Paper Number 2008-01-1350

Fully Flexible Electromagnetic Valve Actuator: Design, Modeling, and Measurements

David Cope, Andrew Wright Engineering Matters, Inc.

Christopher J. Corcoran Corcoran Engineering, Inc.

Kenneth Pasch Pasch Engineering Design

David Fischer DMF Associates

Copyright © 2008 SAE International

ABSTRACT

Design, modeling, and measurements are discussed for a new direct-drive, electromagnetic valve actuator (EVA) for internal combustion engines. The actuator offers greater dynamic performance than previous designs and achieves fully flexible valve timing and lift, which provides great benefits in engine performance including increased engine efficiency, reduced emissions, and improved low-end performance. The innovative actuator consists of simplified stationary permanent magnets, stationary coils, and a moving steel plunger that transmits significant bi-directional forces to the valve throughout the stroke. Single- and double-actuator magnetic configurations, non-linear electromagnetic finite element analyses, system bond graphs, Simulink® simulations, and performance measurements of the actuator controlling intake valves on a cylinder head will be discussed. Future work, including on-engine dynamometer testing and an actuator design variant for exhaust valves, is briefly discussed. INTRODUCTION

A new actuator for the gas exchange valves of a multi-valve multi-cylinder automotive engine has been fabricated, assembled, and demonstrated and is shown in Figure 1. Bench top and cylinder head testing has been performed and the results will be presented and discussed, and are in close agreement with the modeling analyses. The system power requirements are low enough to be operable from a standard 12V

automotive electrical system. These results demonstrate the actuator achieves the desired power, force, and control performance requirements for fully flexible EVA.

Figure 1. Fully flexible electromagnetic valve actuator.

In this paper, the designs, analyses and simulations are first discussed and then actuator measurements are described. The analytical and computer modeling consisted of electromagnetic, electrical and mechanical designs and analyses, including lumped parameter bond graph analysis, and dynamic and finite element simulation. The measurements consisted of laboratory bench top, cylinder head, and baseline engine dynamometer measurements.

1

Operation of the gas exchange valves is crucial to efficient operation of the internal combustion engine. Traditional fixed camshaft operation of the valves amounts to a one-size-fits-all approach to valve actuation, representing a static compromise over the changing engine operating conditions, particularly load and speed. Multi-step camshafts are improvements which offer distinct cam profiles employed as the engine operating conditions change. Mechanical cam phasing approaches significantly offer timing variations of the valve operations, but utilize the same cam profile. Other valve control approaches have also been undertaken. Specific valve actuation strategies, such as Early Intake Valve Closing (EIVC), Late Intake Valve Closing (LIVC), Late Intake Valve Opening (LIVO), and Variable Max Valve Lift (VMVL) have been experimentally verified to improve engine performance. These approaches and strategies are well-described in the literature. [1,2,3,4,5,6,7,8,9] It should be noted that current cam valve train technology has evolved to an extremely well-developed and thoroughly tested system setting a very high standard for replacement technologies. [4] Valve actuation includes valve timing, valve lift, and the duration of lift (or dwell). A fully flexible valve actuator (FFVA) approach can maximize the performance of the engine. FFVA allows for tailored valve actuation over engine load, speed, ambient temperature, air pressure (altitude), fuel type, and other variables. FFVA provides a completely variable valve motion, including variable valve timing, variable lift, and variable duration, limited only by the dynamic performance of the valve actuator. In addition, as demonstrated below, valve trajectory can be arbitrarily designed, i.e., valve open to partial valve lift, then increased lift, then reduced lift, and finally valve closed (zero lift). The well-known benefits of FFVA are improved fuel economy, increased low-end torque, and reduced emissions. [10, 11] Two complementary methods for electromagnetic valve actuation exist: commanded-holding and commanded-acceleration. In theory, both methods require only the initial energy input for operation. In detail, however, the power requirements and achievable dynamics of each method differ based upon the specific actuator properties such as force per current over the actuator stroke. A commanded-holding actuator relies on stored mechanical spring potential energy to be released into kinetic energy of an armature for transition, then storing once again the kinetic energy into mechanical spring potential energy to be held until the next desired transition. The control capabilities of this type of actuator are shown in Figure 2 indicating regions of strong control, near the end-of-stroke, and region of weak control away from the ends. In these regions of weak control, i.e., low force, control of the actuator is difficult. In addition to there being a weak force (small magnitude), the force is attraction-only (pull-only) since it

is based upon electromagnetic reluctance attraction. This means the actuator cannot dynamically adjust the direction of applied force, i.e., it cannot repel (push away) the armature.

