Sunteți pe pagina 1din 12

184

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 52, NO. 1, JANUARY 2003

Nonlinear Air-to-Fuel Ratio and Engine Speed Control for Hybrid Vehicles
John R. Wagner, Member, IEEE, Darren M. Dawson, Senior Member, IEEE, and Liu Zeyu
AbstractInternal combustion spark ignition engine management systems regulate the fuel, spark, and idle air subsystems to achieve sufficient engine performance at acceptable fuel economy and tailpipe emission levels. Engine control units also monitor other engine processes, using a suite of sensors, and periodically check the system actuators operation to satisfy legislated onboard diagnostics. The majority of production engines regulate the air-to-fuel ratio using a speed-density, or air-flow, control strategy. In this approach, the mass of air drawn into a given cylinder is calculated using the engine speed, manifold absolute pressure, and inlet air temperature. Based on the air mass, appropriate fuel amounts are injected to achieve stoichiometric operation. However, the wide range of operating conditions, inherent induction process nonlinearities, and gradual component degradations due to aging have prompted research into model-based algorithms. In this paper, a nonlinear model-based control strategy will be proposed for simultaneous air-to-fuel ratio control and speed tracking in hybrid electric vehicles. The motivation for engine speed management resides in the integrated control of the engine and a continuously variable transmission for increased efficiency. The proposed backstepping controller uses an observer to reduce the inputs to manifold air mass (e.g., manifold absolute pressure and inlet air temperature) and engine speed. The underlying engine model describes the air intake, fuel injection, and rotational dynamics. For comparison purposes, an existing multisurface sliding mode controller and an integrated speed-density air-to-fuel controller with attached engine speed regulation have been implemented. The performance of each controller will be studied using an analytical engine model with representative numerical results presented and discussed to provide insight into the overall performances. Index TermsDisturbances, engines, modeling, nonlinear control, road vehicles, simulation.

ESC IAC

MAP

PRI

SGN SI TC

NOMENCLATURE AFI AFR AFS Air-to-fuel influence. Air-to-fuel ratio. Air-flow sensor. Model parameters. Intake manifold parameter. Torque constant. Continuously variable transmission. Controller error signal. Air-to-fuel ratio error. Engine speed error. engine control unit.

CVT

ECU

Manuscript received October 2000; revised April 2002. The authors are with the Automotive Research Laboratory, Department of Mechanical and Electrical/Computer Engineering, Clemson University, Clemson, SC 29634-0921 USA. Digital Object Identifier 10.1109/TVT.2002.807156

Electronic spark control. Frequency (Hz). Idle air control. Effective engine inertia (Nm rad s ). Robustness parameter. Derivative gain. Integral gain. Proportional gain. Manifold absolute pressure (kPa). Intake manifold air mass (kg). Intake manifold air mass flow rate (kg). Maximum air flow rate (kg/s). Inlet manifold air flow rate (kg/s). Manifold exit air mass to cylinder (kg). Manifold exit air mass flow rate (kg/s). EGR flow rate (kg/s). Commanded fuel flow rate (kg/s). Inlet fuel flow rate (kg/s). Fuel inlet mass (kg). Engine speed (rpm). Oxygen sensor. Atmospheric pressure (kPa). Manifold absolute pressure (kPa). Normalized pressure ratio characteristic. Air gas constant (KJ kg K). Sliding surface. Sign function. Spark ignition. Normalized throttle characteristic. Indicated torque (Nm), temperature (K). Engine load torque (Nm). Controller output signal. Lyapunov function. Cylinder volume (m ). Engine volume (m ). Intake manifold volume (m ). Oxygen sensor binary signal output. Crank angle (rad). Engine speed (rad/s). Throttle angle (%). Nominal throttle angle (%). Stoichiometric ratio. Robustness parameter. Air-to-fuel ratio. Induction process model error. Maximum induction process model error. Fuel delivery model error. Maximum fuel delivery model error.

0018-9545/03$17.00 2003 IEEE

WAGNER et al.: NONLINEAR AIR-TO-FUEL RATIO AND ENGINE SPEED CONTROL FOR HYBRID VEHICLES

185

Fig. 1.

Engine management system.

Volumetric efficiency. Manifold air density (Kg m ). Time delay. Time constant (s). Controller gain. Engine speed disturbance (rpm). Standard deviation of the air-to-fuel ratio.

