System and method of controlling air-charge in direct injection lean-burn engines

[0002] This invention relates to engine control systems and, more particularly, to a system and method of controlling air-charge in lean-burn, spark-ignited engines that compensates for uncertainties in the intake and exhaust system including EGR valve flow area and volumetric efficiency.

[0003] The objective of the air-charge control in lean-burn, spark-ignited engines is to operate the electronic throttle and the EGR valve in a manner so as to provide the desired gas flow to the engine and the desired burnt gas fraction in this flow for NO_xreduction. The conventional approach to the air-charge control is open-loop whereby the desired mass flow rates through the EGR valve and throttle are calculated as functions of the desired burnt gas fraction in the intake manifold and of the desired gas flow into the cylinder. The desired values of the EGR valve position and of the throttle position are backtracked using an orifice equation with known effective flow area map. Suppose F_1,dis the desired burn gas fraction, W_cyl,dis the desired gas flow rate into the cylinder. The desired mass flow rate through the throttle and the desired mass flow rate through the EGR valve can be calculated as follows:

[0004] W_th,d=(W_cyl,d−W_egr,d),

[0005] W_egr,d=(F_1,d(W_cyl,d+W_f)W_cyl,d/W_f/(1+λ_s)),

[0006] where λ_sis the stoichiometric air-to-fuel ratio and W_fis the engine fueling rate. The desired throttle position, u_th, is determined from the intake manifold pressure p₁, ambient pressure, p_amb, and ambient temperature, T_amb, while the estimate of exhaust manifold temperature T₂and exhaust manifold pressure p₂as well as a measurement of p₁are used to calculate the EGR valve position u_egr. Typically, the values of F_1,d, W_cyl,dare functions of engine speed, engine torque, combustion mode, etc., determined by calibration tables that are optimized for fuel economy and emissions.

[0007] There are several problems with the conventional approach. First, the soot deposits in the EGR valve and pipes change the effective flow area of the valve. This problem is particularly severe with lean-burn, direct-injection, spark-ignition engines. Second, depending on the EGR valve and throttle type, calibration drifts may render their actual position uncertain. Due to these uncertainties, the desired mass flow rates may not be achieved with the open-loop approach; and the emission performance of the engine at a given engine speed and engine load point may be shifted away from the desired nominal performance. Even without the calibration drift, when the pressure drop across the throttle or the EGR valve is small, the open-loop procedure may lead to excessive chattering of the desired throttle position or EGR valve position due to noise in the pressure measurements. Other sources of uncertainty, such as changes in the engine back pressure due to the exhaust tract clogging, may also affect the engine operation in substantial ways.

[0008] More specifically, consider FIGS. 1-3. The dashed lines in these figures correspond to the steady-state values of the NO_xmass flow rate, engine exhaust temperature and engine torque achieved by the conventional open-loop controller as a function of the unknown multiplier, θ_egr, on EGR valve effective flow areas. The value of θ_egr=1 correspond to the nominal case. Constant values of engine speed N=2000 rpm, fueling rate W_f=2 kg/hr, spark timing δ=25 deg BTDC and constant desired values of the in-cylinder flow W_cyl,d=80 kg/hr and burnt gas fraction F_1,d=0.1 were used in the engine model simulation. With 40 percent of EGR valve flow area reduction, there is a significant increase in NO_xemissions by 50 percent (see FIG. 1). FIGS. 2 and 3 demonstrate that the engine exhaust temperature and the engine brake torque are affected by changes in θ_egr. The variations in θ_egrtranslate into variations in the burnt gas fraction delivered to the engine and, therefore, engine combustion variations.

[0009] In accordance with the present invention, a method is disclosed that produces robust engine operation in spite of these uncertainties. The method relies on adding feedback to the open-loop scheme to compensate for uncertainties in the EGR valve flow area and provides an adaptation algorithm to improve the estimate of in-cylinder flow and throttle open-loop control. The solid lines in FIGS. 1-3 that show much reduced, almost non-existent sensitivity of the engine emissions, exhaust temperature and engine torque to changes in θ_egrcorrespond to one of the embodiments that will be described hereinafter. Some of the embodiments use an estimate of the flow through the EGR valve that is derived from an intake manifold pressure measurement.

