Experimental Thermal and Fluid Science 34 (2010) 217–226 Contents lists available at ScienceDirect Experimental Thermal and Fluid Science journal homepage: www. elsevier. com/locate/etfs Development of a novel passive top–down uni? ow scavenged two-stroke GDI engine G. Ciccarelli *, Steve Reynolds, Phillip Oliver Mechanical and Materials Engineering, Queen’s University, Kingston, Ontario, K7P 2M4 Canada a r t i c l e i n f o a b s t r a c t The design and performance characteristics of a novel top–down uni? ow scavenged gasoline direct-injection two-stroke engine are presented.
The novelty of the engine lies in the cylinder head that contains multiple check valves that control scavenging air? ow into the cylinder from a supercharged air plenum. When the cylinder pressure drops below the intake plenum pressure during the expansion stroke, air ? ows into the cylinder through the check valves. During compression the cylinder pressure increases to a level above the intake plenum pressure and the check valves close preventing back-? ow into the intake plenum. The engine head design provides asymmetrical intake valve timing without the use of poppet valves and the associated valve-train.
In combination with an external Roots-type supercharger that supplies the plenum and exhaust ports at the bottom of the cylinder wall, the novel head provides top–down uni? ow air scavenging. Motoring tests indicated that the check valves seal and the peak pressure is governed by the compression ratio. The only drawback observed is that valve closing is delayed as the engine speed increases. In order to investigate the valve dynamics, additional tests were performed in an optically-accessible cold ? ow test rig that enabled the direct measurement of valve opening and closing time under various conditions. O 2009 Elsevier Inc.
All rights reserved. Article history: Received 11 September 2008 Received in revised form 26 October 2009 Accepted 27 October 2009 Keywords: Two-stroke engine Uni? ow scavenging 1. Introduction The advantages of the two-stroke engine are its high power to weight ratio and its simple, low-manufacturing cost design. The higher power density is the result of torque delivered to the crankshaft each crank revolution, as opposed to every other crank revolution as achieved in a four-stroke engine. The challenge for a two-stroke engine lies in the ef? cient removal of exhaust gases from the cylinder while simultaneously introducing a fresh charge.
The conventional two-stroke engine achieves this by using the crankcase as a fresh charge reservoir and the piston as a pump. During the compression stroke the crankcase pressure drops and a fresh premixed air–fuel charge is inducted into the crankcase through a reed valve. This air–fuel mixture is compressed in the crankcase during the expansion stroke. As the piston approaches bottom dead center (BDC) the exhaust ports located in the cylinder wall are uncovered, starting the blowdown process. Shortly thereafter the transfer port, also located in the cylinder wall, connecting the cylinder and the crankcase volumes is uncovered.
The air–fuel mixture is forced into the cylinder displacing the remaining combustion products out the exhaust port. This purging of combustion * Corresponding author. Tel. : +1 613 533 2586; fax: +1 613 533 6489. E-mail address: [email protected] queensu. ca (G. Ciccarelli). 0894-1777/$ – see front matter O 2009 Elsevier Inc. All rights reserved. doi:10. 1016/j. exptherm? usci. 2009. 10. 029 products is referred to as scavenging and represents the key process in a two-stroke engine. Insuf? cient scavenging results in excessive exhaust gas residual concentration that leads to slow, inef? ient burning and mis? res, the effects and remedies of which have been widely studied . Conventional two-stroke designs also suffer from lower fuel ef? ciency due to signi? cant amounts of air–fuel mixture being lost out the exhaust pipe. This short-circuiting effect also contributes to high emissions due to unburned hydrocarbons ? owing out the exhaust along with the combustion products. In the last decade signi? cant improvements in fuel ef? ciency and emissions have been achieved by scavenging the combustion products using air and the introduction of gasoline direct-injection (GDI).
In the late 1990s the Orbital Engine Company of Australia designed such a two-stroke engine for automotive application. It was a crankcase air-scavenged GDI engine that used an air-assisted fuel injection system . In more recent years there has been a resurgence of interest in two-stroke engines operating with GDI and externally-compressed air scavenging. For example, the ELEVATE engine used cylinder ports to control scavenging with a charge-trapping valve at the exhaust port for asymmetrical timing . The use of an external compressor allows for wet-sump lubrication, similar to that used in four-stroke automotive engines.
In 2004, Moriyoshi, Arai, and Morikawa reported on the development of a uni? ow scavenged GDI engine that operates with a cam driven single 218 G. Ciccarelli et al. / Experimental Thermal and Fluid Science 34 (2010) 217–226 Table 1 Engine speci? cations. Bore A stroke (mm) Displacement (cc) Clearance volume (cc) Trapped displacement (cc) Effective compression ratio Exhaust ports open/close Fuel injection pressure 86. 0 A 70. 0 407 32 305 10. 5 120° ATDC/120° BTDC 7. 5 MPa (75 bar) poppet valve in the cylinder head to control scavenge air ? w from a Lysholm type external compressor . Exhaust ports were located at the bottom of the cylinder as with conventional twostroke engines. Performance data including power output, fuel consumption, and hydrocarbon emissions obtained from the engine compared favorably with data from a conventional loop scavenged carbureted two-stroke and four-stroke engines of similar size. The drawback of this design is that it uses a poppet valve and associated valve-train which moves away from the simplicity and low cost of a conventional ported two-stroke engine.
