The gas dynamics of the flow past an exhaust valve has been investigated using Schlieren photography. An experimental setup was designed and constructed, which allowed optical access to the valve head and seat region as well as to the exhaust port. The setup was constructed so that the shock structures of a steady flow, with a static valve, could be compared to the structures found in experiments with a more realistic dynamically discharging cylinder, with a moving valve. The steady flow experiments were carried out at a valve lift to a port diameter ratio of 0.155 with cylinder pressures up to 325 kPa. The dynamic valve experiments were performed with an initial cylinder pressure of 300 kPa and at an equivalent engine speed of 1350 rpm. The steady flow experiments belonged to one of the two flow regimes, depending on the cylinder pressure: regime I, a wall-bounded supersonic jet (for low cylinder pressures) or regime II, a fully expanded supersonic nozzle-flow (for high cylinder pressures). By comparing the images from the dynamic valve experiment to those of the steady flow experiments, it was shown that the flow in the dynamic experiments exhibits more similarities with regime I. However, large differences in the shock structures between the steady flow in regime I and the dynamic valve flow remain. This indicates that experiments using a steady flow and a fixed valve lift do not encompass the essential physics found in real engine flows and should be avoided.

This work deals with the gas dynamics of the exhaust valve and port during the initial part of the exhaust stroke of an engine cylinder. This section starts with a brief introduction to engine cylinder flows with definitions of some pertinent parts and quantities, whereafter some earlier work on shocks and so called shock trains in conduits is reviewed. Finally some previous results, that motivated this study, will be summarised.

The most common way (in the automotive industry) to increase the fuel efficiency of a four stroke engine is through “downsizing” (reduction of the cylinder volume), either by decreasing the volume of the cylinders and/or by decreasing the number of cylinders. One way to keep the performance of a downsized engine is to recover some of the energy in the exhaust gases with a turbocharger turbine and use the compressor instead to increase the amount of air in the cylinder.

Figure 1 shows a simplified model of a diesel-engine cylinder. During the intake stroke, when the intake valve is open and the piston is moving down toward the bottom dead centre (BDC), air enters the cylinder through the intake port (coloured blue). The shape of the intake port and valve determines the large-scale flow structures in the cylinder when the piston reaches BDC. These structures are modified during the compression stroke but still have a large effect on the combustion process and have thus been extensively studied. After the combustion and expansion (power stroke), the exhaust valve opens and allows the residue gases to leave the cylinder through the exhaust port (coloured red).

FIG. 1.

Illustrative image of a simplified diesel engine cylinder.

FIG. 1.

Illustrative image of a simplified diesel engine cylinder.

Close modal

In comparison to the flow during the intake stroke, the flow through the exhaust valve does not have a major effect on the combustion and has thus been studied to a lesser degree. The shape of the exhaust valve and port as well as the valve strategy (timing and speed of opening) does, however, affect how efficiently the cylinder can be emptied. This has an effect on the pumping losses of the engine [the lower the pressure in the cylinder as the piston moves up toward the top dead centre (TDC), the better] as well as on the flow properties at the turbine inlet (which affects the efficiency of the turbocharger).

To design an efficient engine system, it is important to minimise aerodynamic losses during the exhaust stroke and to understand its flow physics. Typically, the exhaust valve opens at approximately 50 crank angle degrees (CAD) before the piston reaches BDC and closes at approximately 15 CAD after the piston reaches TDC.1 This allows for a small overlap where the inlet and exhaust valves are both open. The exhaust stroke itself can be divided into two phases: the blowdown and the displacement phases. The blowdown phase refers to the initial part of the exhaust stroke, where the pressure ratio across the valve is high and the flow is driven by the high cylinder pressure. During this phase, the piston is always close to BDC, so the cylinder volume is relatively constant. The displacement phase refers to the later part of the stroke, where the displacement of the piston drives the expulsion of gas from the cylinder. The overlap when both the intake and exhaust valves are open helps to flush out the last of the exhaust gases from the cylinder.

During at least part of the blowdown phase the pressure ratio across the valve is higher than the critical limit for supersonic flow. As a result, the flow reaches sonic speed at the smallest area (close to the valve seat, see Fig. 1) and then expands supersonically into the port. In the port, the supersonic flow will decelerate through a complex process involving one or several shock waves.