Figure 2. Typical control capabilities of commanded-holding actuator showing large regions of weak actuator control. (AU = Arbitrary Units) A specific example of the commanded-holding actuator, the Pischinger design, uses solenoids to hold the valve either in the fully open or fully closed position while mechanical springs are in full compression [12]. When the solenoid releases the armature, the valve travels through the spring equilibrium point then decelerates to the end of the travel at which time another solenoid holds the armature in place. In this way, electromagnetic force is used only to hold the solenoid and spring force is used to accelerate/ decelerate the valve. In order to achieve the fast transition times required at high rpm, the spring constant is tuned to a resonance of the combined mass of the armature and valve. The selection of solenoid actuation is well suited to this method because of the one-way actuation (holding) and the fixed stroke range of appreciable force. The disadvantage of this method is the dependence on resonant transition allows only variations in the timing, but not the lift or speed of transition. While holding force alone is adequate for actuation, the inability to push the armature requires complex control for soft valve seating [13]. An actuator exhibiting aspects of both commanded-holding and commanded acceleration is the single-phase multipole actuator. See, for example, reference 16. In this type of actuator, typically associated with moving permanent magnets, there are mid-stroke regions of valve stroke with useful force and, conversely, low force, as illustrated in Figure 3. As in the previous actuator, control of the valve is difficult in regions of low force. In regions of high force, this actuator can both attract and repel the armature. In contrast to both of the above actuators, the commanded-acceleration actuator electrically exerts a significant bi-directional thrust on the armature throughout the valve stroke, and then reclaims the kinetic energy through regeneration. This is shown in Figure 4.

2

Figure 3. Typical performance capabilities of single-phase multipole actuator.

Figure 4. Performance capabilities of the present commanded-acceleration actuator. As can be seen, the commanded acceleration profile actuator provides high force capabilities over the entire stroke. There are no dead spots where the valve is outside of the control of the actuator. This constant force capability greatly simplifies the control system and improves overall actuator performance in that the system response is much less dependent upon the position of the actuator, there is no requirement to avoid certain regions of operation, and the actuator has no areas of weak control which must be compensated. The commanded-acceleration actuator, with force authority over the entire stroke, is capable of controlling the valve lift directly. A bi-directional force approach is also better suited to controlling the valve seating velocity. Potential difficulties with this method include obtaining enough force for high speed valve acceleration and maintaining reasonable power requirements. The benefits of the commanded-acceleration method include achieving the valve actuation strategies mentioned above utilizing variable valve lift, variable timing, and seating velocity control. ACTUATOR DESIGN AND ELECTROMAGNETICS

3

The electromagnetic actuator consists of stationary permanent magnets, stationary copper coils, and a

moving steel plunger that transmits bi-directional forces to the valve. The valve actuator is very rugged in design and suitable for integration into commercial automobiles. Experimental data presented below confirm basic operational capabilities, with both variable timing and lift. We have chosen to utilize a stock engine with stock engine components (valves, springs, etc.) instead of using highly specialized components made of exotic materials. This essentially amounts to a camless engine retrofit project and means the moving mass is relatively large. The project was initially focused on the intake valves since this provides the greatest performance gain in terms of fuel economy and low end performance. An exhaust valve actuator with increased peak force is certainly possible and is briefly mentioned below.

Figure 5. Electromagnetic Valve Actuator Design cross-section showing permanent magnet magnetization directions as the heavy arrows. The actuator is not axisymmetric, but has x- and y- planes of symmetry as it moves in the z direction (vertical). The fully flexible single valve actuator design is shown in Figure 5. There are essentially two magnetic circuits, one with permanent magnets and one with electromagnetic coils, which utilize some of the same components. These magnetic circuits are illustrated in Figure 6 and Figure 7, which show the magnetic flux lines of each magnetic circuit. Note these magnetic circuits are different from those suggested by Lequesne. [17] Significant differences of the present design are improved permanent magnetic circuits (long axially magnetized permanent magnets instead of thin radially magnetized permanent magnets) and reduced coil inductance, especially for the double actuator design. For the single valve actuator, the magnets are rectangular prisms with a central cylinder removed.

Figure 6. Schematic of permanent magnet flux paths showing the moving plunger has equal magnetic flux lines leaving from its upper and lower surfaces.

Figure 7. Schematic of electromagnetic coil flux paths with positive current showing the moving plunger has flux lines exiting its upper surface and entering its lower surface. The motion of the ferromagnetic plunger is a result of the vector superposition of these magnetic flux lines and controlled by the current. For example, for the current polarity shown in Figure 7. the permanent magnet (PM) flux above the plunger is reinforced by the coil flux and the flux below the plunger is canceled and the force on the plunger is upwards. By reversing the coil current direction, the flux in Figure 7 reverses (the flux in Figure 6 is constant) and the PM flux below the plunger is reinforced and the flux above the plunger is canceled resulting in a downward force. This is, in principle, a non-linear superposition of magnetic flux, since the materials may exhibit non-linear BH curves at the excitation levels involved.