This study explores the feasibility of using multiple check valves built into the cylinder head, in place of a camshaft driven poppet valve, in a top–down uni? ow-scavenged engine. Freshair-only scavenging in combination with high pressure GDI and a dedicated oil pan style lubrication system ensures the possibility of a clean burning engine that could rival four-stroke engine emissions. The primary objective of this study is to demonstrate the feasibility of the check valve concept and the basic operation of the engine. 2. Engine design and operation A single-cylinder 1975 Yamaha 400 cc two-stroke engine was modi? d for top–down uni? ow scavenging and GDI operation. The engine crankcase and internal components including the crankshaft, main bearing, connecting rod and piston are from the original engine. The original engine was crankcase scavenged requiring a fuel–oil blend to lubricate the main bearings and cylinder walls. This type of lubrication was not available for the current engine so a small amount of oil was manually added to the crankcase through lubricating holes before each engine test. The aluminum piston has a slightly convex crown designed for the original engine scavenging method.
The engine block and cylinder head were specially constructed to incorporate the unique features of the engine. Passages were machined into the block and head in order to circulate cooling water. As seen in Fig. 1, two diametricallyopposed sets of exhaust ports are located at the bottom of the cylinder, each set consisting of four 15 mm diameter closely spaced round ports. The engine speci? cations are listed in Table 1. The passive uni? ow scavenging feature of the engine is achieved by the unique head design. The head contains a matrix of 16 ‘‘check valves” that react to the pressure differences between the intake plenum and the cylinder.
The air? ow pathway obtained during scavenging with the piston at bottom dead center (BDC) is shown schematically in Fig. 2. Each check valve consists of a 10 mm diameter channel opening up to a 2. 6 mm long, 16 mm Fig. 2. Schematic showing the scavenging air ? ow pathway. Fig. 1. Photograph of the top of the aluminum piston within cylinder after disassembly. Shiny part indicates where the injector spray impacts the piston as a high momentum jet. diameter cavity. The cavity contains a 1. 15 mm thin, 12. 74 mm square steel platelet that weighs 1. 23 g. A photograph of the bottom side of the head is provided in Fig. a. The end of the spark plug and the fuel injector in the center of the head can be made out on the photograph. A photograph of the platelet next to a 19 mm diameter coin is provided in Fig. 3b. The ring that appears on the soot coated platelet in Fig. 3b corresponds to the elevated rim of the check valve cavity top surface where the seal is made. A 2 mm thick stainless steel retaining plate fastened to the bottom of the head supports the platelets from below, see Fig. 2. The retaining plate is fastened to the head by two small bolts and can be easily removed to access the platelets.
The retaining plate has a cluster of four 5. 5 mm diameter holes for each of the 16 check valves, see Fig. 4. The schematic shows the position of the retaining plate holes relative to the check valve cavities (dotted lines). The retaining plate has a rectangular slot cut out at the center for the spark plug and fuel injector. Each platelet is free-? oating so its angular orientation relative to the corresponding four retaining plate holes is not ? xed. A roots-type supercharger driven by an independent AC motor was used to supply scavenging air. The supercharger was connected to a large 3. gallon surge tank mounted directly above the intake plenum, see Fig. 2. The purpose of the surge tank was to dampen any pressure ? uctuations produced by the engine. A ? ow turbine meter was located on the suction side of the supercharger to measure the scavenging air volumetric ? ow rate. The delivery ratio for the engine was calculated based on the measured volumetric ? ow rate and the ambient air density. Exhaust headers located on each side of the engine, were used to collect the exhaust gas exiting the two sets of four exhaust ports. A Siemens Deka fuel injector operating at a fuel pressure of 75 bar was used for all tests.
At atmospheric pressure the injector produces a hollow-cone fuel spray with a cone angle of 40°. The G. Ciccarelli et al. / Experimental Thermal and Fluid Science 34 (2010) 217–226 219 Fig. 3. Photographs of (a) the bottom of the cylinder head without the retaining plate and (b) platelet with a 19 mm diameter coin for scale. Fig. 4. Schematic of the retaining plate, hidden circles correspond to the locations of the check valve cavities. fuel was pressurized inside of a 2 l stainless steel Swagelok sample bottle with nitrogen gas supplied from a standard compressed gas cylinder equipped with a pressure regulator.
The fuel injector was mounted into the cylinder head as shown in Figs. 2 and 3. The fuel line was routed from the sample bottle into the surge tank. The fuel used for all tests contained 10% ethanol by volume. For ? red tests a standard starter was used to turn a ? ywheel that was mounted on the crankshaft just outside the crankcase. A DENSO coil-over-plug ignition unit supplied the ignition energy to the spark plug that was mounted on the cylinder head as shown in Figs. 2 and 3. The spark plug operated reliably during the testing without severe fouling problems. The spark plug gap was set to 0. 6 mm for all tests. Motoring tests were performed using an AC motor pulley drive system. A range of engine speeds was achieved using different pulley diameters on the crankshaft. The engine control unit (ECU) used for this project was a MOTEC M4. Ignition timing, injection timing, and fuel pulse width were adjusted as required for each test. The ECU communicated with a personal computer via a serial cable. The computer displayed real-time engine operating conditions and in combination with the ECU allowed for real-time tuning. The engine was connected to a super? w hydraulic dynamometer that was located inside of an engine cell. A Kistler 6117BCD17 spark plug mounted pressure transducer was used to measure cylinder pressure. The piezoelectric pressure transducer has a range of 0–200 bar and a linearity of ±0. 1% full scale output. The resulting measurement uncertainty is ±0. 2 bar. A Kistler 5010 charge ampli? er converted the capacitance signal into a voltage signal for the data acquisition. The charge ampli? er was set to 5 bar/V for motoring tests and 10 bar/V for ? red tests. The time constant was set to ‘‘long” and the sensitivity set to 16. 1 pC/bar.