[0010] FIGS. 1-3 show comparisons between a prior art system configuration (dashed), and the system of the present invention (solid) with respect to feedgas NO_xemissions sensitivity, feedgas temperature sensitivity, and engine torque sensitivity;

[0011] FIG. 4 shows a block diagram of the system of the present invention;

[0012] FIG. 5 shows time histories of the intake manifold pressure and EGR valve flow rate (solid lines);

[0013] FIG. 6 shows time histories of throttle position and EGR valve position (solid lines);

[0014] FIG. 7 shows some simulation results for the least squares adaptive algorithm; and

[0015] FIG. 8 shows a flowchart of the method of the present invention.

[0016] Referring again to the drawing and initially to FIG. 4, a block diagram of the control system of the present invention is shown. The system comprises an electronic engine controller generally designated 10, that includes ROM, RAM and CPU as indicated. The controller 10 controls a set of injectors 12, 14, 16 and 18 which inject fuel into a four-cylinder, spark ignited, lean-burn gasoline engine 20. The fuel is supplied by a high pressure fuel system (not shown), and is injected directly into the combustion chambers in precise quantities and timing as determined by the controller 10. Air enters the combustion chambers through the intake manifold 22, and combustion gases are exhausted through the exhaust manifold 24. An electronically controlled throttle valve 26 is positioned by the controller 10 to control the air mass flow into the intake manifold 22. An airmeter or air mass flow sensor 28 is positioned upstream of the valve 26 and provides a signal to the controller 10 that calculates a value indicative of the mass of air flowing into the induction system. The controller 10 transmits a fuel injector signal to the injectors to produce a desired engine torque and maintain a desired air/fuel ratio.

[0017] To reduce the level of NO_xemissions, the engine is equipped with an exhaust gas recirculation (EGR) system 30. The EGR system 30 comprises a conduit 32 connecting the exhaust manifold 24 to the intake manifold 22. This allows a portion of the exhaust gases to be circulated from the exhaust manifold 24 to the intake manifold 22 in the direction of the arrow. An EGR valve 34, controlled by the controller 10, regulates the amount of exhaust gas recirculated from the exhaust manifold. In the combustion chambers, the recirculated exhaust gas acts as an inert gas, thus lowering the flame and in-cylinder gas temperature and decreasing the formation of NO_x. On the other hand, the recirculated exhaust gas displaces fresh air and reduces the air-to-fuel ratio of the in-cylinder mixture.

[0018] In the controller 10, command signals to the actuators for positioning the EGR valve 34 and the throttle valve 26 are calculated from measured variables and engine operating parameters by means of control algorithms. Sensors and calibratable lookup tables provide the controller 10 with engine operating information. For example, MAP sensor 36 provides a signal to the controller 10 indicative of the pressure in the intake manifold 24. A temperature sensor 38 provides a signal to the controller 10 indicative of the temperature of the gas in the intake manifold. The sensors 36 and 38 may be combined if desired. Additional sensory inputs can also be received by the controller 10 such as engine coolant temperature, engine speed, throttle position, and ambient temperature and barometric pressure. Based on the sensory inputs and engine mapping data stored in memory, the controller controls the EGR and throttle valves to regulate the intake airflow.

[0019] A heated exhaust gas oxygen (HEGO) sensor 40 or universal exhaust gas oxygen (UEGO) sensor detects the oxygen content of the exhaust gas generated by the engine, and transmits a signal to the controller 10. Sensor 40 is used for control of the engine A/F, especially during any stoichiometric operation. An exhaust system, comprising one or more exhaust pipes, transports exhaust gas produced from combustion of an air/fuel mixture in the engine to a conventional close-coupled, three-way catalytic converter (TWC) 42. The converter 42 contains a catalyst material that chemically alters exhaust gas that is produced by the engine to generate a catalyzed exhaust gas. The catalyzed exhaust gas is fed through an exhaust pipe 44 to a downstream NO_xtrap 46 and thence to the atmosphere through a tailpipe 48.