Most studies on exhaust valve flows are concerned with the discharge coefficient (CD), which is the ratio between the actual and the ideal (i.e., isentropic) mass flows through the valve. CD thus assumes a value between 0 and 1 and is used in engine simulation software as a scaling factor of the open valve area, to compensate for the aerodynamic losses; see, e.g., the work by Woods and Khan.2 There are, however, some studies where attempts to study the flow structure in the exhaust port have been made. For instance, Tanaka3 performed flow visualisation on a 2D geometry to study the flow structure present in the exhaust flow. Also large eddy simulations have been performed4 which showed that the structures reported by Tanaka also existed in the 3D geometry. However, both of these studies were for low cylinder pressures and the flow never reached the supersonic condition. There does not seem to be any previous work in the open literature dealing with the shock structures in the exhaust port.

Ideally, when decelerating an internal inviscid flow from supersonic to subsonic velocity, the pressure recovery is performed across a single normal shock wave. However, in a conduit, there is a build up of a boundary layer along the walls, which can interact with the shock wave. The interaction is especially prominent for turbulent boundary layers as they give rise to strong mixing of high and low momentum flow.

For Mach numbers below 1.2, the influence of the shock wave is limited. However, as the Mach number increases, the interaction between the shock wave and the boundary layer becomes stronger. For Mach numbers between 1.2 and 1.3, a small separation bubble is formed and the shock wave starts to bend (convex in the flow direction) and the distance between the shock wave and the wall increases. At Mach numbers between 1.3 and 1.5, the region of separation increases and the boundary layer shows less tendency to reattach. As a result of the increased region of separation, the shock wave starts to form bifurcated ends with a normal section in the centre of the channel (often called a λ-shock). For Mach numbers above 1.5, a large scale separation of the boundary layer occurs after the first shock wave, forming an “aerodynamic nozzle,” which allows the flow to once again accelerate to supersonic velocities outside of the boundary layers. In order to accommodate the high pressure downstream, another shock wave is formed. This may occur several times forming a series of shocks (a shock train) of continuously decreasing shock strength.

After the last shock wave, the pressure continues to rise over a region of mixed supersonic and subsonic flows (called the mixing region) and the flow is decelerated to subsonic velocities everywhere without any shock waves. The pressure increases due to the mixing, reaching a maximum at some point, where the pressure decrease due to wall friction (the so called Fanno flow5) becomes greater than the pressure rise due to the mixing. The local maximum of the pressure marks the end of the mixing region. The region containing both the shock train and mixing region is referred to as a pseudo-shock. A good overview of the physics related to the pseudo-shock phenomenon and a review of previous research, up to the late 1990s, are given by Matsuo et al.6 

The shape of the shock train is determined mainly by the Mach number and by the ratio of boundary-layer displacement thickness (δ1) to conduit half width (h) (δ1/h, known as the confinement parameter). For large enough Mach numbers, the bifurcation, of the first shock wave in the shock train, reaches the centre of the channel (the shock does no longer have any normal section) and this kind of shock trains is called “oblique shock trains” or an “X-type shock train.” When the first shock wave of the shock train has a normal section, the shock train is called a “normal shock train” or a “λ-type shock train.” A study by Carroll and Dutton7 concerning the influence of the Mach number and the confinement parameter showed that if the Mach number was kept constant (M = 1.6 in this case) and δ1/h was increased (δ1/h = [0.08, 0.14, 0.27, 0.32, 0.40, 0.49]), the number of shock waves and the spacing between two consecutive shock waves increased as well. Similarly, the number of shock waves and the spacing between them increase if the Mach number is increased.

The position of the shock train has been shown to oscillate around a time-mean position. There are three competing theories for the cause of these oscillations: turbulent fluctuations in the incoming flow,8 pressure fluctuations in the low subsonic regions far downstream of the shock train which propagates upstream,9 or unsteadiness of the separation at the foot of the first shock wave.10 This is still an open topic of research.

The flow past the exhaust valve has traditionally been assumed to be quasi-steady so that the pressure-loss coefficients can be obtained from steady measurements at different valve lifts (). The reasoning behind this assumption is that the flow speed (which is of the order of the speed of sound) is much higher than the maximum valve speed (a few meters a second) so that the effects of the valve motion can be neglected.