The intake plenum pressure was measured by an Omega PX219-030A5V, 0–2. 1 bar absolute pressure transducer with an accuracy of ±0. 25% full scale. The resulting measurement uncertainty was ±0. 005 bar or ±0. 5 kPa. The transducer was calibrated with two known pressures: 0 bar produced by a vacuum pump, and the local atmospheric pressure as measured by a barometer. The crankshaft position was calculated using the same trigger wheel signal used by the ECU. A 58-tooth trigger wheel with a two-tooth gap in combination with a magnetic sensor produced a sinusoidal voltage signal. The trigger wheel resulted in a tooth every 6°.
Top dead center (TDC) was found using a dial indicator to measure the piston location with the head removed. A TDC index angle of 266° was used to relate the index tooth (? rst tooth after the gap) to TDC for the calculations. An estimated uncertainty for this value is ±0. 5°. For the setup used the sensor voltage drops past 0 V when a tooth is aligned with the sensor. A 16-Bit National Instruments (NI) USB-6210 data acquisition system was used to record voltage signals from the cylinder pressure transducer, intake plenum pressure transducer, crankshaft trigger wheel sensor, and injector current clamp.
The data acquisition system was controlled by lab view. Each manually triggered event sampled 40,000 data points at a sample rate of 50 kS/s. This resulted in 0. 8 s of data or 16 to 26 cycles for engine speeds between 1250 rpm and 2000 rpm. 3. Motored engine tests The novel feature of the engine is the completely passive operation of the intake valves, i. e. , they are not driven off a predictable cam pro? le as with conventional poppet valves. In all engines, the valve timing is critical to the engine performance. In this engine valve activity is dif? cult to measure directly because the valves are located in a con? ed space exposed to combustion temperatures and pressures. The cylinder pressure–time history recorded during engine motoring at 675 rpm is shown in Fig. 5. Disturbances, or ‘‘noise”, in the pressure signal are initiated at roughly 90° BTDC and 80° ATDC. The vibration sensitive piezoelectric 220 16 14 G. Ciccarelli et al. / Experimental Thermal and Fluid Science 34 (2010) 217–226 35 30 1000 rpm 2000 rpm 3000 rpm 3600 rpm Absolute Cylinder Pressure (bar) Cylinder Pressure (bar) 12 10 8 6 4 2 0 -2 -180 -150 -120 A B C 25 20 15 10 5 0 TC -90 -60 -30 0 30 60 90 120 150 180 CrankAngle(deg) Fig. 5.
Cylinder pressure obtained during engine motoring at 675 rpm with an absolute plenum pressure of 141 kPa. -5 -180 -150 -120 -90 -60 -30 0 30 60 90 120 150 180 CrankAngle(deg) Fig. 6. Absolute cylinder pressure for motored engine with a plenum pressure of 134 kPa (abs). Pressures are absolute with the following offset: 1000 rpm offset by 0 bar, 2000 rpm offset by 5 bar, 3000 rpm offset by 10 bar, 3600 rpm offset by 15 bar. pressure transducer was mounted directly on the cylinder head. It was hypothesized that the noise in the pressure signal is produced by vibrations in the cylinder head picked up by the transducer.
The noise in the pressure signal occurs at times where the valves would be expected to open and close. Vibrations are induced in the head when the platelets impact the retaining plate and the valve seat during opening and closing, respectively. Note that this method does not identify when the valves begin to move, only when they are fully open or closed. The pressure signal in Fig. 5 starts with the piston at A180°, i. e. , BDC, when the check valves and the exhaust ports are fully open and the cylinder is at atmospheric pressure.
As the piston moves towards TDC, the exhaust port area decreases and the cylinder pressure slowly increases. At 120° before top dead center (BTDC) the exhaust ports completely close and the cylinder pressure rises sharply reaching the plenum pressure. Note that the cylinder pressure rises slightly above the plenum pressure due to the ram effect. The noise starting at point A in Fig. 5 indicates when the platelets ? rst impact the valve seat. Roughly 24 crank angles (CA) elapse from when the exhaust closes at 120° BTDC to when the platelets impact the valve seat. During this time the reverse air? w imposes a drag force on the platelet accelerating it upwards toward the valve seat. The noise duration in this test is roughly 30 CA. It is dif? cult to discern at what point a complete seal is obtained but it appears that the cylinder pressure continues to rise shortly after the start of the noise. The peak pressure occurs before TDC, opposed to at TDC for the isentropic compression case, as a result of heat transfer and leakage past the piston and possibly through the valves . As the piston moves down towards BDC, the cylinder pressure drops below the plenum pressure at about 75° after top dead center (ATDC).
The pressure difference causes the platelets to accelerate towards the retaining plate. Impact with the retaining plate is made at 80° ATDC. From this point until 120° ATDC where the exhaust ports begin to open, the valves appear to be in a neutral state, ? uctuating between open and closed, as indicated by the noise. In this neutral state, between points B and C in Fig. 5, the platelets may be affected by the pressure waves within the cylinder since the average pressure difference across them is very small. At 120° ATDC the exhaust port starts to open and the cylinder pressure drops sharply to atmospheric pressure.
This prompts the platelets to impact the retaining plate at point C in Fig. 5. A set of motoring tests was performed to determine the effect of engine speed on the valve timing and peak cylinder pressure. Fig. 6 shows the cylinder pressure recorded over one crankshaft revolution for four different engine speeds. Note the pressure traces have a vertical offset in order to separate the signals for clarity. As the motoring speed increases, the valve closing noise (point A in Fig. 5) starts closer to TDC indicating that the platelets close later in the cycle.
This behavior is expected since the valve response time is governed by platelet inertia, and for a given valve closing time this translates to more crank shaft rotation for higher engine speeds. Over the motored engine speed range tested of 1000 to 3600 rpm the valve closing time increased by 17 CA, see Fig. 7 for a summary of the data. This has a signi? cant impact on the effective compression ratio and thus on the ? red engine performance. The pressure data in Fig. 6 also reveal that the valve opening noise, point B in Fig. 5, appears later in the cycle for higher engine speeds.