[0020] As is usually done, the control and adaptation algorithms discussed hereinafter are specified in continuous time. It is understood, however, that for the actual implementation the algorithms are discretized assuming either a time-synchronous or a crank-synchronous sampling procedure, with compensation of the delay introduced by sampling (if necessary). The signals used by the algorithms are filtered versions of the measured signals. The filters remove the periodic oscillations and noise in the signals so that the mean values of the signals are available for control and adaptation. First or second order linear filters or crank-synchronous sampling are standard ways of deriving the mean value from a signal. Hereinafter, all the signals are identified and referenced as their mean values. The controllers that use the integral action, such as proportional-plus-integral (PI) controllers, use antiwindup compensation. It is introduced to handle control signal saturation due to actuator limits.

[0021] A relatively simple approach to reduce sensitivity to uncertainties in the EGR path (EGR valve flow area, exhaust pressure and temperature) is to add feedback on the deviation of the intake manifold pressure, p₁, from the desired pressure or set-point, p_1,d,

[0022] u_egr=u_egr,d−k_p(p_1,d−p₁)−k_i∫₀^t(p_1,d−p₁(σ))dσ,

[0023] u_th=u_th,d,

[0024] where u_egr,dis the desired EGR valve position (a function of engine speed and engine demanded torque) and u_th,dis the desired throttle position (also function of engine speed and engine demanded torque). The controller adjusts the EGR valve position (due to the integral term) to correct for the uncertainty in the EGR valve flow area. The set point, p_l,d, can be derived from W_cyl,dby inverting the “speed-density” equation (see Eq. 1 below) or can be specified directly as a function of engine speed, demanded engine torque, combustion mode and other variables.

[0025] Where a mass airflow sensor, such as MAF sensor 28 in FIG. 4, is available to measure the flow rate through the throttle, an improved design is possible. First, an estimate of the mass flow rate through the EGR valve is made using the measurements of the intake manifold pressure, p₁, mass flow rate through the throttle, W_th, and an estimate of the total in-cylinder flow Ŵ_cyl. The total mass flow rate into the cylinder can be estimated from the speed-density equation and is a function of engine speed (N), intake manifold pressure (p₁) and estimated or measured intake manifold temperature (T₁), e.g., $\begin{array}{cc}{\hat{W}}_{\mathrm{cyl}}=k_0\left(N\right)\frac{p_1}{T_1}N,& \left(1\right)\end{array}$

[0026] where k₀is a calibratable function. The estimator for the EGR flow rate is described by the following equations: $\begin{array}{cc}\stackrel{.}{ϵ}=-{\alpha }_0ϵ+{\alpha }_0\left(-{\hat{W}}_{\mathrm{cyl}}+W_{\mathrm{th}}\right)+\frac{{\alpha }_o^2}{c_m}{p}_1,& \left(2\right)\\ {\hat{W}}_{\mathrm{egr}}=\frac{{\alpha }_0}{c_m}{p}_1-ϵ,& \left(3\right)\end{array}$

[0027] where α₀>0 is a calibratable parameter and $c_m=\frac{{\mathrm{RT}}_1}{V_1},$

[0028] where T₁is the intake manifold temperature and V₁is the intake manifold volume. It should be noted that the estimate Ŵ_egrdoes not require knowledge of the exhaust pressure or exhaust temperature or of the EGR valve effective flow area. Furthermore, the precise knowledge of the value of c_mis not needed for steady-state accuracy but is desirable in transients. Feedback is then imposed on the deviation of the estimate of the EGR flow rate, Ŵ_egr, from the set point, W_egr,d.