Figure 8, Figure 9, and Figure 10 show the results of finite element analysis (FEA) of actuator operation using commercially available electromagnetic simulation software. [14] These figures are 2D calculations, but the remainder of simulation figures in this paper are the results of 3D calculations.

Figure 8. Single Valve Actuator magnetic flux lines calculated with Ansoft’s Maxwell FEA analysis program. The plunger is centered and the coil is energized with positive current. Due to the reinforcement of the flux at the upper surface and the cancellation of the flux at the lower surface, the net force on the plunger is upward (positive).

Figure 9. Single Valve Actuator magnetic field magnitude in interior regions showing fields concentrated in upper section of actuator. Leakage fields are less than 0.1T. The limited spatial volume available for installing a single valve actuator on a cylinder head is a challenging constraint. For multi-valves per cylinder, installing two closely adjacent non-interfering actuators is even more challenging. Therefore, an actuator configuration comprising a single actuator activating two valves (double valve actuator) was investigated and designed.

4

Figure 10. Single Valve Actuator magnetic field vectors. The plunger is centered and the coil is energized with positive current. Due to the reinforcement of the flux at the plunger upper surface and the cancellation of the flux at the lower surface, the net force on the plunger is upward (positive). Figure 11 shows a Double Valve Actuator in cross-section, similar to Figure 5. Note especially the magnetizations of the stationary permanent magnets and the stationary coils. Since both valves are energized in a similar manner, the magnetic fluxes of the two actuators (permanent magnets and coils) reinforce and there is a single upper permanent magnet and single lower permanent magnet. The coils could be combined as one racetrack coil for common excitation and a single figure 8 coil for differential excitation. The two valves in a double actuator may be independently controlled, but since they do share magnetic flux, there is an influence of one on the other which must be accounted for in the actuator controls.

Figure 11. Double Valve Actuator Design cross-section showing permanent magnet magnetization directions via the heavy arrows.

5

Figure 12 shows an isometric view of the double actuator. The permanent magnets have the shape of a rectangular prism with two nearly tangent cylinders

removed. Figure 13 shows magnetic field magnitude FEA results of actuator excitation.

Figure 12. Double Valve Actuator isometric view.

Figure 13. Double Valve Actuator top view showing magnetic field magnitude in horizontal mid-plane.

Figure 14 shows a side view of the actuator mounted on a cylinder head. This view also shows the air intake extension which, for experimental purposes, permits un-throttled supply of fuel-air mixture for low-lift tests.

Figure 14. Double Valve Actuator side view mounted on engine head.

ACTUATOR SIMULATION RESULTS

Magnetic finite element analysis software was used to compute the total electromagnetic actuator force (Lorentz force plus reluctance force) on each plunger as a function of current and position. Figure 15a shows the results of these calculations. [For convenience, the Lorentz force is defined as the force resulting from the actuator current and the reluctance force is defined as the current-independent force.] The center curve (with current turned off, defined as the reluctance force) shows a force that varies approximately linearly with the displacement from center (effectively a spring force with a negative spring constant). If the plunger is more open than closed, it tends to further open the valve. Alternatively, if the plunger is more closed than open, it tends to further close the valve. This actuator force provides a valve-closed holding force without expenditure of energy. The magnetic reluctance force is modeled as a spring with a negative spring constant (kmag = -62.5N/mm). This force is intentionally partially balanced by the addition of a mechanical spring with positive spring constant (kmech = +60N/mm) and still allowing a zero-power valve-closed holding force. This power-free holding force can easily be adjusted for turbo charging and other purposes. The addition of the linear mechanical springs adds a position-dependent force with constant slope (value = kmech ) to each of these three curves, effectively rotating all three. This raises the force at lift-off (at 0.0mm) by the amount labeled F0 as shown.

Figure 15a. Raw actuator electromagnetic forces on the plunger. The middle curve is the magnetic reluctance force curve with the current turned off. Energization of the current magnitude and polarity controls the actuator force and direction.

6

The resultant actuator forces including all three contributions are presented in Figure 15b with the lift-off force of ~350N as shown. Note that using a lesser spring rate will cause a greater holding force, while using a greater spring rate will provide greater valve

acceleration for faster transitions, but will require current to hold the valve in the closed position. Since the power associated with low current levels is small, this may be a desirable trade-off in some situations, i.e., for very high rpm designs.

