volumetric flow studies in a 4-stroke water...

16th Int Symp on Applications of Laser Techniques to Fluid Mechanics Lisbon, Portugal, 09-12 July, 2012

- 1 -

Volumetric Flow Studies in a 4-stroke Water-Analogue IC-Engine Using High-

Speed Scanning-PIV

David Hess1,*, Stefan Tag2, Christoph Brücker1

1: Institute for mechanics and fluid dynamics, Technical University Freiberg, Germany 2: Service-Center Ostthüringen, TÜV Thüringen e.V., Germany

* Correspondent author: [email protected] Abstract Experiments are carried out in a water analogue 4-stroke internal combustion engine (ICE) with four fully variable intake and outlet valves. Due to the incompressibility of the fluid only the intake stroke is determined. The main test rig, provided by Volkswagen AG, allows the investigation of different opening cycles, equivalent to the different load cases of a real engine in a short period of time. A constant rotational flow structure with its axis perpendicular to the cylinder axis, the so called tumble flow, is of great importance for a reproducible ignition. High-resolution scanning is used to measure a full 3D-field at a high spatiotemporal resolution. A high-speed Nd:Ylf-LASER in combination with a rotating polygon mirror generates 100 parallel and partially overlapping light sheets per volume at 10 kHz. Each light-sheet has a thickness of about 2 mm and an overlap of 75% with the previous and the successive one. The light sheet images are recorded by a Phantom 12.1 running at 10,000 fps with a maximum resolution of 960 x 600 pixels. This leads to an illuminated volume of about 50 mm in scanning direction. This volume can be reconstructed using a stacking technique, which is then analysed by 3D Least Squares Matching (LSM) to retrieve the velocities as well as the velocity gradient matrix. The measurements show a shift of the tumble centre as well as weakened tumble strength during the piston stroke. Hence, the fully developed flow consists of different vortices interacting with the characteristic tumble flow, therefor it is important to investigate fully time resolved 3D data. Different flow characteristics, which lead to a tumble break up, are studied and described in this paper. 1. Introduction The flow within an internal combustion engine (ICE) does not only influence the cycle to cycle fluctuations, it also affects efficacy and produced pollutants after the ignition. Former studies based on multiplanar 2-D measurements of Khalighi B (1991), Voisine M, et al. (2010) and Holographic-PIV studies performed by Konrath R, et al. (2002) show the flow field only in one region or only in a thin range in one duty cycle. Therefore, it is difficult to capture the evolution of flow and possible reasons for cycle-to-cycle variations. Most gasoline ICE´s with direct injection (DI) generate a rotational flow structure with its axis perpendicular to the cylinder axis, called tumble. It is a dominant flow induced by the downward moving piston during the intake stroke. Air is drawn through the inlet valves and channelled to the opposed side of the cylinder wall where the flow gets “parallel” to the piston. However, cycle to cycle variations are influencing the tumble breakup during the following compression stroke. Goryntsev D, et al. (2009) showed that the maximum cycle to cycle variation is highest at the end of the intake stroke. Hasse C, et al. (2010) also showed high grades of cycle to cycle variations with a comparison of a numerical approach with measurements.


- 2 -

In order to keep the tumble stable until the ignition, it is necessary to gain more detailed information about the evolution of the tumble, it´s dissipation and the influence of the fluctuations on the tumble. Therefore, it is important to investigate the full three dimensional field with an adequate temporal resolution. High-resolution scanning PIV with a large number of planes (≥100) is used, for a volumetric voxel-based reconstruction. This resolves the flow during the intake cycle with a depth of 50 mm within the cylinder. 2. Experiment Overview ICE characteristics The experiments are carried out in an original four cylinder head with a displacement of 1.6 liters. The cylinder head has four valves in total, two at a time for the inlet and exhaust valves. The ICE has a bore of 76.5 mm and an original stroke of 86 mm. To better fit the image section of the high speed cameras, the stroke was decreased to 75 mm. In total, three different image sections are used: a total perspective using standard high-speed PIV is used to get a quick overview of the flow within the cylinder. Cross sections normal Z- and X-direction in the mid plane of the cylinder are investigated. Furthermore, those measurements are used to validate the scaling and the 3D measurements. To ensure detailed information of the flow, two overlapping image sections are used for the 3D measurements. A draft of cylinder with the image sections is shown in Figure 1.