This is the result of increased peak cylinder pressure with engine speed. An increase in the peak cylinder pressure with increased engine speed is normally attributed to a reduction in time for heat losses. However a reduction in compression ratio with engine speed due to later valve closing has the opposite effect on the peak pressure. The observed increase in peak pressure with engine speed, see Fig. 7, indicates that the reduction in time for heat transfer dominates the reduced compression ratio effect. A larger peak cylinder pressure results in more time required for the cylinder pressure to drop below the plenum pressure.
This is not expected to have a signi? cant effect on the ? red engine because the peak pressure occurs after TDC and thus the cylinder pressure does not drop below the plenum pressure before the exhaust valve opens at 120° ATDC at which point the cylinder pressure drops precipitously during blowdown. 25 105 Peak Cylinder Pressure (bar) 20 100 95 15 90 10 85 5 Peak Cylinder Pressure Intake Closing Angle 80 0 500 1000 1500 2000 2500 3000 3500 75 4000 Engine Speed (rpm) Fig. 7. Effect of engine speed on peak cylinder pressure and valve closing time. Intake Closing Angle BTDC (deg) G. Ciccarelli et al. Experimental Thermal and Fluid Science 34 (2010) 217–226 221 Fig. 8. Illustration of check valve looking up at the retaining plate. (a) Platelet in the optimum orientation for air? ow. (b) Platelet rotated 45° resulting in a poor orientation for air? ow. It is well known that scavenging ef? ciency increases with air delivery ratio. At a given engine speed the air delivery ratio is governed by the air mass ? ow rate into the engine. Unlike for a regular poppet valve where the ? ow area is axisymmetric with respect to the valve axis, the check valve ? ow area can be different depending on the platelet orientation.
The two orientations that result in the maximum and minimum ? ow area are shown in Fig. 8a and b, respectively. During the motoring tests it was determined that the platelets always assumed the position shown in Fig. 8b. Even if the platelets were initially placed in the position shown in Fig. 8a, after a short motoring time the platelets rotated 45° to the orientation shown in Fig. 8b. The presence of carbon soot circular spots at the platelet corners, corresponding to the retaining plate holes, indicates that the unfavorable platelet orientation occurs for the ? red engine operation as well.
Note that as the platelet travels in the cavity, as the valve opens, the ? ow is not signi? cantly affected by the platelet orientation. The air mass ? ow rate measured with the supercharger running and the piston at BDC with the platelets oriented in the less favorable position shown in Fig. 8b was 33. 6 g/s, 12% less mass ? ow rate than when positioned as in Fig. 8a. The ? ow coef? cient for the engine head was measured on an in-house suction-type ? ow bench that is described in detail in . With the platelets oriented in the optimum position shown in Fig. 8a, the ? ow coef? ient corresponding to a pressure drop of 508 mm water across the head was measured to be 0. 55, which is slightly less than that for a standard poppet valve. 4. Fired engine performance tests Preliminary ? red testing was performed with the engine connected to the dynamometer with measurements of engine speed, brake torque and cylinder pressure. In order to aid engine startup, the engine coolant was initially heated to a temperature of 60 °C using a block heater. The engine was started with excess fuel, i. e. , the fuel pulse width (FPW) was set to between 6 ms and 7 ms.
For the engine charged with air at 20 °C a stoichiometric mixture is formed with a FPW of 2. 9 ms. The start of injection (SOI) was set to coincide with the closing of the exhaust ports at 120° BTDC. Once the engine started, the coolant was allowed to heat up to an engine operating temperature of roughly 95 °C. This relatively high temperature was used in order to promote fuel evaporation. After the engine was at operating temperature, a load was applied on the engine by the dynamometer to bring it to the desired engine speed. In this section engine performance data is presented for a FPW of 5. 5 ms.
Note the engine runs stably with a FPW down to 4. 5 ms with a corresponding 10% drop in the measured torque. The engine was tuned by setting the spark ignition timing for maximum brake torque. The engine operating conditions and performance data at four engine speeds is provided in Table 2. The pressure traces for 15 cycles are superimposed for each engine speed in Fig. 9. The pressure traces in Fig. 9 were ? ltered to remove the noise associated with the valve opening and closing. As would be expected there is some variability from cycle to cycle. It can be seen that the pressure peaks all occur in the range of 10°–27° ATDC.