In our previous work,11,12 this view was challenged by carrying out and comparing measurements of the mass flow rate past the valve both during the steady state and dynamical conditions. The measurements were carried out on a fixed volume (2 l) cylinder, where the valve could be operated dynamically or locked at a fixed valve lift. On the one hand, when the rig was operated dynamically, the valve was closed and the cylinder pressurised to a target pressure. The valve was then opened according to a pre-programmed lift curve. On the other hand, in the steady state experiments, the valve was locked at a fixed valve lift and the cylinder was connected to a compressor system which was run at a range of mass flow rates (covering the flow rates found in the dynamic experiments). That study showed a large difference when comparing the ratio of the cylinder pressure and the port pressure (often called the back pressure which is measured in the pipe downstream the valve head) under static and dynamic valve operations. This is illustrated in Fig. 2 where the ratio of cylinder stagnation pressure (p0,cyl) and back pressure (pback) is plotted as a function of p0,cyl. The figure shows the pressure ratio at a valve-lift-to-port-diameter (d) ratio /d = 0.143 for both static valve conditions and at four different equivalent engine speeds. In those experiments, only the lift sequence is modelled, i.e., from a fully closed to a fully open valve. For a typical engine, the exhaust valve opens before BDC and closes after TDC, and here it is assumed that the valve is open corresponding to a change in the crank angle of approximately 250°. For the 1350 rpm case, the time from closed to fully open corresponds to approximately 15 ms.

FIG. 2.

Cylinder-to-back-pressure ratio as a function of cylinder stagnation pressure for static (full drawn lines) and dynamic flow cases (symbols) obtained at /d = 0.143. Four different dynamic cases are shown, corresponding to different equivalent engine speeds. Data are from Ref. 12.

FIG. 2.

Cylinder-to-back-pressure ratio as a function of cylinder stagnation pressure for static (full drawn lines) and dynamic flow cases (symbols) obtained at /d = 0.143. Four different dynamic cases are shown, corresponding to different equivalent engine speeds. Data are from Ref. 12.

Close modal

Steady flow data for two more /d are also shown in Fig. 2. It can be seen that the flow past the static valve (solid lines with no markers) belongs to one of two different flow regimes. Regime I is characterised by a close to linear increase of the cylinder to back pressure ratio with increasing cylinder pressure. At some cylinder pressure, the flow enters regime II in which the pressure ratio is high and independent of further increase in the cylinder pressure. The large and constant values of p0,cyl/pback in regime II indicate that the flow is supersonic in this regime. Since the minimum flow area (AT) increases with /d, the exact value of p0,cyl/pback will decrease with increasing /d as it is determined by the ratio of the local flow area (A) and the minimum flow area (p0,cyl/pbackA/AT). On the other hand, the near-linear increase in the pressure indicates that the flow is, at least partly, subsonic in regime I. It can also be seen that the flow past the dynamically operated valve does not enter regime II.

Surface-flow visualisation on the valve and port was reported in Ref. 12. The results indicate that conical shock waves are formed in the exhaust port. The images in Fig. 3 show the valve for flow in regime I (left) and regime II (right). In both flow regimes, a deposit of paint can be seen close to the valve head (marked with 1 in Fig. 3). In regime I, this is the only shock related deposit of paint. In regime II, two additional deposits can be seen on the valve stem. The first one is located just upstream of a slight increase in the valve-stem diameter (marked with 2), and the second one is located further downstream on the valve stem (marked with 3). Another deposit of paint could be seen on the port wall in between the two deposits on the valve, marked with a red ellipse in Fig. 4.

FIG. 3.

Photos of surface-flow visualisation performed on the full (360°) geometry. The image on the left side shows the valve for flow within regime I, and the image on the right shows the surface flow for regime II.

FIG. 3.

Photos of surface-flow visualisation performed on the full (360°) geometry. The image on the left side shows the valve for flow within regime I, and the image on the right shows the surface flow for regime II.

Close modal
FIG. 4.

Photo of surface-flow visualisation performed on the outlet pipe wall for the full (360°) geometry in regime II.

FIG. 4.

Photo of surface-flow visualisation performed on the outlet pipe wall for the full (360°) geometry in regime II.

Close modal

Based on the results of the surface-flow visualisation, a schematic of the shocks in the port region could be conceived, see Fig. 5. It was hypothesised that the deposit of paint on the valve head was created by a weak normal (or close to normal) shock wave. When the flow is in regime I, that shock wave terminates the supersonic flow. However, in regime II, the upstream pressure (cylinder pressure) is large enough for the flow to remain supersonic even after the weak shock wave. The slight increase in the valve-stem diameter is hypothesised to cause an oblique shock wave which is reflected on the port wall back to the valve stem. From the pressure-ratio measurements shown in Fig. 2, it appears that the flow during the dynamic valve operations belongs to regime I.

FIG. 5.

Sketch of the conceived shock geometry in accordance with the observed surface flow visualisation. The valve, seat, and port inner wall geometries are accurately represented in the sketch.

FIG. 5.

Sketch of the conceived shock geometry in accordance with the observed surface flow visualisation. The valve, seat, and port inner wall geometries are accurately represented in the sketch.