Figure 15b. Electromagnetic forces with the mechanical spring (kmech = +60N/mm) added to the system to (mostly) cancel the magnetic reluctance force. With the addition of this spring, the electromagnetic lift-off force is equal to +350N as shown in the upper left corner of the figure. Note a small zero-power holding force of 10N exists at closed valve position (0mm). Additional closing forces could be provided actively or by choosing a smaller spring constant. A high level of control during the valve trajectory allows various acceleration profiles to be utilized. Two important acceleration profiles are discussed in the reference. [15] The first profile is a traditional constant acceleration profile, which requires the minimum force applied to the valve to achieve the desired dynamics. The second profile is a linearly decreasing acceleration profile, which can be demonstrated to require the minimum electrical energy for the desired dynamics. [15] These two profiles are shown in Figure 16. The solid curve is the constant force profile required to accomplish the transition in the allotted time. The dotted curve is the actuator acceleration which consumes the minimum energy to achieve the transition. As can be seen, the peak force required for the minimum energy profile is 50% greater than that required for the constant force curve. At low rpm, where low forces are required, the actuator would optimally operate with the minimum energy profile. However, as the rpm increases and greater forces are required to achieve the increasing demands of the valve dynamics, it may be required to transition to the constant force profile with its lower required peak force. The energy savings of the minimum energy profile amounts to 25%.

7

Figure 16. Two bounding valve acceleration profiles for valve-opening transition. To achieve the same valve dynamics, the solid curve indicates the minimum force profile and the dashed curve indicates the minimum energy profile. (Shown for 4,000 rpm and 8.0mm transition.) The performance of the actuator is critically dependent upon the total mass of the components that move with the valve. Using a stock automotive engine with stock engine components (valves, springs, etc.) has several favorable attributes, such as low component cost and ready component availability, but the trade-off is relatively large component mass. This was acceptable for this investigation for two primary reasons: the peak rpm with large mass is still acceptably high (greater than 5,000 rpm), and the improvement with lower masses is readily calculated. Figure 17 shows the breakdown of moving mass for the stock component actuator.

Figure 17. Moving mass breakdown for the single valve actuator and stock engine components. The slug is the portion of the actuator to which the force is applied. The stock valve has a mass of 48g. Total moving mass is 141g.

It should be noted that in order to achieve an actuator with constant dynamic performance (acceleration of the moving mass at the desired speeds), the required force, current, and power scale with mass as follows,

(massF FI∝ , ∝ , )222 massFIP ∝∝∝ . This indicates that minimizing the moving mass certainly has very significant benefits in force and required electrical power. [Note this scaling law assumes acceleration is dictated by the dynamic requirements of moving a certain distance in a certain time period.] Figure 18 and Figure 19 show the actuator required force and power as a function of engine rpm for the two acceleration profiles (fixed mass). Figure 20 shows a plot of achievable rpm as a function of mass (fixed force).

Figure 18. Required maximum force vs. rpm for acceleration profiles of Figure 16 and mass = 141g. The solid curve is the constant force profile and the dashed curve is the minimum energy (minimum average power) profile.

Figure 19. Average electrical power per valve vs. rpm for acceleration profiles of Figure 16 with moving mass = 141g. Solid curve indicates the constant force profile and dashed curved indicates the minimum energy profile.

Figure 20. Achievable rpm vs. moving mass for the two acceleration profiles. Results indicate the actuator ideally can achieve 5,500 rpm using stock engine components and the minimum force profile. When the moving mass is reduced to 120g the actuator should achieve the desired 6,000 rpm. In order to predict system response during high speed operation, a model was developed to determine the frequency response of the actuator modeled as a simple spring coupled mass with viscous damping. Note this analysis was performed with a greater moving mass (150g). The results are presented in Figure 21 which shows the actuator position response per unit current (mm/A) as a function of frequency (Hz). With the ability to activate the coils with currents up to 10A, a simple inspection of Figure 21 indicates that the actuator is capable of achieving significant response amplitudes at frequencies up to 150 Hz (6,000 rpm). A more complete prediction of valve performance will be achieved when the results of this model are coupled with a detailed Fourier analysis of the desired valve trajectory and with the anticipated reduction in moving mass.

Figure 21. Actuator frequency response (displacement per current at a given frequency) for a mass = 150g. The predicted response is the solid line and the dots are measured data.