A. Subscripts and Superscripts Estimate. Set point. Average. I. INTRODUCTION HE legislated regulations for emissions, fuel economy, and diagnostic standards in spark ignition internal combustion engines by the U.S. Environmental Protection Agency and State of California Air Resources Board have been one of the key

drivers for automotive technology. Engine, or powertrain (if integrated transmission control is featured), management systems typically consist of an engine control unit and a common core of sensors and actuators (Fig. 1). The principal engine sensors include crank position/engine speed, manifold absolute pressure or airflow sensor, throttle position, coolant temperature, and exhaust gas oxygen (O2), while the principal engine actuators include the fuel injectors and electronic spark [1]. The ECU most likely contains a hybrid 32-bit microprocessor (i.e., designed specifically for automotive applications) with on-chip high-speed timing circuits, and interfaces with analog and digital input/output resources, power-supply circuitry, and vehicle bus communication hardware. Closed-loop fuel control, either speed-density or air flow, is generally implemented using the engine speed, air temperature, MAP or air-flow rate, and O2 sensor information to regulate the injection process [2]. A speed-density control system with ESC is presented in Fig. 2. In this diagram, air flows through the throttle body via the throttle plate (either manually or electronically adjusted) into the intake manifold with auxillary air

186

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 52, NO. 1, JANUARY 2003

Fig. 2.

ECU and internal combustion engine schematic.

flow regulated by the idle air subsystem. The ECU sends out pulse-width modulated signals to the fuel-injection hardware. Fuel may be injected into the air mixture using throttle body, intake manifold (i.e., port fuel), or direct cylinder injection strategies. The ECU delivers an ESC digital signal to the ignition module and coil to ignite the airfuel mixture at the appropriate instance. The spark ignition control system typically uses lookup tables with engine speed and load (i.e., MAP or AFS) inputs to determine the dwell angle and spark advance outputs. Once the combustion process is completed, the exhaust gases are expelled into the exhaust manifold and pollutants removed in the catalytic converter. Located in the exhaust manifold is an oxygen sensor that generates a binary feedback signal in relation to the oxygen concentration in the exhaust gases. The fuel control system adjusts the fuel flow rate to achieve a stoichiometric air-to-fuel ratio using table lookups with proportional-integral (PI) control. Although these closed-loop air-to-fuel ratio control strategies offer satisfactory performance, the wide range of engine operating conditions and the inherent nonlinear induction process creates a nontrivial calibration process. Therefore, a need exists to leverage the advantages of model-based controls. A variety of research has been conducted during the past decade on model-based air-to-fuel ratio control for improved fuel economy and reduced emissions. Engine models for model-based control applications have been developed by Moskwa and Hedrick [3] and later refined by Cho and Hedrick [4] based on the initial work of Dobner [5]. Some of the proposed controller designs include sliding-mode fuel injection by Cho and Hedrick [6], multivariable optimal control by Hendricks et al. [7] and Han et al. [8], adaptive control by Beaumont et al. [9], and pneumatic and thermal state estimators by Maloney and Olin [10]. Recently, Pfeiffer and Hedrick [11] developed a multisurface sliding-mode controller to simultaneously regulate the fueling process and engine speed. In this project, a model-based engine controller has been developed to manage the air-to-fuel ratio and crankshaft speed using a backstepping problem formulation with attached nonlinear observer [12]. Specifically, the desired intake manifold air mass may be computed such that the engine speed tracks the target. Next, the throttle angle may be selected to force the difference between the desired and actual intake manifold air masses to zero such that the speed error approaches zero. It can be shown

Fig. 3. Parallel hybrid vehicle (rear wheel drive) with internal combustion spark ignition engine driving a CVT and alternator/generator.

that the observer estimated cylinder fuel flow rate does not require the oxygen sensors binary signal. What is the motivation for simultaneous engine speed and air-to-fuel ratio management? This question may be answered in the context of parallel hybrid vehicle powertrains (refer to Fig. 3), which may feature a small SI engine attached to a CVT and generator (e.g., [13][15]). For this paper, only the engine and transmission are considered. A CVT offers a continuum of infinitely variable gear ratios between established minimum/maximum limits to accommodate the vehicles needs with the available engine power in comparison to fixed gear ratio transmissions. Although basic CVT designs have difficulty with high torque/low speed requirements, hybrid configurations offer multiple power paths to satisfy the demand (e.g., [16] and [17]). Effective control of the gear ratio and power path clutches allows the designation of engine torque/speed operating ranges to optimize overall system efficiency for the given operating conditions. In Fig. 4(a), the engine performance map of a 1.0-L spark ignition engine coupled to an automatic transmission, following a prescribed drive cycle, has been generated using ADVISOR [18]. Note that the engine operates primarily in the 18.1% to 30% efficiency range although a maximum efficiency of 33% was available. In contrast, the same engine and driving cycle were explored with a CVT [refer to Fig. 4(b) for efficiency gains]. The number of operating points below 22.1% has been significantly reduced, but more importantly, the engine predominately operates in a narrower speed neighborhood. In other words, during a significant portion of the driving cycle, the engine speed was essentially constant while the torque output was