Figure 1: ICE overview

The piston and the valves are driven by synchronised single linear motors, to easily change the valve timings. This makes the test setup variable for different investigations and cam shaft modifications become redundant. The linear motors can follow different path-time diagrams as consecutive points, loaded on each servo drive where the maximum valve stroke is 10.5 mm. The linear motor (Mannesmann MDD065) has a maximum theoretical speed of 0.66 m/s. Due to the weight of the piston and the drag caused by the sealing, the maximum speed is reduced to about 0.4 m/s. Furthermore, high loads reduce the position accuracy by influencing the closed-loop control. Thus, the maximum speed is limited to 0.25 m/s, respectively minimal intake stroke duration of three seconds, if the maximum velocity remains constant for the whole intake process.


- 3 -

Normalised to the crank shaft angle in degree (CAD), the velocity is 30°CAD/s for one complete rotation. Thereby, the absolute position accuracy is about 0.075 mm at maximum piston velocity. A general example of the valve timings can be seen in Figure 2. Here, the solid line shows the piston displacement from its top dead centre (TDC) position at 0° CAD to its bottom dead centre (BDC) at 180° CAD. While the downward movement, the piston aspirates water-glycerine mixture from a reservoir. The inlet valves (indicated in dashed lines) start to open about 60° CAD earlier. After 180° CAD the exhaust valves open and with the beginning piston upward movement the scavenging of the cylinder starts pushing the liquid back into the reservoir. A valve-opening overlap cannot be used, in order to not influence the measurements.

Figure 2: Example of the valve timing 75mm piston displacement, no phase shifting

As mentioned earlier, the tumble motion of the air is of great importance in ICE´s. Thus, the cylinder head is equipped with an additional tumble flap that provokes a tumble motion, if load case demands it. This flap blocks the lower part of the split inlet port in order increase the flow velocity in the upper part of the port. Hence, more air streams over the upper part of the inlet-valve plate directed directly to the opposite side of the cylinder, which increases the tumble motion. Test setup A Pyrex-silicon construction forms the cylinder bore in an octagonal shape, ensuring good optical access. Furthermore, the silicon eases a refractive index (n) matching. This is necessary, due to the curvature of the cylinder, as the light path to the cameras is deviated at the interface between the fluid and the solid body of the cylinder wall. The influence of the different refractive index is compensated with a water-glycerine mixture (42% glycerine), which is used as working fluid. Spherical Vestosint particles with a diameter of about 100 µm are seeded as tracers into the fluid circuit. A high-speed Laser system is used with a single camera and a scanning light sheet (up to 200 scans per volume), to scan the volume of the cylinder (see Figure 3). The scanning illumination is generated by a polygon mirror (see Hess D, et al. 2010). A stepper motor cranks the polygon to maintain a constant rotational speed. The laser is synchronized to start illuminating a new volume at the beginning of each facet. The laser beam is formed into a light sheet before it is reflected at the polygon. The laser (Nd:YLF) laser has an energy output of 30 mJ per pulse at pulse rate of 1 kHz. Each light sheet of one volume scan is reflected from the rotating facet in a different angle. Thus, the full measurement volume is illuminated. The single light sheets are recorded by a Phantom V12.1 high speed camera, at a frame rate of 10,000 frames per second with a resolution of


- 4 -

960x 600 pixels. Thereby, complete volume representations are gathered with 80 Hz repetition rate. The schematic overview of the optical setup can be found in Figure 3 and the complete test stand is shown in Figure 4. Here, the adjustment of the scanning illumination, the camera and the actual test rig is clarified.