Close modal

In Sec. II, the new experimental setup is described which allows flow visualisation of the flow in the port using Schlieren photography. Section III shows results both for pressure-ratio measurements across the valve as well as visualisation results both for the static and dynamic cases as well as an analysis of the images and a comparison between the static and the dynamic cases. Finally the results are summarised in Sec. IV.

In order to further study and also verify the conceived shock patterns in the exhaust port downstream the valve, suggested in our previous studies,11,12 a new setup was designed and constructed, which allows for Schlieren visualisation of the shock structures. The setup was designed to allow for both static and dynamic experiments, similar to those performed in our previous work.

To gain optical access to the convoluted region between the valve head and the seat, the new setup was designed with a 30° conical section instead of the full axisymmetric geometry. An alternative would have been to use a two-dimensional slice; however, then it would not be possible to replicate the geometric flow area distribution of the previous axisymmetric experiments during the opening sequence. This feature is important since the area distribution is not fully determined by the 2D profile as the change of radius (and thus some addition to the area distribution) is not captured. Of course, both a 2D setup and a 30° setup have additional side walls that do not exist in the axisymmetric case, where boundary layers develop and therefore also affect the effective flow area. However, this effect is deemed to be small, at least close to the valve seat. Some minor changes as compared to the axisymmetric experiments were made to the valve seat, including the removal of a tap for measurement of seat pressure and the removal of a small indent directly downstream of the seat. For ease of manufacturing, the geometries were scaled up 1.5 times compared to the original. Most of the setup was made in aluminum whereas the valve part was constructed through three-dimensional printing in plastic.

The setup used in experiments with a dynamic valve is shown in Fig. 6, and a schematic of the setup is shown in Fig. 7. The volume representing the engine cylinder had an equivalent radius of 90 mm, corresponding to a cylinder with a 180 mm bore (B) and 265 mm stroke (S). From now on this volume will be referred to as the cylinder. The cylinder could be sealed with the moving exhaust valve in the closed position. The cylinder had taps for both pressure and temperature measurements. The valve (with an equivalent diameter of D = 61.5 mm) was actuated with an electromagnetic linear motor (LinMot P10-70 × 320U). The valve connected to the linear motor through a valve-actuation rod (VAR), which exited through a sealed passage in the cylinder bottom, and a dual-joint coupling (to compensate for any misalignment between the motor and the valve path). The position of the valve was measured directly on the valve using a Novotechnik TE1-0025 linear transducer, mounted on the side wall, with a quoted repeatability of 2 μm and a linearity of ±0.2% of full scale output (FSO = 25 mm).

FIG. 6.

3D model (top) and photo (bottom) of the setup used in the experiment with a dynamic valve.

FIG. 6.

3D model (top) and photo (bottom) of the setup used in the experiment with a dynamic valve.

Close modal
FIG. 7.

Schematic (relative scales are accurate) of the setup used in the experiment with a dynamic valve.

FIG. 7.

Schematic (relative scales are accurate) of the setup used in the experiment with a dynamic valve.

Close modal

The seat, port, and outlet pipe were made from one solid piece of aluminum, which gave large surfaces to use for sealing and clamping of the side walls. The port had two taps for pressure measurements, and only measurements from the upstream one of these are used in this analysis. The port had an equivalent diameter of 52.5 mm (d) and had a straight outlet with a length of approximately 4d, discharging to the atmosphere.

The side walls were attached by using screws to the port, cylinder wall, and cylinder bottom and sealed using a thin silicone film. The valve side surfaces were sealed by the clamping of the side walls, which would pinch the valve, valve stem, and valve-actuation rod. To aid the sealing, the valve sides had large surfaces, parallel to the side wall bottom surface, which were covered with bearing grease. It also had a device with linear ball bearings which would force the valve down and increase the surface mating with the side walls.

Pressures were measured using Kistler 4045A5 piezoresistive transducers, with a quoted accuracy of ≤0.1% FSO (FSO = 500 kPa). The temperature was recorded using a T-type thermocouple and a Fluke 51-II thermocouple reader, with an accuracy of ±0.3 K. The pressure and valve-position signals were recorded using a 16-bit National Instrument PCI-6250 A/D-converter, at a sampling frequency of 25 kHz in the dynamic experiment and 16 kHz in the static experiments.

When running a dynamic valve experiment, the valve is initially closed and the cylinder is pressurised to 300 kPa using a pressure inlet at the bottom of the cylinder. Once the cylinder conditions were stable, a trigger signal was sent to the linear motor (which opens the valve using a preprogrammed lift curve equivalent to an engine speed of 1350 rpm) and to the camera of the Schlieren optics.