WAGNER et al.: NONLINEAR AIR-TO-FUEL RATIO AND ENGINE SPEED CONTROL FOR HYBRID VEHICLES

187

II. CONTROL PROBLEM STATEMENT Current emission control technology for spark ignition engines requires that the air-to-fuel ratio be approximately (i.e., stoichiometric) to ensure proper three-way catalytic converter operation [19]. A converters performance will deteriorate as deviates from this setpoint. The first control objective (i.e., 4% neighborhood of stoichiois to maintain metric). In addition to fuel control, the problem of engine speed regulation will be investigated. Thus, the second objective is to maintain the engines speed within a 2% neighborhood of the . speed setpoint,
(a)

III. MODEL-FREE FUEL AND ENGINE SPEED CONTROLS A speed-density fueling strategy with proportional-integral control will be presented along with a PID engine speed controller. These algorithms, acting independently, regulate the injector fuel flow rate and throttle angle to establish a basis for comparing possible performance enhancements realized with model-based nonlinear controllers. A. Speed Density With PI Control The ideal gas law [20] allows the manifold pressure to be expressed as
(b) Fig. 4. (a) Torque versus speed efficiency map for 1.0-L SI engine with automatic. (b) 1.0-L engine torque versus speed map for CVT with efficiency contours.

(1) where the air density in the intake manifold can be expressed as

modified using the CVT gear ratio. Once the CVT reached the minimum/maximum gear ratio boundary, the engine speed was varied to achieve the necessary speed or output torque (i.e., horizontal locus of operating points on the graph). Therefore, a hybrid vehicle powertrain configuration, featuring an SI engine and CVT, can meet small disturbances in the vehicles speed and torque demands through gear ratio adjustment while permitting the engine to operate within a prescribed speed neighborhood. Consequentially, a need exists to explore engine control strategies for simultaneous speed and air-to-fuel regulation in a modular architecture to permit application to emerging vehicle propulsion technologies. In this paper, a new model-based nonlinear control strategy will be presented to simultaneously regulate the engine speed and air-to-fuel ratio. The control contributions include a nonlinear backstepping controller, which tracks the desired sinusoidal engine speed trajectory via throttle angle adjustments, as well as a Lyapunov-based nonlinear controller, with attached observer to estimate the fuel flow rate entering the cylinder to maintain the air-to-fuel ratio at stoichiometric. Section II describes the control problem. Two model-free controllers, a speed-density PI algorithm and speed proportional-integral-derivative (PID) algorithm, are presented in Section III to provide insight into control algorithms. Section IV reviews a three-state engine model and applies it in the design of a nonlinear backstepping controller. For comparison purposes, a multisurface sliding-mode controller is also discussed. The performance of each controller is numerically studied in Section V. Conclusions are presented in Section VI.

(2) is a function of the air density The intake manifold air mass and the engines volumetric efficiency such that (3) Finally, the intake manifold mass air-flow rate becomes (4) where the engine speed is normalized by 120 to account for two crankshaft revolutions per cycle and 60 s/min. To compute , this equation the air mass drawn into the engine each cycle instead of must be changed to consider the engine volume . For the air flow per cylinder event, the manifold volume one must use the cylinder volume . The speed-density approach requires three sensors to estimate the air mass and compute the required fuel mass to satisfy stoichiometric operating conditions [21]. These sensors provide the , the engine speed , and the manifold absolute pressure inlet air temperature . A two-dimensional fuel lookup table can be constructed with engine speed and load (i.e., MAP) on the axes. The table entries are further modified to account for the inlet air temperature (i.e., as it effects the air density) and other effects in the fuel algorithm. The basis for the fuel table