The presence of eddy currents in the actuator structure manifests itself in the form of viscous damping. This results in a reduction of the actuator performance, especially at greater speeds. The eddy currents can be reduced through appropriate design of the actuator conductive structures surrounding the magnetic paths in the actuator. ACTUATOR CONTROLS

Actuator control was approached from two parallel viewpoints: traditional block diagram and bond graph analyses. The traditional block diagram model was developed in Simulink® with non-linear electromagnetic forces computed in Maxwell 3D® using a conventional PID controller as shown in Figure 22. A mechanical spring is utilized to counter the actuator magnetic reluctance force, as discussed above. Figure 23 shows the difference between the desired and realized (simulated) valve trajectories indicating very close agreement. Figure 24 shows the force demanded from the actuator as a function of time. As shown, at 5,000 rpm the required force is well within the actuator’s capabilities.

Figure 22. Simplified Block Diagram for PID closed-loop control of valve actuator utilizing the calculated non-linear Lorentz and reluctance forces and mechanical spring forces.

Figure 23. Difference between the desired valve position and simulated resultant valve position profiles for the block diagram of the previous figure at 8.0mm and 5,000 rpm showing very close agreement.

8

Figure 24. Actuator force profile for the full 8mm stroke at 5,000 rpm. The full force of the actuator in this analysis was +/-300N. When the valve is closed (8ms to 15ms), the magnetic reluctance force holds the valve closed without any expenditure of energy. Bond graphs are a graphical analytical technique based upon continuity of power and are very useful for multi-domain analyses, such as associated with an electromagnetic valve actuator with mechanical springs. The domains considered include magnetic, mechanical, and electrical. Figure 25 shows a simplified lumped parameter magnetic model of the actuator. This magnetic model, as well as the mechanical and electrical models (not shown), are used to create the bond graph shown in Figure 26.

Figure 25. Simplified lumped parameter magnetic circuit model used for bond graph analysis. Leakage flux paths are estimated analytically and shown as the flux lines that travel outside of the

actuator. The magnetic nodes are labeled Mn: n = 0 - 9.

Figure 26. Actuator bond graph including magnetic, mechanical, and electrical elements, and making topological reductions. The “O” junctions are generalized efforts (i.e., magnetomotive, force, voltage) and the “1” junctions are generalized flows (i.e., flux, velocity, current). The bond graph is inserted into a close-loop control block diagram as shown in Figure 27. The major block titled “EVA Subsystem” contains the reduced bond graph model with many interrelated physical aspects of the model. One of the great benefits of the bond graph approach is the transparency of the bi-directional energetic flows. This allows parameters such as actuator back-emf to be readily extracted.

Figure 27. Simplified block diagram of closed loop bond graph dynamic simulation utilizing electromechanical gap interactions, coupled spring-

9

mass subsystems, non-linear magnetics, and PID current loops. Figure 28 shows simulated results for actuator current, valve position, valve velocity, actuator force, and applied voltage, respectively at 4000 rpm.

10

Figure 28. Bond graph simulation results for actuator current, valve position, valve velocity, actuator force, and applied voltage (including BEMF). Commanded and predicted results are shown for position and velocity. The results of a bond graph lumped parameter simulation show the practicality of this electromagnetic actuator. Specifically, the large moving mass (141g) associated with stock engine components can be exercised at over 4,000 rpm and still achieve well-controlled valve dynamics. These simulation results are in agreement with the traditional block diagram simulation discussed above with the detailed FEA magnetic forces used. Furthermore, these results and the block diagram simulation are in agreement with on-cylinder head measurements. The bond graph simulation for reduced moving mass, corresponding to non-stock engine components, shows that 6,000+ rpm is achievable, if the moving mass can be reduced to below 120g. This is, indeed, a significant reduction in mass, but it can be achieved with commercially available components. Table 1 below summarizes the double actuator operation with stock engine components on a single valve basis. Note since the power is reasonably small, the energy source could be a 12V battery. A voltage converter could be used as a convenience, but is not necessary for operation. The number of turns and wire gauge of the coils would simply be changed to lower the required voltage (along with the increase in required current).

Table 1. Intake Valve Actuator Characteristics

Parameter Value Actuator width (transverse to cylinders) 75mm Actuator length (two valves, along cylinders) 95mm Actuator height 51mm Number of coils 2 Number of coil turns 630 Resistance per coil 6.3Ω Inductance 10.5mH Voltage 100V Current capability 10A Mass of actuator (single valve) 1.45kg Moving mass without valve (stock springs) 93g Moving mass with stock valve and springs 141g Distance between two valves 37.5mm Distance between two cylinder centers 85mm Stroke 8.0mm Minimum non-zero lift < 0.1mm Maximum actuator force 350N Minimum seating velocity ~10mm/s Power consumption 1,000 rpm < 10W Power consumption 6,000 rpm 150-200W

ACTUATOR BENCHTOP PERFORMANCE

A single valve actuator was constructed for bench testing utilizing the above design. Figure 29 shows the components of the actuator spread out before assembly and Figure 30 shows the assembled actuator. Figure 31 shows the actuator in a test set-up for measuring the position and force.