[0029] A decentralized proportional plus integral controller takes the form:

[0030] u_th=u_th,d−k_p1(p_1,d−p₁)−k_i1∫₀^t(p_1,d−p₁(σ))dσ,

[0031] u_egr=u_egr,d−k_p2(W_egr,d−Ŵ_egr)−k_i2∫₀^t(W_egr,d−Ŵ_egr(σ))dσ,

[0032] where k_p1, k_p2, k_i1, k_i2are the controller gains. The responses shown in FIGS. 1-3 by solid lines and that indicate reduced sensitivity to EGR valve clogging are due to this controller. The above form for the controller is not the only one possible. A nonlinear proportional plus integral controller can be designed using the speed-gradient approach where an attempt is made to maximize the rate of decay of a cost, “energy-like” function: $Q=\frac{1}{2}{\left({\gamma }_1\left(p_1-p_{1,d}\right)\right)}^2+\frac{1}{2}{\left({\gamma }_2\left({\hat{W}}_{\mathrm{egr}}-W_{\mathrm{egr},d}\right)\right)}^2.$

[0033] The objective is to make Q decay as rapidly as possible to zero, thereby resulting in p₁tracking p_1,dand Ŵ_egrtracking W_egr,d. Specifically, we can define

[0034] φ(p_1,d−p₁,p₁)=γ₁(p₁−p_1,d){tilde over (W)}_th,

[0035] Ψ(p_1,d−p₁,p_1,W_egr,d−Ŵ_egr)=γ₁(p₁−p_1,d){tilde over (W)}_egr+α₀γ₂{tilde over (W)}_egr[Ŵ_egr−W_egr,d].

[0036] Here {tilde over (W)}_this a function of p₁, {tilde over (W)}_egris a function of p₁,p₂and T₂so that

[0037] W_th={tilde over (W)}_thu_th, W_egr={tilde over (W)}_egru_egrθ_egr,

[0038] where θ_egris the unknown parameter. Note that errors in estimating {tilde over (W)}_egrcan be tolerated by the proposed control design (similar to unknown θ_egrvalue). The controller takes the form,

[0039] u_th=u_th,d−k_p1φ−k_i1∫₀^tφdσ,

[0040] u_egr=u_egr,d−k_p2Ψ−k_i2∫₀^tΨdσ.

[0041] FIGS. 5-6 illustrate the transient responses of this controller when the unknown restriction of the EGR valve flow area corresponds to θ_egr=0.8. In FIG. 5, the estimate of EGR valve flow rate is shown by the dashed line, and the setpoints by the dash-dotted line. The controller is turned on at t=2 sec. In FIG. 6, the feedforward values of throttle position and EGR valve position (open loop commands) are shown by the dash-dotted lines. The actual EGR valve position differs from the open-loop feed-forward command in steady-state because of the compensatory action of the feedback controller. It should be noted that the estimator for W_egris able to keep up with only relatively slow variations in W_egr. While (in simulations) it is possible to Increase the estimator bandwidth by increasing the value of α₀, a “slow” estimator is also acceptable since tracking of the set points is required only in steady state. In reality, the bandwidth of the estimator would always be limited by the sampling rate. The controller achieves the commanded values of the intake manifold pressure and of the EGR valve flow rate despite the uncertain EGR valve flow area.

[0042] The feedback controllers discussed above are sensitive to errors in estimating W_cyl. Reduction in in-cylinder flow rate can occur due to deposits as the engine ages. The effect of not estimating W_cylcorrectly on the operation of the controllers specified is that W_thwill still be regulated to W_th,d(derived from W_cyl,d), but W_egrwill be regulated to W_egr,d+(W_cyl−Ŵ_cyl) where Ŵ_cylis an estimate of W_cylcalculated on the basis of the “speed-density” equation. This will lead to errors in tracking W_egr,dand W_cyl,d. An adaptive control algorithm is introduced to adjust the estimate of W_cylas the engine ages.

[0043] When the in-cylinder flow rate is uncertain, equation (2) and (3) cannot reliably estimate the EGR flow rate. The orifice equation cannot be used for EGR estimations either because the estimates of exhaust manifold pressure and temperature (that rely on the estimate of the in-cylinder flow rate) are also uncertain. Hence, to adapt the in-cylinder flow rate equation (1), EGR should be disabled by closing the EGR valve or waiting for engine operating conditions (such as acceleration) when the EGR valve is closed.

[0044] In this case,

[0045] {dot over (p)}₁=c_m(W_th−W_cyl),  (4)

[0046] and it is assumed that

[0047] W_cyl=θ_cyl{tilde over (W)}_cyl,

[0048] where θ_cylis an unknown multiplier. The current estimate of θ_cylis {circumflex over (θ)}_cylso that an estimate of W_cylis given by:

[0049] Ŵ_cyl={circumflex over (θ)}_cyl{tilde over (W)}_cyl.