188

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 52, NO. 1, JANUARY 2003

entries resides on the commanded fuel flow rate, which may be expressed for the engine as (5) where has been selected as the stoichiometric condition. To implement this strategy, a PI control algorithm [22] is often selected to compute the commanded fuel flow rate. In the ideal case, the error signal may be defined as (6) is the fuel flow rate (refer to Section IV-A). This exwhere pression implies that the air and fuel flow rates may be measured directly, which is not the case in production engines. However, oxygen sensors generate a binary voltage signal dependent on the oxygen concentration in the exhaust gas. In other words, a ) results in a sensor output of approxrich mixture (i.e., ) produces 0 V imately 0.8 V while a lean mixture (i.e., with a very narrow transition zone. For simulation purposes, the O2 sensors output will be modeled as SGN (7)

where represents the nominal throttle angle to be stored , , and in table lookup. The three controller gains have been determined through extensive numerical testing using an analytical engine simulation starting with the Zeigler and Nichols [24] tuning rules. IV. MODEL-BASED FUEL AND ENGINE SPEED CONTROLS The application of model-based control strategies to nonlinear dynamic automotive processes should provide improved performance during transient operation. Although current model-free algorithms achieve satisfactory emission levels using oxygen sensor feedback, the engines inherent transport delay poses a challenge during transients. In this section, a three-state engine model that describes the intake air, rotational dynamics, and fuel dynamics will be discussed. This model serves as the foundation for backstepping and sliding-mode controllers that simultaneously regulate the air-to-fuel ratio and the engine speed. A. Real-Time Engine Model To initiate the model-based controller design process, a real-time engine model must exist that describes the air and fuel flows. This reduced-order model, proposed by Moskwa and Hedrick [3] based on Dobners [5] initial work, contains , three states representing the intake manifold air mass engine speed , and fuel flow rate entering the combustion . The intake manifold air flow is equal to the air chamber flow entering the intake manifold minus the air flow leaving the intake manifold (11) The inlet air flow is a function of the throttle angle and normalized pressure ratio PRI where (12)

such that a high/low signal is realized when the mixture is lean/rich, respectively [11]. The oxygen sensor signal has a delay time , which may be represented as . The PI air-to-fuel ratio speed density control structure becomes

(8) The proportional and integral terms add/remove additional fuel when the exhaust mixture is lean/rich, as indicated by O2 and are scheduled sensor output. The controller gains into lookup tables for various engine operating conditions through extensive dynamometer and vehicle testing. The calibration process tends to be time consuming since the engines performance must be evaulated for each table entry. B. PID Engine Speed Control Most production engine management systems currently regulate the crankshaft speed at idle conditions using a servo-motor IAC circuit around the throttle valve. However, engine speed tracking (or torque management) may be necessary to facilitate vehicle platooning on the automated highway [23]. In this section, a PID controller will be presented to track a sinusoidal engine speed resulting in a commanded throttle angle. Let the engine speed error be expressed as (9) and are the setpoint and actual engine speeds, where respectively. Note that the engine speed may also be expressed equivalently in rpm. The PID controller structure becomes MAP (10)

is the normalized throttle characteristic and is the maximum air-flow rate. The normalized pressure ratio influence [25] may be expressed as PRI EXP (13)

The air flow leaving the intake manifold is (14) where and are functions of the intake manifold geometry (15) Substituting (12) and (14) into (11) yields the first differential equation PRI (16)

WAGNER et al.: NONLINEAR AIR-TO-FUEL RATIO AND ENGINE SPEED CONTROL FOR HYBRID VEHICLES

189

The second state of the model, engine speed, is developed by applying Newtons law to the mechanical dynamic system AFI (17)

is the load torque, which may be viewed as an external where is the torque constant, which relates the torque disturbance. produced to the exit mass air flow rate (i.e., engine treated as an air pump whose torque output is proportional to the flow). scales the torque value to Finally, the model parameter AFI account for the charge mixtures composition (e.g., rich or lean) on the available torque. The third model state describes the mass fuel flow rate enas a first-order lag based tering the combustion chamber on the ECU commanded fuel flow rate to account for the combined transport delay and lag [26] (18)

For the speed-tracking controller, the control input is the throttle angle such that the engine speed tracks the desired . However, the throttle cannot control the engine speed engine speed directly. In other words, the throttle controls the . engine speed through the air mass in the intake manifold Thus, the control process can be separated into two steps: first, such that tracks ; and second, design the design throttle angle such that the intake manifold air mass tracks . Define the two error signals and as (24a) (24b) Taking the time derivative of (24a) and substituting in (23) yields (25) which may be rewritten as