Figure 3: Optical setup

Figure 4: Complete overview of the experimental setup, with illumination / acquisition and ICE test-rig

Data processing Each light sheet has a thickness of about 2.4 mm and a shift of 0.51 mm. The single images of one volume scan can be stacked together to span a voxel volume, if several assumptions are fulfilled. Firstly the light sheets need to be regular, which means that the intensity distribution over the height and width has to be constant. Furthermore, the shift of each consecutive light sheet needs to be the same over the whole scanning depth. Secondly, each light sheet needs to overlap the subsequent light sheet for at least 33 %. Thirdly the scanning velocity needs to be much higher, than the flow velocity within the measurement volume (see Hess D, et al. 2011 & Thomas L, et al. 2011).To calibrate the position of each light sheet within the measurement volume and to check the correct overlap, a target is inserted in the cylinder and illuminated by scanning light sheets. Figure 5 shows an arrangement of all 100 used light sheets in one image. In order to gain a better visibility, every plane is shifted for 10 pixels in horizontal direction and every 2nd plane is shifted vertically for 100 pixel, due to the high overlap. However, if the scanned planes are less than the image resolution, every image needs to be stacked multiple times to maintain the aspect ratio of the reconstructed voxel volume compared to the measured volume. In case of the ICE measurements, the result is a voxel volume with a typical size of 960 x 600 x 600 voxel, representing a volume of about 80x 50 x 50 mm³ for the zoomed image sections. Due to the overlap of the light sheets and multiple stacked images, one must admit that the particles become elongated in scanning direction and form ellipsoids. Since the light sheets are overlapping, the ellipsoidal particles have a Gaussian intensity distribution over their length. This intensity distribution of each ellipsoidal particle can be used to apply a Gaussian regression in order to replace the ellipsoids with spherical blobs (Brücker Ch, et al. 2012). Going in scanning direction line by line through the voxel volume, a Gaussian fit determines the centre of each ellipsoidal particle fragment in one line, as shown in Figure 6. A weighting or smoothing function can be applied to the voxel volume to locate the absolute particle centre position. Afterwards, a Gaussian spot with a size of 3 x 3 x 3 voxel can be inserted at this absolute centre positions.


- 5 -

Figure 5: Arrangement of the light sheets. The images are overlaid in a staggered arrangement to illustrate the overlapping planes.

Figure 6: Gaussian regression to replace ellipsoidal

particles with Gaussian spots. Left row shows the ellip-

soidal particle shape over 4 light sheets and its intensity

distribution with Imax for the representing the centre of the

particle. Right row shows the reduced particle with its

centre at the determined Gaussian fitting of Imax . The reprocessed voxel volumes of each consecutive time steps are analysed using a 3D-LSM, to evaluate the flow directions, velocities and the velocity gradient matrix. The LSM is an iterative process where the grey values of voxel cuboids from time step are shifted, rotated sheared and stretched until they best fit the voxel representation of time step two (Maas HG, et al. 1994). One advantage of this method is that the velocity gradient matrix is a direct solution of the algorithms and it must not be computed with all its problems with numerical derivations. Typical cuboid sizes of 65³ voxel are used for the LSM. With a median seeding density of 0.1 particles per pixel the overlapping percentage of the cuboid can be chosen to a higher level (75 % and more) without a risking oversampling occurrence. Thus, up to 140,000 vectors can be gained from one time step, at low computational costs. This corresponds to 0.7 vectors / mm³ for the zoomed image sections and ensures an adequate spatial resolution to investigate the in cylinder flow. 3. Results Due to the fact that the water-glycerine mixture has a higher viscosity than water, it is difficult to scale the model results to a real ICE setup. In order to check the influence of this effect on the in-cylinder flow, the water-glycerine flow is compared to test case, where pure water is used as fluid. Thus, the resulting Reynolds number is eight times higher and more comparable to an air flow within the ICE. The influence of the different refractive indexes can be corrected by using a cylindrical lens between the cylinder and the camera. Retracing can also recalculate the distorted image in one plane, if calibrated. After correcting the refracted image, it is possible the centre planes of the two different measurements can be compared without any spatial deviations at border area of the cylinder wall.