In the static experiments (see Fig. 8), the cylinder bottom was replaced with an area transition section so that the cylinder could be connected to the CICERO laboratory compressor system at the Fluid Physics Laboratory at KTH.13 Also, the linear motor was replaced with a locking mechanism holding the valve at a fixed valve lift.

FIG. 8.

3D model of the setup used in experiments with a static valve.

FIG. 8.

3D model of the setup used in experiments with a static valve.

Close modal

Schlieren photography utilises the property that the refractive index (n) of a gas is directly proportional to the density (ρ),

n=1+ρK,
(1)

where K is the Gladstone-Dale constant. More specifically, if light travels through a region with a density gradient normal to the ray path, the wave fronts will bend toward the region of larger density and thus bend the ray paths in that direction.14 

In a Schlieren system, a point light source is placed in the focal point of a lens, giving parallel light going to the test section. Any normal density gradient causes the ray path to deviate from the undisturbed ray path and not go through the focal point of a second lens on the other side of the test section. If an opaque screen (Schlieren knife) is placed very close to the focal point of the second lens, the light that has passed through a density gradient will hit the knife and not pass through to the camera, thus generating a dark region in the image. Depending on the angle of the knife (if a straight knife is used), different density gradients will be highlighted better.

In the static experiments, several different knife angles and knife types were used in order to obtain an optimal contrast from the images. The different knife angles, for which Schlieren images will be shown, are given in Fig. 9. In the dynamic experiments, an iris was instead used as a Schlieren knife. Photos were taken using a Photron FASTCAM APX-RS high-speed camera at a frequency of 6000 frames per second.

FIG. 9.

Sketch showing the different Schlieren knife angles used in the static valve experiments (denominated knife angle 1–3). The gray dot symbolises the focal point of the Schlieren system.

FIG. 9.

Sketch showing the different Schlieren knife angles used in the static valve experiments (denominated knife angle 1–3). The gray dot symbolises the focal point of the Schlieren system.

Close modal

The result section is divided into four different subsections. In Sec. III A, the pressure ratio measurements are shown for the new setup and compared with the previous axisymmetric results. In Secs. III B and III C, the Schlieren images are discussed for the static and dynamic cases, respectively, in order to show the distinct features of each case. Finally in Sec. III D, a thorough discussion about the differences of the static and dynamic shock and flow structures is attempted.

Figure 10 shows the same axisymmetric pressure-ratio data as in Fig. 2. In the 30° setup, static measurements were made for /d = 0.155 and these data are shown as red circle marks. The comparison for the static valves for the two different setups shows a qualitative and quantitative agreement for low cylinder pressures (regime I) with an almost linear increase in the pressure ratio. At a cylinder pressure slightly above 270 kPa, the data from the 30° setup indicate a shift into regime II. As can be seen from the axisymmetric case, the behavior is rather sensitive to the /d ratio and an exact correspondence should not be expected. In addition, the quantitative value of p0,cyl/pback will depend not only on the valve lift but also on the total pressure loss past the valve. This means that due to the increased friction from the addition of the side walls, the 30° geometry is expected to show a lower value for p0,cyl/pback compared to the full geometry.

FIG. 10.

Cylinder stagnation pressure to back-pressure ratio as a function of cylinder stagnation pressure for the full 360° geometry (black markers and coloured lines) and the 30° geometry (red and cyan markers).

FIG. 10.

Cylinder stagnation pressure to back-pressure ratio as a function of cylinder stagnation pressure for the full 360° geometry (black markers and coloured lines) and the 30° geometry (red and cyan markers).

Close modal

In the figure, the blue squares mark the cases for which Schlieren images are shown in this article.

The black symbols connected with black lines mark the pressure ratio across the dynamic valve for the axisymmetric case at the position /d = 0.143. Correspondingly, the cyan asterisk marks the pressure ratio over the dynamic valve at /d = 0.154 for the 30° setup.

A Schlieren image of the static valve flow in regime I is shown in Fig. 11, for a cylinder pressure of 214 kPa. The image shows traces of oblique shock waves just downstream of the choking region (located between the flat section of the valve and the seat). The extended dark spots in this region are related to large accelerations of the flow and thus large density gradients. The flow can be seen to separate from the port wall (top of the image) downstream of the seat, where a sharp corner is located. Close to the valve (bottom of the image), Mach waves can be seen, indicating that the flow is supersonic in this region but subsonic close to the port wall (further discussed in Sec. III D).

FIG. 11.

Schlieren images of the exhaust port for flow in regime I, and flow is from right to left. p0,cyl = 214 kPa.

FIG. 11.

Schlieren images of the exhaust port for flow in regime I, and flow is from right to left. p0,cyl = 214 kPa.