This expression may be rewritten as (19) The term represents the first-order time lag, which may deterwhere mined by the relationship (20) Substituting the results of (27) into (26) provides where B. Nonlinear Backstepping Controller With Nonlinear Observer The backstepping controller design problem begins with the restatement of the three-state engine model equations into a suitable mathematical form [27]. It is assumed that the manifold air mass and engine speed are measurable quantities. Furtherand AFI are assumed to more, the model parameters be constant over the prescribed operating range. Specifically, the engine speed fluctuation will be within 5% of the mean value. Thus, the volumetric efficiency should also fluctuate in a small range such that it can be approximated as a constant for a given inlet temperature. Similarly, the air-to-fuel ratio must be maintained within 4% of stoichiometric. Therefore, changes in the should be small, and it too air-to-fuel ratio influence AFI will be approximated as a constant. The mass air-flow rate in the intake manifold (16) may be represented as (21) and PRI . The where intake manifold exit air-flow rate (14) may be substituted into the engine speed (17) such that AFI or define AFI to express (22) as (23) AFI (22) and may be selected as

(26)

(27)

(28)

(i.e., tracks ), the time derivative To ensure that of (24b) may be substituted into (21) so that (29) Design the control input throttle angle to allow ; thus and then (30) and has been added to cancel in (28). where This expression may be simplified to realize the controller input for as (31) Substituting the results of (31) into (29) provides (32) The stability may be checked with Lyapunovs stability criteria (33a)

(33b) to where continuously decreases from the initial value zero such that and both go to zero when the system reaches steady state (i.e., speed tracking is realized).

190

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 52, NO. 1, JANUARY 2003

The next task is designing the air-to-fuel ratio controller. In this paper, an observer will be implemented to estimate the fuel . Define , which flow rate entering the cylinder allows (19) to be expressed as (34) Next, define the air-to-fuel ratio error [refer to (6)] as (35) which may be expanded using (14) as (36) An observer may be designed to estimate the value of as

to, and maintain it on, a sliding surface [28], [29]. The sliding surface is selected such that a system restricted to this manifold behaves as an equivalent linear system. In general, VSC controllers are robust to external system disturbances and unmodeled system dynamics. The engine speed controller design begins with (14) and (17) so that AFI The sliding surfaces and may be selected as (46a) (46b) The time derivative of (46a) may be substituted in (45) to realize (45)

(37) AFI where . Subtracting (37) from (34) provides which may be expressed using (46b) as (38) AFI Redefine the air-to-fuel ratio error in (36) as (39) Finally, take the time derivative of (39) and, substituting in (16) and (17) AFI (48) (47)

The first design task is to select the desired air mass such that the engine speed tracks the desired engine speed. Let the desired air mass be chosen as (49)

(40) For mathematical notation convenience, define the term

so that (48) may be written as AFI Substituting (16) into the time derivative of (46b) provides (50)

(41) such that (40) becomes (42) Therefore, the control input becomes

PRI The throttle angle control signal with (51) and selecting as

(51) may now be designed starting

PRI where (43) AFI and The stability of the two sliding surfaces demonstrated using Lyaponovs stability criteria and where and (52) may be

. The stability of the controller may again such that be demonstrated using Lyapunov and (44)

(53)

such that continuously decreases from its initial value to zero and ) so (i.e., ). ( C. Sliding-Mode Fuel-Injection Control A variable structure system controller (VSC) is designed next to regulate the air-to-fuel ratio about the stoichiometric point and track an engine speed profile. Sliding-mode controllers alter the nonlinear system behavior through the high-speed discontinuous switching of the control inputs to force the state trajectory

may be substituted from (50) and (52) to yield (54)

which guarantees a stable speed tracking controller. The air-to-fuel ratio controller design starts with (19), which describes the mass fuel flow rate entering the combustion chamber. Define the sliding surface as (55)

WAGNER et al.: NONLINEAR AIR-TO-FUEL RATIO AND ENGINE SPEED CONTROL FOR HYBRID VEHICLES

191

whose time derivative becomes (56) The air flow leaving the intake manifold derivative are of (14) and its time