Figure 29. Single Valve Actuator components in exploded view. The two magnets are shown out of position and laid flat for viewing.

Figure 30. Single Valve Actuator Assembly.

11

Figure 31. Single Valve Actuator in force testing jig (top view). Figure 32 and Figure 33 show the measured actuator force as a function of valve position and coil current. In the latter figure, the current is increased to its maximum value suitable for high speed engine operation (6,000 rpm). For these measurements, the upper and lower coils were energized in series (same current in both coils). Analyses have indicated that the same current in the upper and lower coils is close to, but is not quite, the optimal distribution of current. In the future, the coils could be energized using separate power supplies independently tailoring the current in the upper and lower coils to the valve position. This will lead to further improvements in actuator performance. The force measurements were taken with an Omega DFG60-110 force gage, the position sensor measurements were taken with a Schaevitz MHR-250 LVDT, and the current measurements were made with a Tektronix A622 AC/DC Current Probe. The Lorentz forces were calculated by subtracting the reluctance

forces (zero current forces) from the forces measured at each current value.

Figure 32. Total Actuator Force (with spring pair yielding about 60N holding force) as a function of position with current as a parameter. (0mm indicates the valve is closed and the valve has a maximum lift of 8mm).

Figure 33. Measured actuator Lorentz force as a function of actuator current for different values of valve lift with spring pair. This data demonstrates that high actuator forces are possible for all values of valve lift. ACTUATOR CYLINDER HEAD INSTALLATION

A double valve actuator was designed to control the intake valves on a Ford Duratec 2.3L in-line 4 cylinder, 16 valve engine, as shown in Figure 34. The assembled actuator is shown in Figure 35. Initially, the actuator is mounted on the last cylinder position, i.e., the one furthest away from the timing belt, as shown in Figure 36. Actuator control of a single cylinder operation is an important milestone and should be demonstrated first, followed by control of all four cylinders. To this end the camshaft was cut short, removing the lobes for the fourth cylinder and leaving the lobes for three cylinders,

Actuator

Springs

Load

as shown in Figure 37. The camshaft oil cover was modified to allow the actuator to be mounted next to the working camshaft for the first three cylinders.

Figure 34. Double Valve Actuator shown on intake valves of last cylinder of Duratec 2.3L engine.

Figure 35. Double Valve Actuator Assembly.

Figure 36. Double Valve Actuator installed on the Duratec cylinder head over the final cylinder.

Figure 37. Unmodified stock camshaft (top) and modified (shortened, on bottom) experimental camshaft to make room for the actuator. Note the timing pulse provided by the tab at the end of the stock camshaft was moved to the exhaust camshaft. Modified pocketed pistons, shown in Figure 38, were installed to ensure piston-valve clearance at all times and allow for uncontrolled actuator excursions possible during testing. If valve control breaks down or an actuator misfires, the valve will not interfere with the piston.

Figure 38. Modified piston showing valve cut-outs to prevent piston-valve interference in the case of timing errors during laboratory tests. A digital switched mode power amplifier was used to energize the actuator. The current sensors used to measure the actual current in the coils were LEM LTS6-NP. The actual and commanded currents are compared in a digital PID (Proportional-Integral-Derivative) closed-loop controller to adjust the PWM (Pulse Width Modulation) duty cycle. PWM applies variable duty cycle pulses of either polarity (positive or negative) across the coils to drive the desired current. ON-HEAD VALVE LIFT MEASUREMENTS

As discussed above, valve lift profiles can be arbitrarily controlled. For convenience, initial valve profiles are taken to have opening transitions, constant open dwell times and closing transitions. A sample lift profile is illustrated in Figure 39.

12

Routine actuator position measurements are performed with the LVDT described above. For precise laboratory measurements, the position and velocity were measured with a Polytec OFV-303 Laser Doppler Vibrometer operating at 20kHz. The ultimate goal, however, is to have a sensorless actuator. As mentioned above, the Bond Graph analyses support development of sensorless control actuator. The physical reality is that the magnetically reluctant actuator is essentially a LVDT and hence should allow accurate sensorless measurements and control. Sensorless control development, though, will be postponed until more is understood about the application and actuator dynamics.

Figure 39. Valve lift profile showing Opening, Dwell, Closing, and Closed periods. For convenience, the time base is exaggerated.

Figure 40 shows various actuator lifts with constant valve open dwells. The chosen lifts were arbitrary and there are no unachievable lifts between fully closed and fully open.