- 6 -

Figure 7: Velocity and directional comparison of a pure water (left) and a water-glycerine mixture (right). Colours

indicate the absolute velocity normalised by the piston velocity In Figure 7 this comparison is visualized at 80° CAD, with the pure water flow on the left side and the water-glycerine flow on the right hand side of the image. The colour indicates the absolute velocity scaled with the piston speed. In both cases the flow exhibits a typical tumble motion at the same height and a comparable velocity range. It is apparent that the tumble motion is more distinct in the water application. Furthermore, the tumble centre is shifted to the cylinder centre, as indicated in Figure 8.

Figure 8: Absolute flow velocities profiles at tumble centre height at 80° CAD normalised with the piston speed. Black

line median of 5 cycles with pure water, grey dashed lines cycle to cycle velocities of water-glycerine mixture, black

dashed line median of 10 water-glycerine mixture cycles


- 7 -

It shows normalised absolute flow velocities in the cross-sectional through of the tumble. Averaged and non-averaged velocity profiles from pure water and water-glycerine mixtures are plotted. The water flow cycles are indicated by the black line, an average of ten cycles of the water-glycerine mixture is indicated by a dashed black line. Grey dashed lines indicate the ten different cycles of the water-glycerine mixture. Rapidly decreasing flow velocities and afterwards increasing velocities can be understood as vortex detection criteria. The tumble centre is located at positions where the criterion has strongest gradients and high amplitudes. For the water-glycerine mixture, the tumble centre is located at a bore diameter of 49 mm, determined out of the ten single cycles. In contrast to the behaviour of the tumble centre, the tumble width fluctuates at the side of the inlet valves in a range of 10 mm. Also a lot of inflection points indicate small secondary vortices. When added pure water only, the tumble position is at 42 mm, here this position also does not vary a lot within the single measurements. When comparing the two curves, most parts of the water-velocity curve is located within the fluctuation range of the water-glycerine mixture. This indicates a similar and comparable behaviour of the different liquids, even if the Reynolds number does not match. Furthermore, the two different liquids show the same behaviour during the tumble formation. At higher crank angels two phenomena occur. On the one hand, the tumble degrades and loses kinetic energy. Due to a decreasing piston speed and the attended lower flow rate intake trough the valves. On the other hand the tumble centre moves from the right hand side of the cylinder further to the left hand, as shown in Figure 9. Here, velocity profiles are shown for three different crank shaft angles and each curve is a phase average of five cycles gained from the water measurements.

Figure 9: Velocity profiles at different crank shaft angles indicating the tumble center shift over the crank angle The diagram shows the tumble centre where velocities drop close to zero, together with a shift of the tumble centre position at higher crank shaft angels. Furthermore, the maximum velocity drops at higher crank shaft angles, resulting in more vortices, as fluctuating velocity profiles of 120° CAD


- 8 -

and BDC indicate. Murali Krishna B, and Mallikarjuna J M (2009) also described this effect for in-cylinder air flows. The tumble centre shift can also be observed with water-glycerine as working fluid, whereas this effect is even more distinct. Furthermore, it is obvious, that several small vortices start to disturb the tumble motion at BDC, as the left plot in Figure 10 shows. On the right hand side the same image cross-section represents the flow field with a closed tumble flap. As expected, the main tumble motion has already collapsed into different vortices. In both cases vortices are formed within the first 80° after TDC. Compared to the flow conditions from Figure 8, the small vortices on the left hand side of the tumble centre become aspirated into the tumble at higher crank shaft angles. For the flow regime with opened tumble flap this effect is more distinctive, due to a lower swirling strength of the tumble. Furthermore, the flow rate over the left side of the valve plate is higher leading to more vortices. Figure 10 shows a comparison of the two states. On the left side with a closed tumble flap, leading to higher swirling strength of the tumble. On the right hand side with an open tumble flap. Here, the tumble is already disrupted by secondary flow vortices.