TABLE I LISTING OF THE PARAMETERS FOR THE MODEL-FREE AND MODEL-BASED CONTROLLERS

(57) Substituting (17) into (57) provides V. NUMERICAL RESULTS A suite of subsystems was created in Matlab/Simulink 1 to implement the three engine control strategies. The controller parameters are summarized in Table I. These subsystems were then integrated into an existing analytical and empirical engine model [30], which describes the throttle body, intake manifold, combustion process, and rotational dynamics for an internal combustion spark ignition utility engine. An integras and fourth-order RungaKutta tion time-step of integration routine were selected. The average engine speed rpm with a sinusoidal disturbance of shall be rpm and frequency Hz such amplitude . Two load torque cases were that examined. Case one denotes a constant load torque, and case s. Fitwo explores a 33% torque step increase for nally, the air-to-fuel ratio shall be maintained at stoichiometric. The engine speed and air-to-fuel ratio results for the model-free PID and PI controllers are presented in Fig. 5 for both cases. For convenience, the engine simulation was executed for 2 s prior to recording the numerical results; this permitted the initial simulation transients to decay to steady-state operation. For case one [refer to Fig. 5(a)], it may be observed that the engine speed tracks the desired trajectory ) and the air-to-fuel ratio is within 1.0% (i.e., ) neighborhood of regulated within an 8.0% (i.e., stoichiometric. The numerical results displayed in Fig. 5(b) for case 2 show that the engine speed deviates from the desired value by 1.8% between 6 and 8 s (i.e., load torque disturbance time interval). Similarly, the air-to-fuel ratio has sharp depar(i.e., 29%) and (i.e., 15%) s from the tures at stoichiometric value. The results for the model-based backstepping controller, with an attached observer, are shown in Fig. 6. In case one [refer to Fig. 6(a)], the engine speed tracks the setpoint trajectory within 0.4%, offering performance improvements over the model-free controller. Similarly, the air-to-fuel ratio is maintained within a 4.0% neighborhood of the stoichiometric operating point. This is a significant improvement in comparison to the model-free controller. For case two, shown in Fig. 6(b), the engine speed deviates by 0.9% from the desired operating trajectory during the disturbance time interval, while the air-to-fuel ratio does not exhibit large departures at the transition points. However, the air-to-fuel ratio does contain high-frequency oscillations (i.e., chattering) about the nominal sinusoidal response. In other words, the magnitude of deviation from the stoichiometric value is similar to case one, but the oscillations are greater due
1Registered

(58) Noting that , we can express (56) as (59) which may be rewritten using (55) as (60) Again, the control design should select attractive sliding surface such that is an

(61) In other words, the command fuel flow rate becomes (62) The research conducted by Cho and Hedrick [6] has examined model uncertainty in the fueling process. For this situation, and fuel process the air flow exiting the intake manifold first-order time lag may be expressed as (63) (64) and represent model imprecision and dewhere scribes the best model estimated value. Let these time-varying imprecisions have upper bounds and (65) Finally, the oxygen sensor delay must be included in (62) for the commanded fuel flow rate such that (66) where is the oxygen sensor output and becomes controller gain the time delay. The

(67)

trademark of The Mathworks, Inc., Natick, MA.

192

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 52, NO. 1, JANUARY 2003

(a)

(b) Fig. 5. (a) Engine speed PID and air-to-fuel ratio PI controller results for constant load torque (case one) with air-to-fuel ratio tracking within an 8.0% neighborhood of stoichiometric. (b) Engine speed PID and air-to-fuel ratio PI controller results for a load torque disturbance (case two) with air-to-fuel ratio departures at the transition points.

(a)

(b) Fig. 6. (a) Backstepping controller results for constant load torque (case one); the air-to-fuel ratio tracks within a 4.0% neighborhood of stoichiometric value. (b) Backstepping controller engine speed and air-to-fuel ratio results for a load torque disturbance between 6 and 8 s (case two).

WAGNER et al.: NONLINEAR AIR-TO-FUEL RATIO AND ENGINE SPEED CONTROL FOR HYBRID VEHICLES

193

(a)

(b) Fig. 7. (a) Sliding-mode controller results for constant load torque (case one) with an air-to-fuel ratio within 6.0% stoichiometric value. (b) Sliding-mode controller results for load torque disturbance (case two) with engine speed (rpm) and air-to-fuel ratio versus time (seconds). TABLE II SUMMARY OF ENGINE CONTROLLER PERFORMANCE

to the disturbance rejection attributes of the controller. Overall, the backstepping controller demonstrates good robustness to external disturbances and signal noise. However, in the presence of model uncertainty and/or parameter variations, the controller performance will deteriorate since the underlying model no longer corresponds to the actual plant. Consequentially, an adaptive controller may need to be designed to compensate for system uncertainty through the online estimation of the uncertain parameters. Finally, the sliding-mode controller results are presented in Fig. 7. For case one [refer to Fig. 7(a)], the engine speed tracks the given trajectory within 0.7% while the air-to-fuel ratio is regulated within a 6.0% neighborhood of stoichiometric. This performance is secondary to that exhibited by the backstepping controller; however, it is also an improvement over the model-free controller. In case two, the sliding-mode controller demonstrates superior performance, as evidenced by the results

in Fig. 7(b). The engine speed does not significantly deviate . The air-to-fuel ratio has (i.e., 0.9%) from desired speed increased chattering during the disturbance time interval; however, the magnitude of deviation from stoichiometric is approximately the same as case one. With a load disturbance, the sliding-mode controller performance deteriorated 2.94% and 0.62% for the summed air-to-fuel ratio and engine speed and per Table II, respectively. In conerrors trast, the backstepping controller displayed error magnitudes of 13.04% and 1.68% for these same metrics. The disturbance rejection qualities for the sliding-mode controller are superior to the backstepping controller. The performance of each controller for cases one and two is summarized in Table II. In columns two through four, the average air-to-fuel ratio , standard deviation , and integral of , where , absolute air-to-fuel ratio error are presented. The last two table columns denote the average