Close modal

As the cylinder pressure is increased (moving into regime II), the flow is able to expand past the corner without separating and the flow in the port becomes fully supersonic. The increase in the Mach number due to the larger flow area is the cause of the increase in p0,cyl/pback for regime II. In Fig. 12, Schlieren images for three different cylinder pressures (all in regime II) are shown (cylinder pressure decreasing from top to bottom picture). Different knife orientations are used to enhance various aspects of the shock-wave patterns observed in these images.

FIG. 12.

Schlieren images of the exhaust port for flow in regime II using different angles of the Schlieren knife and cylinder pressures (p0,cyl = 325, 304, 295 kPa, see Fig. 10).

FIG. 12.

Schlieren images of the exhaust port for flow in regime II using different angles of the Schlieren knife and cylinder pressures (p0,cyl = 325, 304, 295 kPa, see Fig. 10).

Close modal

The shock-wave pattern close to the throat, in regime II, is similar to that of regime I. However, here, an oblique shock wave can be seen that originates from the port wall, just downstream of the seat. This shock wave is reflected on the valve surface. Further downstream, a normal shock train can be seen and the starting position of the shock train is affected by the upstream pressure (forcing it downstream) and the confinement ratio (i.e., the effective aerodynamic cross-stream area). When the reflected shock wave meets the shock train, it is terminated. From these images, it can be seen that the oblique shock wave, originating from the port wall and reflecting on the valve surface, is more or less independent of the position of the shock train (except for the fact that it ends as it encounters the shock train). It is assumed that the interaction of this oblique shock wave with the boundary layers of the valve stem and outlet pipe is what was causing the deposits of paint in previous experiments on the full 360° geometry, see Figs. 3 and 4. The fact that the oblique shock wave is reflected on the pipe wall and can be seen again on the valve stem in the 360° experiments is most likely due to the fact that the shock train is located further downstream as the confinement effect of the side walls was not present in those experiments. The reflection on the valve stem occurs slightly further upstream in the 30° experiments compared to the 360° case. This is probably caused by a slight difference in the valve lift and/or that the side walls may have an effect on the angle of the shock wave since the effect of the boundary layers of the walls is to decrease effective aerodynamic area and thereby decrease the bulk Mach number.

The interaction between the boundary layer and the normal shock train can be readily seen in the Schlieren images using knife angle 2 (horizontal knife, positioned just below the focal point). This is because the light that bends as it enters the “low velocity” boundary layer, close to the port wall, will be blocked by the knife, and the separation of that boundary layer will be highlighted (which is larger than the boundary layer close to the valve and thus have a larger impact on the aerodynamic area).

In the case of a dynamic valve opening, the flow undergoes several different states during the process from the closed valve to the fully open position.

The dynamic valve process can be divided into three different types of flow states as the valve opens:

  • an overexpanded jet with two free boundaries,

  • a wall-bounded overexpanded jet, and

  • a fully expanded nozzle flow.

These three states, together with sketches of the shocks and expansions, are shown in Fig. 13 (Multimedia view). For small /d, the flow belongs to state A [Fig. 13, upper figures (Multimedia view)]. As /d increases, the diameter of the jet increases as well. At some point, the diameter of the jet has increased enough for the jet boundary to change from a free boundary to the surface of the valve; the flow thus enters state B [Fig. 13, middle figures (Multimedia view)]. When in state B, the jet is bound and forced to turn by the valve surface. The other jet boundary separates from the port wall and remains a free boundary. The angle of the boundary can be seen to change with /d, tending to decrease the angle between the jet and the port wall with increasing /d. At large enough /d, the flow no longer separates from the port wall and the flow can now expand in the full port geometry, state C [Fig. 13, bottom figures (Multimedia view)]. The flow then terminates its supersonic state with a normal pseudo-shock. The location of the shock train is at first more or less constant in the streamwise direction. As the cylinder pressure falls, the shock train moves upstream before disappearing as the flow becomes subsonic.

FIG. 13.

Different flow states of the dynamic discharge process. The sketches show shocks and expansion waves. Gray regions in the bottom right sketch are subsonic. A video of the dynamic valve experiment can be viewed in the online version of this paper Multimedia view: https://doi.org/10.1063/1.5084174.1

FIG. 13.

Different flow states of the dynamic discharge process. The sketches show shocks and expansion waves. Gray regions in the bottom right sketch are subsonic. A video of the dynamic valve experiment can be viewed in the online version of this paper Multimedia view: https://doi.org/10.1063/1.5084174.1

Close modal

In our previous work, it was hypothesised that the flow during the dynamic valve operation belongs to regime I for medium /d (/d ∈ [0.1, 0.2]). This hypothesis may be confirmed or disproved by comparing the Schlieren images of the two static regimes to that of the dynamic valve experiment.