Figure 10: Flow field comparison of closed (left) and opened tumble flap (right) at BDC.

Since, the cylindrical shape has an influence on the flow. The three dimensional flow field needs to be investigated, in order to understand the vortex formation and vortex interaction. Figure 11 gives an overview of the 3D-flow formation within the cylinder at different crank shaft angels of 40° CAD (a), 70° CAD (b) and 150° CAD (c). The slices represent the velocity magnitude normalised with the mean piston velocity. The flow rate over the inlet-valve plates is illustrated by the black iso-surfaces from the velocity magnitude. The combined flow rate over the valve plates separates into two main regimes. In the X-plane between the inlet-valves a high shear region occurs. The shear is induced by the two inlet-valves, within this region a lot of small vortices are generated, due to flow rate fluctuation between the two valve-ports. Some of the small vortices can be seen in the slices of figure 11 a). Those secondary flow vortices disturb the tumble formation. Hence, the beginning of a large tumble formation starts at about 60° CAD. In Figure 11 b) the stream tracers indicate the tumble as the primary flow condition. However, the tumble expansion in z-direction is not jet fully developed. The tumble formation continues until 100° CAD, when the tumble dilates most of the scanning direction.


- 9 -

Figure 11: Overview of the 3D-flow formation within the ICE at different crank angels, with closed tumble flap: a)

40° CAD (image section 1,) b) 70 CAD (image section 1), c) 150° CAD (image section 2). The transparent black iso-

surfaces shows the flow rate channelled over the inlet-valve plate, indicated by a constant absolute flow velocity

normalized with the piston speed. The blue iso-surface illustrates the Lambda-2 criteria, visualizing the tumble

formation. Additionally the tumble flow is highlighted with the red stream tracers


- 10 -

A small fraction of the fluid streaming into the cylinder is not channelled over the inlet-valve plates to the opposed side of the cylinder. It directly streams downwards parallel to the cylinder wall, at the inlet valve side. This part counteracts the tumble flow, forming several secondary flow vortices. Those vortices are transported by the tumble front. At higher crank-shaft angels those vortices are gathered into the tumble centre. The beginning of this process can be seen on the Z-plane slice in Figure 11 b) and also in Figure 10 on the left side; here, the process is already finished. Some of the gathered vortices are counter rotating to the tumble flow and reduce the overall kinetic energy of the tumble. Thereby the flow velocity of the tumble is reduced in the upper part of the tumble. At higher crank shaft angels, the tumble gets deformed to a u-shaped profile with its ends facing the cylinder head, as Figure 11 c) demonstrates. The blue iso-surface of constant vorticity represents the tumble centre. The figure also depicts a lateral tilt of the vortex core. This tilt is a result of the gathered vortices slowing down the upper part of the tumble. This deformation occurs independent of the tumble flap position. But it is easier to investigate with provoked tumble at closed tumble flap, because of limited secondary flow vortices disturbing the primary tumble flow. The u-profile deformation may have different reasons. First of all, the tumbles ends start interacting with the cylinder walls and slow down due to wall friction. These slow ends are displaced upwards by high velocity fluid channelled from the valves to the tumble. Secondly, the tumble generates a force by induction resulting to the left cylinder wall. The used piston has a plane shape. Hence it would be useful to investigate different piston shapes in future work. This would help discovering the main reason for the tumble centre line deformation. In the subsequent compression and combustion stroke the tumble deformations needs to be avoided. Hence the irregular tumble shape can be more deformed during the compression. The risk of a complete break-up is high, due to high shear rates resulting from the compression. A complex flow structure with a lot of small vortices will be the result and leads to an insufficient combustion and consequently to a decayed efficiency and more pollutant exhausts. 4. Conclusion This work presents a high-speed scanning technique with a high quantity of light sheets (up to 200), applied on a 4-Valve internal combustion engine. A water-glycerine mixture is used as fluid to avoid distortion from refractive index disparities between the fluid and the cylinder wall. The illuminated light sheets are gathered by a single high-speed camera with a frame rate of 10 kHz. The images can be stacked together to span a voxel-volume if the thin light sheets overlap at least 33 %. The investigated volume has a size of 80 x 50 x 50 mm³. Elongated particle shapes in scanning direction, resulting from the overlapping light sheets, are reduced to spheres by a Gaussian fit of the intensity distribution. The voxel volumes are analysed by a 3D Least-Squares-Matching algorithm, in order to evaluate the flow velocity and the velocity gradient matrix. Results show a shift of the tumble centre to the left cylinder wall during the intake stroke. Furthermore, the tumble centre-line is deformed to a u-profile, at higher crank angels. Additionally, the deformed tumble centre line is tilted. The deformation and the tilt result from vortex interaction and wall friction. They insufficient characteristics for the subsequent compression phase. A tumble break up can occur, with adverse effects on the combustion. References Brücker Ch, Hess D, Kitzhofer J (2012): Single-View Volumetric PIV via High-Resolution Scan-ning, isotropic voxel restructuringand 3D Least-Squares Matching (3D-LSM). Reviewed and to be published in: Measurements Science and Technology


