Cycle Adaptive Feedforward Approach Control of an Electromagnetic Valve Actuator Jimmy Tsai, Charles Robert Koch, and Mehrdad Saif Abstract— An electromagnetic valvetrain on an internal com- bustion engine can improve the engine thermal efficiency but requires control to achieve soft landing and to avoid excessive wear and noise. Since the valves open and close repetitively, cycle adaptive control can be utilized. A cyclic adaptive feedfor- ward approach controller for automotive electromagnetic valve is presented. This method uses a Nelder-Mead direct search algorithm with the goal of setting constant initial conditions for the landing control. Simulation and testbench results are presented and they show that the approach control works well for disturbances that are slow compared to the valve travel time. I. INTRODUCTION The coupling between crankshaft and engine valve op- eration presents an area of potential improvement for the internal combustion engine. If this coupling is removed, the engine valve operation can be optimized at different operating conditions using variable valve timing (VVT) [1] [2]. While many variable valve timing systems are avail- able, the promise of improved engine performance, emission and fuel efficiency provides a strong motivation to develop camless valvetrains [3]. Candidate actuators considered for camshaft replacement include hydraulic [4], rotary motor [5], piezoelectric [6], and electromagnetic solenoid actuator [7]. Of all these actuators, the solenoid/electromagnetic valve actuator (EMV) excels in its cost, efficiency, and ruggedness [8]. The EMV actuator considered here consists of two springs, two solenoids, and one shaft that connects a metal armature to the valve (See Figure 1). The springs are pre- loaded in compression evenly to store energy. When there is no current on either coil of the actuator, the armature rest position is in the middle due to the balanced spring forces. To open or close a valve, each of the two solenoids acts as electromagnet to attract the armature to the respective end of the actuator. The energy needed for travel is mostly recovered because it is stored in the springs. The solenoids are needed only for the additional pull to land and hold the armature. Without control, the EMV actuator control tends to suffer from excessive valve seating and the resultant premature- wear and acoustic emission [10]. The control problem arises from the low force and low control authority at large airgaps and high inductance and reduced bandwidth at at small airgaps. Specifically in the EMV actuator, the magnetic Jimmy Tsai is with School of Engineering Science, Simon Fraser Uni- versity, Vancouver, B.C. V5A 1S6, Canada jtsai@sfu.ca Charles Robert Koch is with Department of Mechanical Engineer- ing, University of Alberta, Edmonton, Alberta T6G 2G8, Canada bob.koch@ualberta.ca Mehrdad Saif is with School of Engineering Science, Simon Fraser University, Vancouver, B.C. V5A 1S6, Canada saif@ensc.sfu.ca Fig. 1. The EMV actuator : actual (left) and schematic (right) [9] force drops off inversely proportional with the square of gap distance whereas the system inductance increases with gap distance [11]. Additionally, the exhaust valve has to overcome large pressure disturbances related to valve timing and engine load [12]. For effective control, the controller can be divided into two parts [13]. The landing controller is active close to the catching coil where there is enough control authority for it to track a smooth landing trajectory, while the approach controller operates over the remaining trajectory to keep the initial conditions constant for the landing control. The position-velocity plot in Figure 2 shows where the two controllers act. The important position in the plot is the end of approach control, which is also the beginning of the landing control: x i land = x f appr =2.55mm. The superscripts and subscripts i, f , land, and appr stands for “initial”, “final”, “landing control” and “approach control” respectively. Like the dashed line connecting the disturbed to the ideal valve lift in Figure 2, the approach controller compensates any disturbances so that at x f appr : v f appr = v i land = v d and i f appr = i i land = i d Published works on EMV actuator control, both on the approach controller and the landing controller, are extensive. A simplified relationship between current measurement with the armature velocity over position ratio is presented in [14] to facilitate “sensorless control”, which uses no posi- tion/velocity measurement. An LQ optimal controller based on a linearized system model is used in [15]. Later, the sensorless control is improved by adding in take-off and Proceedings of the 47th IEEE Conference on Decision and Control Cancun, Mexico, Dec. 9-11, 2008 ThC18.3 978-1-4244-3124-3/08/$25.00 ©2008 IEEE 5698