Figure 40. Actuator operation demonstrating various lift capabilities. Lift is continuously controllable to any desired lift up to the maximum designed lift. Figure 41 shows constant lift and variable dwells. The dwell times chosen for presentation were arbitrary and

there are no unachievable dwell times, including continuously open or closed.

Figure 41. Actuator operation at 8mm lift and various dwells. Dwell time is continuously and arbitrarily controllable. Figure 42, although rather complex, shows four lifts and many dwells demonstrating the flexibility of valve motion. All these measurements were made on the cylinder head at 1,000 rpm. Greater rpm's are, of course, possible and generally amount to reducing the spacing between valve lift waveforms.

Figure 42. Actuator operation demonstrating various combinations of lifts and dwells. As a further demonstration of the flexibility of the actuator control in achieving arbitrary waveforms, the actuator commanded the valve open through a controlled series of positions as follows: 0mm, 5mm, 8mm, 3mm, and finally closed once again (0mm). These particular steps were chosen arbitrarily and the actuator could have been program to achieve any arbitrary lift profile within its bandwidth. The results of this demonstration are presented in Figure 43.

13

Figure 43. Experimental demonstration of the valve being controlled in a series of steps to demonstrate the fully flexible control of the actuator to achieve continuously and arbitrarily controllable lift profiles. Again, the time base was adjusted for convenience. For control of the valve during these measurements, the valve was basically assumed to be a spring-coupled moving mass with viscous damping and Coulomb friction. Valve position PD (proportional-derivative) control was implemented with velocity feedback and a feed-forward current command component was used for friction compensation. The feed-forward approach was used since friction was relatively constant and could be anticipated and compensated. The remaining commanded current component was determined by the PD controller. The valve velocity was fed-back as the derivative term. The magnetic (negative) and the mechanical (positive) springs approximately canceled each other, although the non-linear magnetic spring was stronger near the end of stroke. DYNAMOMETER MEASUREMENTS

Final on-engine actuator tests will be performed on an M&W P-6500 dynamometer. For baseline performance tests, an unmodified engine was exercised and measured on the dynamometer. Figure 44 shows the baseline engine undergoing tests. Figure 45 and Figure 46 show the diagnostic equipment used for these measurements. There are two basic types of measurements made with the dynamometer: slow speed and high speed measurements. Slow speed data changes on a time scale of about 1 second and includes fundamental information such as engine torque, engine speed, and fluid temperatures. High speed data changes within an engine cycle and includes such information as: cylinder pressure as a function of cylinder volume, actuator position and current as functions of time, etc. Figure 47 shows baseline slow speed data for engine torque and power measurements and Figure 48 shows baseline PV data.

14

Figure 44. Duratec 2.3L engine installed on an M&W P-6500 dynamometer for baseline tests.

Figure 45. Dynamometer diagnostic equipment. This equipment allowed quick and easy monitoring of the performance of unmodified and modified cylinders (the modified cylinder is the cylinder with the electromagnetic valve installed). By intent, high speed cycle-by-cycle performance data will enable quick tuning of the actuator performance to match the camshaft-driven cylinders. In response to these measurements, adjustments to the EVA trajectory will be made to improve its performance. As of this writing, the on-cylinder head tests are finishing and the actuator is being transitioned to the engine for dynamometer testing to compare with the baseline engine tests. During a complete engine cycle, useful work is done during the combustion power stroke. The measured PV (cylinder pressure vs. cylinder volume) curve, shown in Figure 48, is the baseline engine work per cycle and is measured by the net enclosed area of the curve. On the plot the power stroke provides work output and is represented by a clockwise traversal, i.e., the open loop of data, and the compression portion of the cycle requires work input and is traversed in a counter-clockwise manner.

Figure 46. Close-up of diagnostic data acquisition screen (colors reversed for greater visibility).

Figure 47. Duratec 2.3L 4 cylinder baseline engine torque and power curves as a function of engine speed.

Figure 48. Pressure-Volume diagram for the baseline engine taken at full throttle, low rpm and under inertial load.

FUTURE WORK

Future work includes the following: 1. Installing the actuator on a single cylinder of an

engine for dynamometer testing; then installing it on multiple engine cylinders.

2. Implementing low moving-mass engine components (mass reduction to 108 g identified).

3. Designing an exhaust valve actuator variant concentrating on high peak force capability by increasing Lorentz forces to quickly open the exhaust valve at reduced stroke length.