- 11 -

Goryntsev D, Klein M, Sadiki A, Janica J (2009): Grobstruktursimulation für einen DI-Ottomotor Charakterisierung des Strömungs- und Mischugnsfeldes. MTZ 03I2009 Jahrgang 70, p. 252-257 Hasse C, Sohm V, Durst B (2010): Numerical investigation of cyclic variations in gasoline engines using a hybrid URANS/LES modelling approach. Computers & Fluids 39 (2010) 25–48 Hess D, Brücker Ch, (2010): 3-D Scanning PIV of the Flow within a Two-Stroke Water Analogue Combustion Engine, Proc. 15th Int. Symp. on Applications of Laser Techniques to Fluid Mechanics Hess D, Brücker Ch, Kitzhofer J, Nonn T (2011): Single-View Volumetric PIV using High-resolution Scanning and Least Squares Matching. 9th International Symposium on Particle Image Velocimetry – PIV´11 Khalighi B, (1991): Study of the intake tumble motion by flow visualization and particle tracking velocimetry. Experiments in Fluids 10, 230 236 Konrath R, Schröder W, Limberg W (2002): Holographic particle image velocimetry applied to the flow within the cylinder of a four-valve internal combustion engine. Experiments in Fluids 33 781-793 Maas HG, Stefanidis A, and Grün A, (1994). From pixels to voxels -tracking volume elements in sequences of 3-d digital images. In Int.Arc.Photogr.Rem.Sens.303/2. Murali Krishna B, Mallikarjuna M J, (2009): Tumble Flow Analysis in an Unfired Engine Using Particle Image Velocimetry Thomas L, Tremblais B, Braud P, David L (2011) Comparison of Algebraic tomography PIV and scanning PIV for fluid flow. Proc. FVR 2011, Forum on recent developments in Volume Reconstruction techniques applied to 3D fluid and solid mechanics Voisine M, Thomas L, Borée J, Rey P (2010): Spatio-temporal structure and cycle to cycle to cycle variations of an in-cylinder tumbling flow. Experiments in Fluids 2011 50:1393-1407 Westfeld P, Maas H.G, Pust O, Kitzhofer J, Brücker Ch (2010): 3D least square matching for volumetric velocity data processing; Proc. 15th Int Symp on Applications of Laser Techniques to Fluid Mechanics

volumetric flow studies in a 4-stroke water...

Documents