated by slight asymmetry of the fluctuation contours in Fig.7. Evaluation of velocity field fluctuations may permit one to identify flow regions responsible for the droplets break-up in the emulsifier. In the turbulent flow, the droplets break-up rate is related to the magnitude of the kinetic energy dissipation. Distribution of the turbulent kinetic energy, obtained from the direct numerical simulations shows that the maximum value of turbulent kinetic energy is reached in the region just behind the processing element. It is where the intense mixing of the high velocity fluid ejected from the gap with a low velocity of the recirculating flow behind the processing element takes place. It is in agreement with the experimental findings indicating the highest velocity fluctuation about 8mm behind the gap. However, the k-e turbulent numericalmodel points out that themaximumof turbulent energy dissipation rate is at the gap entry, where both direct numerical simulation and micro- PIV experiment indicate a very little amplitude of velocity fluctuations. The conclusionwould be that the assumption of turbulence in numericalmodelling may give misleading results, especially in regions too small to allow for its development. 4. Conclusions Velocity measurements (micro-PIV) indicated almost a uniform velocity flow field in the short micro-channel formed by the gap region of the emulsifier. It means that the turbulence is still not fully developed. A strong recirculation zone with the reversal of flow is found behind the processing element. The turbulent fluctuations of the velocity field and break-up of the flow symmetry observed in this region indicate that probably transition from the laminar to turbulent flow regime occurs there. Numerical modelling confirmed main de- tails of the velocity flow field measured by the micro-PIV method. The DNS 502 S. Błoński, T.A. Kowalewski and turbulent flow models were successfully applied producing a similar ave- raged flow structure. Although, the turbulent flow model overestimates early turbulent transition within the gap. The DNS model confirms the observed delayed transition to turbulence within short contraction of the emulsifier. However, computational overload of the DNS simulations (about 1month for 2ms flow time) prohibited deeper quantitative analysis of the flow pattern. It appears necessary to extend computational time and largely expand the area covered by the experimental interrogation to explain in details the observed discrepancies of measured and computed velocity profiles. Acknowledgments This investigation was conducted in the framework of EMMAproject supported byAustrianMinistryofScienceandEducation, contractno.:GZ45.534/1-VI/6a/2003 CONEX. The micro-PIV expertise was developed in the framework of EUThematic Network PIVNET2. References 1. Brenn G., 2005, Emulsions with nanoparticles for new materials, Scientific Report of EMMA Project, Graz University of Technology, TUGraz 2. Błoński S. Korczyk P, Kowalewski T.A., 2007, Analysis of turbulence in a micro-channel emulsifier, Int. J. Thermal Scs, doi:10.1016/ j.ijthermalsci. 2007.01.028 3. Fluent 6.3, 2006,Users Guide, Fluent Inc., Lebanon NH, USA. 4. 4. Kowalewski T.A., Błoński S. Korczyk P., 2006, Turbulent flow in a microchannel,Proc. of ASME ICNMM2006, CD-ROMpaper 96090, Limerick, Ireland 5. Li H., Olsen G., 2006, MicroPIV measurements of turbulent flow in square microchannels with hydraulic diameters from 200µm to 640µm, Int. J. Heat Fluid Flow, 27, 123-134 6. Meinhart C.D., Wereley S.T., Gray M.H.B., 2000, Volume illumina- tion for two-dimensional particle imageVelocimetry,Measurement Science and Technology, 11, 809-814 7. Santiago J.G., Wereley S.T., Meinhart C.D., Beebe D.J., Adrian R.J., 1998, A micro particle image Velocimetry system, Exp. Fluids, 25, 316- 319 8. Shinohara K., Sugii Y., Aota A., Hibara A., Tokeshi M., Kitamori T.,OkamotoK., 2004.High-speedmicro-PIVmeasurements of transient flow in microfluidic devices,Measurement Science and Technology, 15, 1965-1970 PIV analysis of turbulent flow in a micro-channel 503 9. SteinerH., TeppnerR., BrennG.,VankovaN.,TcholakovaS., Den- kov N., 2006, Numerical simulation and experimental study of emulsification in a narrow-gap homogenizer,Chemical Engineering Science, 61, 5841-5855 Analiza przepływu turbulentnego w mikro-kanale przy wykorzystaniu cyfrowej anemometrii obrazowej Streszczenie Wpracy przedstawiono analizę turbulentnego przepływuwody przezmikro-kanał emulsyfikatora. Wyniki uzyskane eksperymentalną techniką Cyfrowej Anemometrii Obrazowej (PIV–Particle Image Velocimetry) zaadoptowanej do pomiarówwmikro- skali (micro-PIV) porównano zwynikami symulacji numerycznych przeprowadzonych zarówno poprzez bezpośrednie rozwiązanie równań Naviera-Stokesa (DNS – Direct Numerical Simulation), jak i wykorzystując hipotezę Reynoldsa uśredniającą fluk- tuacje prędkości i ciśnienia (model k-ε). Część eksperymentalna pracy przeprowa- dzona została na unikalnym stanowisku pomiarowymbazującym namikroskopii epi- fluorescencyjnej i szybkim obrazowaniu analizowanego przepływu. Charakterystyki przepływu wyznaczone eksperymentalnie i numerycznie symulacją DNS wykazały, że turbulizacja przepływu pojawia się dopiero w końcowym przekrojumikro-kanału, osiągającmaksimumw kanale wylotowym emulsyfikatora. Manuscript received February 9, 2007; accepted for print March 19, 2007