The location of the peak pressure relative to TDC is an indication of the nature of the combustion, e. g. , improper ignition timing or slow ? ame speed can cause late peaks. From Table 2, on average the peak pressure occurs at 16° ATDC for all the tests except for the 1250 rpm case, where the peak occurs at 21° ATDC. For typical SI engines, peak cylinder pressure occurs at about 16° ATDC with optimum timing . Three out of the four tests agree with this. The late peak recorded for the 1250 rpm test indicates that the timing could be advanced more. The indicated torque and power versus engine speed are shown in Fig. 0. The maximum power is 6. 5 kW at 2000 rpm. There is a continuous drop in torque with engine speed from the maximum Table 2 Performance test conditions and results. Test 1 Head coolant temperature (°C) Intake plenum temperature (°C) Intake plenum pressure (kPa gauge) Engine speed (rpm) Ignition timing (BTDC) Air mass ? ow (g/s) Delivery ratio Intake valve closed (deg BTDC) Intake valve open (deg ATDC) Brake torque (N m)a Brake power (kW)a Indicated torque (N m) Indicated power (kW) ISFC (g/kW h) IMEP (kPa) COV in IMEP (%) Average peak pressure location (deg ATDC) a Test 2 96 44 51. 1528 20 30. 6 2. 48 89 148 31. 3 5. 0 33. 0 5. 3 861 510 2. 7 16 Test 3 98 50 52. 0 1755 20 30. 9 2. 18 89 148 32. 0 5. 9 33. 1 6. 1 859 511 2. 9 16 Test 4 98 51 53. 8 2039 25 30. 6 1. 86 87 148 28. 7 6. 1 30. 3 6. 5 940 469 5. 9 16 97 42 52. 5 1267 10 30. 5 2. 96 92 144 34. 4 4. 5 36. 9 4. 9 771 570 2. 9 21 Power required to run the blower is not taken into account. 222 G. Ciccarelli et al. / Experimental Thermal and Fluid Science 34 (2010) 217–226 a Cylinder Pressure (bar) 45 40 b Cylinder Pressure (bar) 45 40 35 30 25 20 15 10 5 0 -180 35 30 25 20 15 10 5 0 -180 -150 -120 -90 60 -30 0 30 60 90 120 150 180 -150 -120 -90 -60 -30 0 30 60 90 120 150 180 Crank Angle ATDC (deg) Crank Angle ATDC (deg) c Cylinder Pressure (bar) 45 40 35 30 25 20 15 10 5 0 -180 d Cylinder Pressure (bar) -150 -120 -90 -60 -30 0 30 60 90 120 150 180 45 40 35 30 25 20 15 10 5 0 -180 -150 -120 -90 -60 -30 0 30 60 90 120 150 180 Crank Angle ATDC (deg) Crank Angle ATDC (deg) Fig. 9. Pressure traces for 15 cycles at different engine speeds and 5. 5 ms FPW: (a) 1250 rpm, (b) 1500 rpm, (c) 1750 rpm, (d) 2000 rpm. 40 35 10 Indicated Torque (N-m) 8 30 25 20 15 10 2 5 0 1000 Torque Power 6 4 250 1500 1750 2000 0 2250 . EngineSpeed(rpm) Fig. 10. Indicated torque and power versus engine speed for 5. 5 ms FPW. value of 36. 9 N m obtained at 1250 rpm. The engine could not be tested below 1250 rpm due to limitations of the dynamometer, as a result, the true peak torque could not be measured. The engine does not produce any measurable torque above 2000 rpm where unstable combustion sets in. The continuous drop in torque with engine speed, even at these relatively low engine speeds, is due to a decrease in time available for scavenging, fuel evaporation, and air–fuel mixing.
Higher engine speeds reduce scavenging time and lead to larger amounts of residual gas in the mixture. The delivery ratio trend ob- served in Table 2 re? ects this as less air is delivered to the engine with increasing engine speed. Increased amounts of residual gas means there is less air to mix with the fuel and slower ? ame speeds, both resulting in less torque. Increasing engine speed also reduces the time available for fuel evaporation and air–fuel mixing. Therefore, less well-mixed combustible air–fuel mixture is available at the time of ignition resulting in lower torque.
Higher engine speeds also reduce the time available for fuel evaporation. Injection ends at 79° BTDC at 1250 rpm and 54° BTDC at 2000 rpm. Piston wetting may increase since the piston is closer to TDC at the end of injection for higher engine speeds. The IMEP and ISFC results are shown in Fig. 11. Naturally, the IMEP follows the same trend as the indicated torque. A maximum IMEP of 570 kPa was achieved at 1270 rpm. The coef? cient of variation (COV) in the IMEP is roughly 3% for the three slower engine speeds and jumps to 5. 9% at the highest engine speed, see Table 2 for exact values.
A minimum ISFC of 771 g/kW h was obtained at 1270 rpm for this test session. The ISFC increases with increasing engine speed at a ? xed FPW of 5. 5 ms. This means that the returns in power for the fuel added are diminishing. Typical production small-bore carbureted two-stroke engines have BMEP in the 500 kPa range and assuming a typical mechanical ef? ciency of 0. 75 results in an IMEP of about 667 kPa. This value is higher than the 570 kPa found for the engine in this study. Furthermore the minimum ISFC value measured is well above that for engines of similar size.
This indicates that there are signi? cant inef? ciencies in the current engine design. Upon disassembly of the engine, it was observed that most of the piston was covered in carbon, due to the fuel-rich operation, except for an oval shaped area that IndicatedPower(kW) G. Ciccarelli et al. / Experimental Thermal and Fluid Science 34 (2010) 217–226 1000 223 800 600 400 200 IMEP ISFC 0 1000 1250 1500 1750 2000 2250 EngineSpeed (rpm) Fig. 11. IMEP and ISFC in verses engine speed at 5. 5 ms pulse width. was relatively clear of carbon, see the photograph of the combustion chamber in Fig. 1.
It appears that the fuel impacts the piston directly as a high momentum jet scouring away carbon deposits formed during the previous cycle. The fuel is de? ected towards the cylinder wall producing a fuel ? lm on the surfaces. Because the piston has a convex shape one can assume there is a substantial accumulation of liquid fuel around the piston and above the piston ring. Much of this liquid fuel is blown out the exhaust ports that can be seen at the bottom of the cylinder wall in Fig. 1. Some of the fuel evaporates when the fuel contacts the hot surface to form a fuel rich mixture next to the surface.
The subsequent combustion of the liquid fuel and the fuel rich mixture is responsible for the carbon deposits observed. Fuel jet impingement on the piston occurs for all the SOI times because of the relatively narrow cone angle of 40°. The engine performance can be dramatically improved by using a fuel injector with a wider cone angle and including a shallow bowl in the piston in order to produce squish near TDC. An alternative is to operate the engine in strati? ed mode where the fuel spray is de? ected by an appropriately design piston towards the spark plug.