4. Developing sensorless actuator control. CONCLUSIONS

In summary, a new electromagnetic actuator capable of controlling the intake valves on an internal combustion engine was designed, manufactured, and demonstrated. The ability to control the force over the entire range of motion provides the capability to specifically tailor the lift profile for all speed and load conditions. This will also enable the elimination of the throttle. The control electronics were demonstrated to open and close the valve in a wide variety of lift heights and dwell times. This new capability will provide a significant improvement in engine performance in terms of low end torque, emissions, and fuel economy. Based upon the analyses and measurements performed, this fully flexible electromagnetic valve actuator will operate up to 5,000 rpm on the intake valves using stock engine components. Utilizing low mass valve components will enable actuator operation at over 6,000 rpm. An exhaust valve actuator is possible through optimizing the specific actuator design for increased peak force. The next step in development of this technology will be installing the electromagnetic actuator on a multi-cylinder multi-valve engine and utilizing the actuator to control intake gas exchange during engine operation. This demonstration is currently being performed. ACKNOWLEDGMENTS

The authors would like to thank Neil Tischler of Tischler Resources for mechanical design and manufacturing inputs. The authors would also like to thank Jon Baroud for help with experimental testing and Steve Reynolds for assistance with actuator integration and testing. In addition, the authors would like to thank the manuscript reviewers for useful and constructive comments. The authors acknowledge the support of the National Science Foundation through its SBIR program under grant DMI-0522170.

15

16

REFERENCES

1. Wang, Y., Introduction to Engine Valvetrains, SAE International, Warrendale, PA, 2007, p. 25ff. ISBN-13: 978-0-7680-1079-4

2. Stone, R., and Kwan, E., “Variable Valve Actuation Mechanisms and the Potential for their Application,” SAE Paper 890673, 1989.

3. Klein, F. et al, “The Influence of the Valve Stroke Design in Variable Valve Timing Systems on Load Cycle, Mixture Formation and the Combustion Process in Conjunction with Throttle-free Load Governing,” SAE Paper 981030, 1998.

4. Sellnau, M., and Rask, E., “Two-Step Variable Valve Actuation for Fuel Economy, Emissions, and Performance, SAE Paper 2003-01-0029, 2003.

5. Ahmad, T., Theobald, M., “A Survey of Variable-Valve-Actuation Technology,” SAE Paper 891674, 1989.

6. Dresner, T., Barken, P., “A Review and Classification of Variable Valve Timing Mechanisms,” SAE Paper 890674, 1989.

7. Schechter, M., and Levin, M., “Camless Engine,” SAE Paper 960581, 1996.

8. Pierik, R., Burkhard, J., “Design and Development of a Mechanical Variable Valve Actuation System,” SAE Paper 2000-01-3307, 2000.

9. Kreuter, P. et al, “The Meta VVH System – A Continuously Variable Valve Timing System,” SAE Paper 980765, 1998.

10. Carney, D. "Internal Combustion Engineering," Automotive Engineering International, May 2004, p. 42;

11. "European Centers of Power," Automotive Engineering International, June 2004, p. 67.

12. Pischinger F., Kreuter, P., "Arrangement for Electromagnetically Operated Actuators", U.S. Patent #4,515,343, May 7, 1985, www.uspto.gov.

13. Wang, Y., Peterson, K., et al, “Modeling and Control of Electromechanical Valve Actuator,” SAE Paper 2002-01-1106, 2002.

14. Ansoft Corp., Pittsburgh, PA 412-261-3200, www.ansoft.com.

15. Cope, D., Wright, A., “Electromagnetic Fully Flexible Valve Actuator”, SAE Paper 2006-01-0044, 2006.

16. Hoffman, Bernhard, “Fully Variable Valve actuation with electromagnetic Linear Motor”, SIA Conference on Variable Valve Actuation, 30 November 2006, IFP Rueil.

17. Lequesne, B., “Variable Lift Operation of Bistable Electromechanical Poppet Valve”, US Patent 4,829,947, 16 May 1989, www.uspto.gov.

CONTACT

The authors may be contacted at: Engineering Matters, Inc. 375 Elliot St., Suite 130K Newton, MA 02464 617-965-8974 Dr. David Cope: dcope@engineeringmatters.com DEFINITIONS, ACRONYMS, ABBREVIATIONS

Commanded-holding: Class of actuator offering variable timing and fixed lift based upon an electromagnetic solenoid holding a mass against the force of compressed springs. Commanded-acceleration: Class of actuator offering variable lift and variable timing based upon continuous application of electromagnetic force throughout actuator stroke. EVA: Electromagnetic Valve Actuator FFVA: Fully Flexible Valve Actuator ICE: Internal Combustion Engine Lorentz force: Electromagnetic force dependent upon applied current. NSF: National Science Foundation PM: permanent magnet Reluctance force: Electromagnetic force independent of current.