194

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 52, NO. 1, JANUARY 2003

engine speed and the integral of the absolute engine speed , where . For case one, the error first controller (i.e., model-free PI/PID) provided satisfactory performance when the system parameter magnitude changes were maintained within a neighborhood of the stoichiometric value. The model-based sliding-mode controller exhibited approximately a 25% performance improvement (i.e., examination of the air-to-fuel ratio and speed tracking for case one) in comparison to the model-free controller. Finally, the backstepping controller maintained outstanding speed tracking and air-to-fuel ratio control with approximately a 50% improvement (i.e., comparison of case ones speed tracking and air-to-fuel ratio) over the lookup table controller. Overall, the backstepping controller demonstrated the best performance among the three controller designs examined in this paper. For case two, it may be observed from Figs. 57 that the disturbance rejection characteristics of the model-free controller had longer recovery times (i.e., approximately 0.3 s). In contrast, the sliding mode and the backstepping controllers exhibited better robustness when subjected to a load torque disturbance. Although each controller satisfied the speed requirement in Section II, only the backstepping controller met the air-to-fuel ratio control objective on a transient basis.

VI. CONCLUSION Three different control strategies have been presented to regulate the speed and air-fuel ratio in internal combustion spark ignition engines. These controllers range from model-free, or lookup table, controllers using the classical speed-density approach to model-based methods using nonlinear control theory. Based on the numerical simulation results presented in the previous section, the backstepping controller provided the best overall performance for speed tracking and air-to-fuel ratio control followed by the sliding mode and the model-free controllers. It is interesting to note that the backstepping controller used an attached observer to avoid the inclusion of oxygen sensor feedback and its time delay. For system disturbances (i.e., load torque fluctuations), the sliding-mode controller exhibited slightly better performance than the backstepping controller. In summary, the model-based control strategies demonstrated attractive operating properties and merit further consideration for hybrid vehicle engine control.

[7] E. Hendricks, T. Vesterholm, and S. Sorenson, Nonlinear, closed-loop, SI engine control observers, SAE paper 920 237, 1992. [8] M. Han, R. Loh, L. Wang, A. Lee, and D. Stander, Optimal idle speed control of an automotive engine, SAE paper 981 059, 1998. [9] A. J. Beaumont, A. D. Noble, and A. Scarisbrick, Adaptive transient air-fuel ratio control to minimize gasoline engine emissions, in Proc. FISITA Congr., London, U.K., 1992. [10] P. J. Maloney and P. M. Olin, Pneumatic and thermal state estimators for production engine control and diagnostics, SAE paper 980 517, 1998. [11] J. M. Pfeiffer and J. K. Hedrick, Nonlinear algorithms for simultaneous speed tracking and air-fuel ratio control in an automobile engine, SAE paper 99010547, 1999. [12] J. Wagner, D. Dawson, and Z. Liu, Nonlinear air-to-fuel ratio and engine speed control for spark ignition engines, in Proc. ASME Dynamic Systems and Control Division, Orlando, FL, 2000. [13] N. Iwai, Analysis on fuel economy and advanced systems of hybrid vehicles, JSAE Rev., vol. 20, no. 1, pp. 311, 1999. [14] S. Cikanek, K. Bailey, and B. Powell, Parallel electric hybrid vehicle dynamic model and powertrain control, in Proc. Amer. Control Conf., vol. 1, 1997, pp. 684688. [15] A. Corbett and C. Mellors, Hybrid electric machines, Inst. Elect. Eng. Colloq., no. 152, June 1996. [16] V. H. Mucino, J. E. Smith, B. Cowan, and M. Kmicikiewicz, A continuously variable power split transmission for automotive applications, SAE paper 970 687, 1997. [17] S. Hanachi, A study of the dynamics of a split-torque, geared-neutral transmission mechanism, J. Mech. Design, vol. 112, pp. 261270, Sept. 1990. [18] ADVISOR, National Renewable Energy Laboratory, Golden, CO, 2.2.1 ed., 1999. [19] C. D. Falk and J. J. Mooney, Three way conversion catalysts: Effect of closed-loop feedback control and other parameters on catalyst efficiency, SAE paper 800 462, 1980. [20] C. F. Taylor, The Internal Combustion Engine in Theory and Practice, Volume 1: Thermodynamics, Fluid Flow, Performance. Cambridge, MA: MIT Press, 1977. [21] E. Chowanietz, Automobile Electronics. London, U.K.: ButterworthHeinemann, 1995. [22] G. F. Franklin, J. D. Powell, and A. Emami-Naeini, Feedback Control of Dynamic Systems. Reading, MA: Addison-Wesley, 1994. [23] J. K. Hedrick, M. Tomizuka, and P. Varaiya, Control issues in automated highway systems, IEEE Contr. Syst. Mag., vol. 14, no. 6, pp. 2132, 1994. [24] J. G. Zeigler and N. B. Nichols, Optimum settings for automatic controllers, Trans. ASME, vol. 64, pp. 759768, 1942. [25] J. Heywood, Internal Combustion Engine Fundamentals. New York: McGraw-Hill, 1988. [26] D. Cho and J. K. Hedrick, A nonlinear controller design method for fuel-injected automotive engines, ASME J. Eng. Gas Turbines Power, vol. 110, no. 2, pp. 313320, 1988. [27] J. J. E. Slotine and W. Li, Applied Nonlinear Control. Englewood Cliffs, NJ: Prentice-Hall, 1991. [28] J. J. E. Slotine and S. S. Sastry, Tracking control on nonlinear systems using sliding surfaces with application to robot manipulators, Int. J. Contr., vol. 38, no. 2, pp. 465492, 1983. [29] V. I. Utkin, Sliding Modes and Their Application in Variable Structure Systems. Moscow, Russia: MIR, 1978. [30] J. R. Wagner and W. Cheek, Model development for spark ignition utility engines to support control algorithm designs, in Proc. ASME Dynamic Systems and Control Division, Nashville, TN, 1999, pp. 167178.