Current research on this engine is geared in this direction. 5. Valve dynamics testing In a traditional engine the valve timing and duration are prescribed by the came pro? le, however in this novel engine such data is not available. An optically-accessible cold ? ow scavenging test rig was constructed to gather data on the check valve opening and closing timing as well as the platelet displacement time history. Valve displacement data is obtained during blowdown and scavenging that occurs after exhaust opens (EO), and during compression that occurs after exhaust closes (EC).
The test rig consists of the same basic elements as the engine and is built to the same scale, see Fig. 12 for a schematic of the test rig. The intake plenum is fed by a high pressure line regulated by a Watts R216 Precision Regulator which ensures a constant boost pressure in the plenum through the course of a test. The plenum pressure is monitored with an Omega PX219 strain gage-type pressure sensor. The cylinder head, machined out of clear cast acrylic, contains only a single valve body which houses a single mild steel check valve platelet.
The platelet is held up by the same stainless steel retaining plate used in the engine. The retaining plate is fastened to the bottom of the head by two steel bolts that penetrate through the head. Only a single valve is used so that the platelet motion could be visualized through the cylinder head. Clear cast acrylic was used for the cylinder head and the cylinder to allow the displacement of the check valve platelet to be visualized with a Photron FASTCAM 1024 PCI high speed digital video camera recording at 10,000 frames per second.
Ten frames from a video captured during the closing of the valve are provided in Fig. 13. The platelet starts to move in frame 1 and hits the top of the valve cavity between frames 7 and 8. After reaching its maximum displacement it rebounds and moves down towards the retaining plate in frames 8–10. Eventually the platelet changes direction again and moves up. This cyclic motion can repeat several times. The valve displacement measurements have an uncertainty of 10% due to the spatial resolution of the video camera.
The cylinder contains two diametrically-opposed banks of four exhaust ports located at BC which open at 120° ATDC and close at 240° ATDC. The cylinder pressure is monitored by a PCB Piezotronics 113A26 pressure transducer mounted near the cylinder head. As seen in Fig. 12 the piston is connected by a linkage, consisting of a connecting rod and a crank arm, to a drive shaft on which a Novotechnik P6500 potentiometer acts as a crank position sensor. The pressure sensors, crank position sensor as well as an accelerometer were linked through a National Instruments NI USB-6210 data acquisition (DAQ) board to a PC for data collection.
The PCB Piezotronics accelerometer was mounted on the cylinder head such that a vibration signal was transmitted when the check valve platelet struck the retaining plate or the cylinder head in the case of the check valve opening or closing, respectively. The accelerometer was used to synchronize the two data acquisition systems used, the high speed digital video camera and the DAQ. While the engine operates at a frequency between 1000 and 2000 rpm, the scavenging test rig is designed to operate as a single shot apparatus.
Two separate tests were performed to look at the dynamics of valve opening and valve closing. For studying the opening of the valve, the piston is initially held in place by a cable that is fastened to a 10 cm diameter drive shaft-mounted pulley. The opposite end of the cable is fastened to a steel plate that is held in place by an electromagnet. The intake plenum is pressurized to the appropriate scavenging pressure and the cylinder is pressurized to a higher pressure which is comparable to the pressures found in the engine during motoring or ? red engine cycles.
This action closes the check valve and pushes down on the piston making the test rig ready for a test. The electromagnet is deactivated releasing the constraint on the piston and allowing the gas in the cylinder to expand and accelerate the driveshaft. Although there is only one valve in the cylinder head of the test rig, there should be no difference in the response of the platelet compared to the case where numerous valves are employed since it is the pressure differential between the intake plenum and the cylinder that governs the reaction of the platelet.
The plenum pressure is constant throughout the course of a test and the cylinder pressure is determined by the initial cylinder pressure and the initial piston IMEP (kPa) / ISFC (g/kW-h) Fig. 12. Schematic of the optical valve test rig. 224 G. Ciccarelli et al. / Experimental Thermal and Fluid Science 34 (2010) 217–226 Fig. 13. Images taken from video recorded at 10,000 frames per second showing the platelet motion during valve closing (only every third frame shown). position. Thus the effect of employing one valve is negligible and simpli? es the experiment greatly.
The main difference between the test rig and the engine operating condition is that air is used in place of hot combustion products in the cylinder. This will directly in? uence the cylinder blowdown that starts once the exhaust ports start to open. The valve opening tests were performed with the plenum at a pressure of 150 kPa abs, similar to the boost used in the supercharged engine. In order to make a valid comparison between these tests and the engine, crank speed and cylinder pressure during valve opening need to be as close as possible.
The cylinder pressure transient recorded during engine operation at a speed of 1250 rpm is shown in Fig. 14. At roughly 145° ATDC there is substantial noise in the pressure signal which in the previous section was attributed to vibrations in the head mounted pressure sensor caused by the valve platelet contacting the retaining plate when fully open. A similar cylinder pressure transient, over the valve opening time between 135° ATDC and 145° ATDC, was achieved in the test rig by pressurizing the cylinder to 520 kPa at a starting crank position of 70° ATDC, see Fig. 4. Note the difference in pres- sure drop during blowdown between 110° and 135° is due to the large difference in speci? c heat ratio and speed of sound of the combustion products present in the engine and air used in the test rig. At roughly 143° ATDC there is some noise in the test rig pressure signal similar to that observed in the engine pressure signal. The signal noise level is substantially lower because the pressure transducer in the test rig was mounted in the plastic head which does not transmit vibrations as ef? ciently as in the metal engine head.
The measured crank sensor signal for this initial condition is shown in Fig. 15. The spike visible before 1. 5EA03 s is due to contact being lost within the potentiometer which serves as the crank position sensor. The crank angular velocity is fairly constant over the period that the valve opens, varying by less than 1° while the valve is in motion between 135° ATDC and 145° ATDC. Over this crank angle range the crank speed is roughly 1225 rpm. The measured test rig cylinder pressure transient shown in Fig. 14 is reproduced in Fig. 16.