In Fig. 14, Schlieren images for the two static regimes (at /d = 0.155, the two bottom images) and the dynamic experiment (at /d = 0.154, the top image) are shown. The approximate boundaries of the jets for the flow in regime I and the dynamic experiment are highlighted with red dashed lines and some Mach waves by magenta lines. Figure 14 shows that the dynamic valve experiment resembles the flow in regime I (thus, the flow in regime I belongs to flow state B). Some differences can, however, be seen. The angles of the oblique shock waves, close to the valve head and seat, can be seen to differ. Furthermore, the initial angle of the free boundary of the jet (at the point of separation from the port wall) is smaller in the static flow (i.e., the boundary of the jet is closer to the port wall). This is likely caused by the shift in the position of where the oblique shock wave is reflected against the free boundary, which will act to turn the boundary to become more parallel with the major axis of the setup.

FIG. 14.

Schlieren images of the dynamic valve experiment at /d = 0.154 (top, p0,cyl = 277 kPa) and the static valve experiments at /d = 0.155 in regime I (middle, p0,cyl = 214 kPa) and regime II (bottom, p0,cyl = 325 kPa). The red dashed line, in the top two images, indicates the estimated boundary of supersonic velocities in the formed jet. The magenta lines highlight some Mach waves, with corresponding Mach angles.

FIG. 14.

Schlieren images of the dynamic valve experiment at /d = 0.154 (top, p0,cyl = 277 kPa) and the static valve experiments at /d = 0.155 in regime I (middle, p0,cyl = 214 kPa) and regime II (bottom, p0,cyl = 325 kPa). The red dashed line, in the top two images, indicates the estimated boundary of supersonic velocities in the formed jet. The magenta lines highlight some Mach waves, with corresponding Mach angles.

Close modal

The expansion of the jet also differs between the static regime I and the dynamic valve case. In the dynamic experiment, the shock wave is reflected against the free boundary further downstream and the jet has a larger cross-sectional area. This leads to a larger expansion of the flow in the dynamic case and thus smaller Mach angles, as shown by the relation between the Mach angle (μ) and the Mach number (M)

M=1sinμ.
(2)

Using the angles found in the images (μ = 43°, 52°, and 32°), it is possible to make a quantitative comparison of the Mach numbers for the different flows. It can thus be shown that the highest Mach number is found in regime II (M = 1.89) where the flow can expand into the entire exhaust port (and thus belong to flow state C). The second highest Mach number is found in the dynamic valve case (M = 1.47), which, as discussed earlier, has a more expanded jet compared to the static regime I flow. The static regime I flow thus has the lowest Mach number (M = 1.27).

To further highlight these features, the unsteadiness of the Mach and shock waves can be used. By plotting the difference between two images separated by a small time (Δt = 1/6000 s in this case), the boundaries of the jets and the Mach waves are enhanced, see Fig. 15. For the dynamic case, the valve moves slightly which can readily be seen by the bright mark on the lower flat side of the valve (to the right in the figure). Note that it is not truly the Mach wave that is shown but rather the displacement of the Mach wave over a small time period. For the static valve experiments, this has little effect, but in the dynamic valve case, the valve actually moves slightly and thus it is better to determine the Mach angles using the original images rather than these.

FIG. 15.

Images, generated by subtracting two subsequent images from each other (same cases as shown in Fig. 14), of the dynamic valve experiment (top) and the static valve experiments in regime I (middle) and regime II (bottom). The cyan lines mark the approximate location the valve and port wall.

FIG. 15.

Images, generated by subtracting two subsequent images from each other (same cases as shown in Fig. 14), of the dynamic valve experiment (top) and the static valve experiments in regime I (middle) and regime II (bottom). The cyan lines mark the approximate location the valve and port wall.

Close modal

Shock visualisation experiments have been performed on an idealised exhaust valve flow of an internal combustion engine. The experiments focused on two main topics: (i) Exploring the difference between measurements using a steady flow and a static valve and measurements made on a discharging cylinder with a dynamic valve (i.e., a moving valve); (ii) explaining the difference between different flow regimes previously found for static valve operations, where the exhaust port experienced a sudden large drop in pressure for high cylinder pressures.

The static valve experiments were performed using a high pressure flow bench (pressures up to 500 kPa and flow rates up to 0.5 kg/s) whereas the dynamic valve experiments were made by discharging a pressurised volume, equivalent to an engine cylinder for which that valve was originally used. Shock visualisation images were captured using high-speed (6000 frames/sec) Schlieren photography.