REFERENCES
[1] Bosch, Automotive Electric/Electronic Systems. Wallendale, PA: Society of Automotive Engineers, 1994. [2] E. Hendricks, M. Jensen, P. Kaidantzis, P. Rasmussen, and T. Vesterholm, Transient A/F ratio errors in conventional SI engine controllers, SAE paper 930 856, 1993. [3] J. J. Moskwa and J. K. Hedrick, Automotive engine modeling for real time control applications, in Proc. Amer. Controls Conf., June 1987. [4] D. Cho and J. K. Hedrick, Automotive powertrain modeling for control, J. Dyn. Syst., Meas., Contr., vol. 111, no. 4, pp. 568576, 1989. [5] D. J. Dobner, A mathematical engine model for development of dynamic engine control, SAE paper 800 054, 1980. [6] D. Cho and J. K. Hedrick, Sliding mode fuel-controller: Advantages, J. Dyn. Syst., Meas., Contr., vol. 113, no. 3, pp. 537541, 1991.

John R. Wagner (M00) received the B.S., M.S., and Ph.D. degrees in mechanical engineering from the State University of New York at Buffalo and Purdue University, West Lafayette, IN. He joined the Department of Mechanical Engineering, Clemson University, Clemson, SC, in 1998. He was previously on the Engineering Staff at Delphi Delco Electronics Systems working on automotive control systems from 1989 to 1998. His research interests include nonlinear and intelligent control systems and system health monitoring strategies with application to automotive and mechatronic systems. He is the Leader of the Automotive Research Laboratory at Clemson University and serves as the SAE faculty advisor.

WAGNER et al.: NONLINEAR AIR-TO-FUEL RATIO AND ENGINE SPEED CONTROL FOR HYBRID VEHICLES

195

Darren M. Dawson (S89M90SM94) received the B.S. and Ph.D. degrees in electrical engineering from Georgia Institute of Technology, Atlanta, in 1984 and 1990. He is a Professor in electrical and computer engineering at Clemson University, Clemson, SC. From 1985 to 1987, he was with Westinghouse as a Control Engineer. He leads the Robotics and Manufacturing Automation Laboratory, which is jointly operated by the Electrical and Mechanical Engineering Departments. His areas of interest include nonlinear-based robust and adaptive and learning control with application to electromechanical systems.

Liu Zeyu is a Mechanical Engineering Graduate Research Assistant in the Department of Mechanical Engineering, Clemson University, Clemson, SC. He is investigating nonlinear model reduction strategies and model-based controller designs for hybrid vehicle spark ignition engines.

S-ar putea să vă placă și