The cylinder pressure initially drops solely due to the expanding cylinder volume to a pressure of 246 kPa at 120° ATDC. At this crank position the exhaust ports start to open and the blowdown phase starts during which the cylinder pressure drops more quickly. The cylinder pressure reaches 400 Cylinder Pressure (kPa abs) 350 300 250 200 150 100 50 0 110 Rig pressure (1225 rpm) Engine pressure (1250 rpm) 180 170 Crank Angles (deg) 160 150 140 130 Crank Position Signal Linearization 110 -1. 50E-03 120 120 130 140 150 160 170 180 5. 00E-04 2. 50E-03 4. 0E-03 Crank Position (deg) Fig. 14. Comparison of measured cylinder pressure recorded in engine operating at 1250 rpm and test rig at 1225 rpm and initial cylinder pressure of 520 kPa at 70° ATDC. 6. 50E-03 8. 50E-03 Time (s) Fig. 15. Crank position sensor signal (1225 rpm). G. Ciccarelli et al. / Experimental Thermal and Fluid Science 34 (2010) 217–226 350 Cylinder pressure Accelerometer signal Valve displacement 1 0. 8 0. 6 250 0. 4 200 0. 2 150 0 100 -0. 2 50 -0. 4 -0. 6 180 225 0 110 120 130 140 150 160 170 Crank Position (deg ATDC) Fig. 16.
Valve opening experiment results with the intake plenum at 150 kPa and initial cylinder pressure of 520 kPa at 70° ATDC. the plenum pressure of 150 kPa at roughly 135° ATDC. From the video recording the valve starts to open (i. e. , platelet starts to move) at 135° ATDC. As the valve opens the pressure continues to drop until a steady value equal to roughly atmospheric pressure is achieved, i. e. , 101 kPa. As mentioned above there is some low level noise in the pressure signal that occurs at 143° ATDC which was attributed to vibrations caused by the valve platelet impacting the retaining plate when fully open.
A more de? nitive measure of the impact of the valve platelet with the retaining plate was obtained by an accelerometer that was mounted directly on the steel bolt that fastens the retaining plate to the head. Note that in Fig. 16 the noise in the pressure and accelerometer signals occur at exactly the same time. The video data and pressure signal timing was synchronized in Figs. 15 and 16 by the onset of the large oscillations in the accelerometer signal and the maximum displacement in the platelet travel. It takes 23 CA after start of EO, or 3. 17 ms, for the valve to fully open at this crank speed.
Note the number of crank angles required for the pressure to drop from EO to 150 kPa depends directly the cylinder pressure at EO and on the crank speed. It is was very dif? cult to reproduce the same initial conditions in the cylinder pressure at the start of EO and the same crank speed exactly for each test in order to obtain statistics on the opening time. In three additional tests where the cylinder pressure at 120° ATDC was in the range of 252–258 kPa, and the crank speed was in the range of 1215–1237 rpm, the valve required 25 CA to open in all three cases, or roughly 3. 0 ms. Note the crank speed and initial cylinder pressure are not independent parameters in the operation of the test rig and thus the valve opening time could not be measured solely as a function of engine speed to compare to the engine results. Based on the data in Fig. 16 the valve is completely open at 143° ATDC. However, the platelet starts to move at 135° ATDC, thus only 8 CA are required for the valve to completely open once the platelet starts to move. The blowdown phase that occurs during the ? rst 15 CA in the test rig would be shorter in a ? ed engine due to the higher speed of sound of the combustion products that results in a higher mass ? ow rate out the exhaust. The duration of the blowdown phase is also governed by the cylinder pressure at EO, which is closely correlated to the maximum cylinder pressure. Speci? cally the higher the peak cylinder pressure the longer it would take for the cylinder pressure to drop to the plenum pressure and thus requiring a longer valve opening time. To investigate the closing of the check valve, the same test rig is employed with the initial piston position at BDC and air supplied to the plenum at 150 kPa.
At this piston position the exhaust ports are Valve Displacement (% opened) 300 fully open and a scavenging ? ow exists through the cylinder. At steady state the cylinder pressure is less than 1 kPa above atmospheric due to the pressure drop across the exhaust ports. At the start of the test the piston is accelerated towards TDC by a torque that is applied to the driveshaft pulley. A cable is attached to the pulley rim and partially wrapped around the top of the pulley with a weight suspended at the end of the cable. When activated the electromagnet prevents the weight from turning the pulley.
Deactivation of the electromagnet triggers the test with the weight dropping and accelerating the piston upwards, ? rst closing the exhaust ports and then building pressure in the cylinder to a level above the plenum pressure such that the check valves close. The crank speed was relatively constant over the period between EC at 240° ATDC and valve closing at roughly 260° ATDC. Using a weight of 5 kg the average crank speed was around 125 rpm. This crank speed is an order of magnitude lower than the engine speeds tested. However, as only one valve exists in the test rig, compared to the 16 valves which are present in the engine, the ? w rate through the single valve is on the order of magnitude of that through a single valve in the engine. The cylinder pressure, valve displacement and accelerometer output measured during a typical valve closing test is presented in Fig. 17. The measured average crank speed during valve displacement is 122 rpm. The cylinder pressure remains roughly constant from the piston starting position of 180° ATDC up to roughly 230° ATDC, which is just before EC at 240° ATDC. The cylinder pressure increases as the pressure drop across the exhaust ports increases with decreasing ? ow passage area.