[0050] An alternative estimate of W_cylis also obtained from the time rate of change of the intake manifold pressure, and {circumflex over (θ)}_cylis adapted based on the difference of the two estimates. This can be done in a number of different ways and one algorithm will now be described. On-line parameter estimation can be done using a variety of the prediction-error-based methods, for example, the Least-Squares (LS) algorithm. Filtering both sides of the system equation through a first order filter one obtains

[0051] Y(t)=φ(t)θ_cyl,

[0052] where $\begin{array}{cc}Y\left(t\right)=c_m\left(\frac{1}{s/\tau +1}{W}_{\mathrm{th}}\right)-\frac{s}{s/\tau +1}{p}_1,& \left(5\right)\\ \phi \left(t\right)=\frac{c_m}{s/\tau +1}{\tilde{W}}_{\mathrm{cyl}}.& \left(6\right)\end{array}$

[0053] Here the standard notation for the filter in terms of its transfer function is used and τ is the time constant of the first order filter. Then the LS estimator with forgetting factor looks as follows:

[0054] {circumflex over ({dot over (θ)})}_cyl=−P(t)φ(t)(φ(t){circumflex over (θ)}_cyl−Y(t)  (7)

[0055] $\frac{d}{dt}\left(P^{-1}\right)=-\lambda \left(t\right)P^{-1}+{\phi }^2\left(t\right),\mathrm{where}$ $\lambda \left(t\right)={\lambda }_0\left(1-\frac{\Vert P\Vert }{k_0}\right),{\lambda }_0,k_0>0\Vert P\left(0\right)\Vert \le k_0.$

[0056] This estimator guarantees the exponential decay of the parameter error {tilde over (θ)}_cyl={circumflex over (θ)}_cyl−θ_cyldue to the persistence of excitation conditions holding generically ({tilde over (W)}_cyl≠0).

[0057] A simpler scheme is the projection algorithm which in discrete time takes the form ${\hat{\theta }}_{\mathrm{cyl}}\left(t\right)={\hat{\theta }}_{\mathrm{cyl}}\left(t-1\right)+\frac{\mathrm{\gamma \phi }\left(t\right)}{\alpha +{\phi }^2\left(t\right)}\left(Y\left(t\right)-\phi \left(t\right){\hat{\theta }}_{\mathrm{cyl}}\left(t-1\right)\right),$

[0058] where α≧0 and 0<γ<2.

[0059] An alternative scheme which relies on the difference between the estimated and measured intake manifold pressure can also be used for the W_cyladaptation. Based on equation (4), one can generate the estimated intake manifold pressure based on the estimated in-cylinder flow as:

[0060] {circumflex over ({dot over (p)})}₁=c_m(W_th−{circumflex over (θ)}_cyl{tilde over (W)}_cyl)

[0061] The difference between the measured intake pressure p₁and the estimated {circumflex over (p)}₁is due to error in the in cylinder flow estimation and can be used to adjust the parameter {circumflex over (θ)}_cyl. A simple form of the projection algorithm in the discrete form is: ${\hat{\theta }}_{\mathrm{cyl}}\left(t\right)={\hat{\theta }}_{\mathrm{cyl}}\left(t-1\right)+\frac{\mathrm{\gamma \phi }\left(t\right)}{\alpha +{\phi }^2\left(t\right)}\left(p_1-{\hat{p}}_1\right),$

[0062] For the implementation of these adaptive algorithms disclosed here, a deadband is also applied to assure robustness of the adaptation scheme in the presence of other uncertainties such as measurement noise. For example, the adaptation to the in cylinder flow equation is only activated when the error between the measured and estimated intake manifold pressure (p₁−{circumflex over (p)}₁) exceed some threshold. Below that threshold, this error could be caused just by measurement noise, therefore it is discarded and adaptation is stopped.