The images show that the difference between the two static flow regimes is caused by the flow ability to expand past the convoluted surface of the valve seat and port if enough upstream pressure is available. As the flow is able to follow the geometry without large scale flow separation, it reaches high Mach numbers and thus low pressures in the port.

The shock structures found in the dynamic valve experiments carry a closer resemblance to those seen in regime I compared to the fully expanded regime II. There are, however, still relatively large differences in both shock structures close to the minimum flow area and in the shape of the supersonic jet formed in the port.

From this, it can be concluded that static valve experiments, using a flow bench, fail to capture the key components of the gas dynamics (shock structures and Mach numbers) in the exhaust valve flow, showing that some fundamental physics are missing in the static valve experiments. This means that basing conclusion on design parameters based on data taken from such experiments might lead to, at best, suboptimal decisions or, at worst, catastrophic failure.

This work has been supported by CCGEx, the Competence Centre for Gas Exchange at KTH. We also like to thank Dr. Michael Lieverts with help on the Schlieren setup and Jonas Vikström for building the new setup. The 3D printing of the valve part of the setup was expertly made by the company FELIMA.

To give an overview of the flow during the dynamic valve experiment, a selection of representative Schlieren images of the dynamic valve process are given in Fig. 16.

FIG. 16.

Schlieren images of the dynamic valve experiments at six different /d, given in the images.

FIG. 16.

Schlieren images of the dynamic valve experiments at six different /d, given in the images.

Close modal
1.
J. B.
Heywood
,
Internal Combustion Engine Fundamentals
(
McGraw-Hill
,
New York
,
1988
).
2.
W.
Woods
and
S.
Khan
, “
An experimental study of flow through poppet valves
,”
Proc. Inst. Mech. Eng. Conf. Proc.
180
,
32
41
(
1965
).
3.
K.
Tanaka
, “
Air flow through exhaust valve of conical seat
,” Rep. Aeronautical Res. Inst., Tokyo Imperial University,
1931
, Vol. 6, pp.
1
24
.
4.
Y.
Wang
,
B.
Semlitsch
,
M.
Mihaescu
, and
L.
Fuchs
, “
Flow induced energy losses in the exhaust port of an internal combustion engine
,”
J. Fluids Eng.
137
,
011105
(
2015
).
5.
J. D.
Anderson
,
Modern Compressible Flow: With Historical Perspective
, 3rd ed. (
McGraw-Hill
,
New York
,
2004
).
6.
K.
Matsuo
,
Y.
Miyazato
, and
H.-D.
Kim
, “
Shock train and pseudo-shock phenomena in internal gas flows
,”
Prog. Aerosp. Sci.
35
,
33
100
(
1999
).
7.
B. F.
Carroll
and
J. C.
Dutton
, “
Characteristics of multiple shock wave/turbulent boundary-layer interactions in rectangular ducts
,”
J. Propul. Power
6
,
186
193
(
1990
).
8.
T.
Ikui
,
K.
Matsuo
,
M.
Nagai
, and
M.
Honjo
, “
Oscillation phenomena of pseudo-shock waves
,”
Bull. JSME
17
,
1278
1285
(
1974
).
9.
R.
Yamane
,
E.
Konda
,
Y.
Tomita
, and
N.
Sakae
, “
Vibration of pseudo-shock in straight duct: 1st report, fluctuation of static pressure
,”
Bull. JSME
27
,
1385
1392
(
1984
).
10.
H.
Sugiyama
,
H.
Takeda
,
J.
Zhang
,
K.
Okuda
, and
H.
Yamagishi
, “
Locations and oscillation phenomena of pseudo-shock waves in a straight rectangular duct
,”
JSME Int. J. Ser. 2, Fluids Eng. Heat Transfer, Power, Combust., Thermophys. Prop.
31
,
9
15
(
1988
).
11.
P. M.
Winroth
,
C. L.
Ford
, and
P. H.
Alfredsson
, “
On discharge from poppet valves: Effects of pressure and system dynamics
,”
Exp. Fluids
59
,
24
(
2018
).
12.
M.
Winroth
, “
On gas dynamics of exhaust valves
,” Licentiate Thesis, TRITA-MEK 2017:03,
Royal Institute of Technology, KTH
,
Sweden
,
2017
.
13.
F.
Laurantzon
,
N.
Tillmark
,
R.
Örlü
, and
P. H.
Alfredsson
, “
A flow facility for the characterization of pulsatile flows
,”
Flow Meas. Instrum.
26
,
10
17
(
2012
).
14.
W.
Merzkirch
,
Flow Visualization
(
Academic Press
,
1974
).