After EC, the cylinder pressure increases due to air ? ow coming in from the plenum as well as a decreasing cylinder volume. There is very low level noise in the cylinder pressure at 258° ATDC which coincides with a large jump in the accelerometer signal. This corresponds to the valve platelet impacting the head when the valve is completely closed. Based on the video recording of the platelet displacement the valve starts to close at 256° ATDC, at the moment that the cylinder pressure is 155 kPa. Interestingly the video records a rebound in the platelet motion after it hits the head.
The ? rst hit of the platelet with the head occurs at 258° ATDC, thus requiring 18 CA for the platelet to traverse the cavity. This closing time is very repeatable. In ? ve tests the platelet starts to move on average at 255. 4 ± 0. 42 CA, and the ? rst hit with the head occurs at 257. 3 ± 0. 42 CA. As shown in Fig. 13 the video records a rebound in the platelet motion after it ? rst hits the head. The platelet motion oscillates up and down for a number of crank angles after the ? rst hit before coming to rest in the closed position. The valve is fully closed t 258° ATDC, thus requiring 18 CA to close after EC. Note the rebounding of the Cylinder Pressure (kPa abs) 180 Rig cylinder pressure Engine cylinder pressure Valve displacement Accelerometer 1. 0 0. 6 140 0. 4 120 0. 2 100 0. 0 80 -0. 2 60 220 230 240 250 260 -0. 4 270 Crank Position (deg ATDC) Fig. 17. Valve closing experiment results with the crank speed of 122 rpm. Valve Displacement (% closed) 160 0. 8 Cylinder Pressure (kPa) 226 G. Ciccarelli et al. / Experimental Thermal and Fluid Science 34 (2010) 217–226 platelet is exaggerated in the test rig because the head is made of plastic.
Based on the video images it takes up to 7 CA for the platelet to come to rest in the closed position after the initial hit with the head. The time required for the valve platelet to completely close to hit the head once the platelet starts to move, roughly 2 CA, is very short compared to the 18 CA required for the valve to close after EC. Clearly this is a limitation of the check valve approach that cannot be improved upon since the bulk of the closing time can be attributed to cylinder pressure increase after EC that is solely governed by the decrease in-cylinder volume.
This pressurization duration is expected to be largely independent of engine speed. Unlike in the case of valve opening it is much more dif? cult to directly compare the test rig results with what is observed in the engine. This is mainly due to the different number of valves and crank speed. For comparison, the pressure measured in the engine running at 1500 rpm is also shown in Fig. 17. As expected there is a noticeable difference in the cylinder pressure transient leading up to valve closing as measured in the test rig and the engine.
The valve closes at 260° ATDC in the engine which is surprisingly similar to the 258° ATDC measured in the test rig. It is possible that the effects from the test rig having a lower number of valves and slower crank speed compared to the engine cancel each other out. 6. Conclusions The engine equipped with the novel passive check valve concept was successfully operated. The opening and closing times of the valve were deduced from the onset of noise superimposed on the in-cylinder pressure time signals obtained under motored and ? red conditions. This approach for detecting the opening and closing valve times was veri? d by tests performed in an optically-accessible cold ? ow test rig where the dynamics of the valve were directly measured photographically and via the use of an accelerometer mounted to the head. Video of the valve platelet motion showed that it is very quick once it starts to move. Most of the valve actuation time is attributed to the time required for the cylinder pressure to decrease to below the plenum pressure during valve opening and to increase to above the plenum pressure during valve closing. This is a basic limitation of the check valve approach that effects both the scavenging time and the compression ratio.
The engine was successfully ? red up to a maximum speed of 2000 rpm. The poor engine performance measured can be directly attributed to poor air–fuel mixture preparation associated with the direct fuel injection. It was found that excessive piston wetting by the fuel jet was occurring. This leads to poor fuel evaporation and air–fuel mixing at all engine speeds. This piston wetting problem worsens with engine speed as the time available for evaporating and mixing decreases and thus can also be a major contributor to the upper engine speed operating limit.
Future work on the engine will involve a redesign of the fuel injection and combustion chamber design. Acknowledgments Funding for the research was provided by Materials and Manufacturing of Ontario and Armin Motors. A patent on the engine is held by Hans Ohlmann who is also the president of Armin Motors. The authors would like to acknowledge contributions to the development of the engine by Dirk Ohlmann and David Rival. References  J. B. Heywood, E. Sher, The Two-Stroke Cycle Engine, Taylor & Francis, Philadelphia, 1999.  Stan D. Shawcross, C. Pumphrey, D. Arnall, A ? ve-million kilometre, 100-vehicle ? et trial, of an air-assist direct fuel injected, automotive 2-stroke engine, Presented at SAE World Congress. Detroit, US, 2000.  D. Blundell, J. Turner, P. Duret, J. Lavy, J. Oscarsson, G. Emanuelsson, J. Bengtsson, T. Hammarstroem, M. Perotti, R. Kenny, G. Cunningham, Design and evaluation of the ELEVATE two-stroke automotive engine, SAE Paper 200301-0403, 2003.  Y. Moriyoshi, M. Arai, J. Katsuta, K. Morikawa, Performance tests of reverse uni? ow-type two-stroke gasoline DI engine, Presented at SAE Small Engine Technology Conference & Exposition, Graz, Austria, 2004. 5] M. Tazerout, O. Le Corre, S. Rousseau, TDC determination in IC engines based on the thermodynamic analysis of the temperature–entropy diagram, Presented at the SAE International Spring Fuels & Lubricants Meeting. Dearborn, MI, USA, 1999.  P. Oliver, S. Reynolds, G. Ciccarelli, Flowbench calibration of a numerical model for a novel uni? ow-scavenged two-stroke GDI engine, Presented at SAE Power Train Conference, Toronto, 2006.  J. B. Heywood, Internal Combustion Engine Fundamentals, McGraw-Hill, 1988.