[0063] FIG. 7 shows some simulation results for the LS adaptive algorithm. These simulation results are for the case when three parameters, the multiplier on the EGR valve flow, θ_egr, the multiplier on the throttle flow rate, θ_th, and the multiplier on the in-cylinder flow rate θ_cylare unknown. In this case, periodic excitation is applied to the throttle and the EGR valve inputs (added to the nominal values) to assure parameter convergence. The simultaneous identification of three parameters requires the knowledge of the values of {tilde over (W)}_th, {tilde over (W)}_egr, and {tilde over (W)}_cyl, such that W_th=θ_th{tilde over (W)}_th, W_egr=θ_egr{tilde over (W)}_egr, W_cyl=θ_cyl{tilde over (W)}_cyl. Since only the value of θ_cylrequires adaptation in the present invention, application of periodic excitation to the throttle and the EGR valve values can be avoided by closing the EGR valve during the adaptation.

[0064] A step-by-step procedure that avoids applying periodic excitation while identifying all three parameters is also possible. First, θ_thcan be adapted using the measurements of the mass airflow. θ_cylcan then be adapted with the EGR valve closed. And finally, θ_egrcan be adapted from the estimate of the EGR flow rate W_egrprovided by the estimator. Since the estimator provides an accurate representation for the EGR flow rate only when the variations in this flow rate are relatively slow, the knowledge of θ_egrcan be used in an open loop estimate of the EGR flow rate, W_egr=θ_egr{tilde over (W)}_egr, which is accurate even when changes in the flow rate are fast. The accurate estimate of the EGR flow rate is important for accurately estimating the burnt gas fraction in the intake manifold and, hence, setting the spark timing correctly.

[0065] Besides the deposits, another problem in estimating W_cylaccurately is the estimation of the intake manifold temperature, T₁. Clearly, if T₁is not measured, its estimate must take into account the information about both the EGR flow rate and the in-cylinder flow rate, both of which are being estimated. To avoid this “circular” estimate dependence, the use of an intake manifold temperature sensor 38, to measure T₁, is provided in FIG. 4. Typical temperature sensors are rather slow, and additional compensation is required. This compensation can be provided as follows. Suppose that the sensor dynamics are

[0066] {dot over (T)}_1,m+λT_1,m=λT₁,

[0067] where T₁is the intake manifold temperature and T_1,mis its measurement where 1/λ is the time constant of the temperature sensor. By regressing engine variables, an estimate of T₁,{tilde over (T)}₁, for example, can be determined as

[0068] {tilde over (T)}₁=f(N,W_f,Ŵ_t,Ŵ_egr).

[0069] This estimate may not be, and does not have to be, very precise; and to improve it, a correction is added from the sensor measurement of the form,

[0070] {circumflex over (T)}₁={tilde over (T)}₁+Δ{circumflex over (T)}₁,

[0071] where $\Delta {\hat{T}}_1=\frac{{\alpha }_0}{\lambda }{T}_{1,m}-ϵ,\stackrel{.}{ϵ}=-{\alpha }_0ϵ+{\alpha }_0\left(-T_{1,m}+{\tilde{T}}_1\right)+\frac{{\alpha }_0^2}{\lambda }{T}_{1,m}.$

[0072] Here α₀is a calibratable constant.

[0073] Referring now to FIG. 8, a flowchart for the method of the present invention is shown. In block 60, the desired values or set points for burnt gas fraction, intake manifold pressure, and mass flow rate to the cylinder are updated. In block 62, the set points for throttle and EGR valve positions are updated. In block 64, the estimated values of flow rate to the cylinder and through the EGR and ETC valves is determined as well as the estimated values of intake manifold pressure and temperature. In block 66, the feedback corrections to the throttle and EGR valve control are performed. If throttle adaptation is enabled, as determined at block 68, and the error between the estimated and measured throttle flow rate is greater than a threshold g₀, as determined in block 70, then the values in the throttle effective flow area lookup table are adjusted to reduce the estimation error as indicated in block 72. If throttle adaption is not enabled, or the throttle flow rate estimation error is less than the threshold g₀, a check is made at block 74 to determined if in-cylinder flow rate adaptation is enabled. If enabled, and the EGR valve is not open as determined in block 76, then if the estimated intake manifold pressure error is greater than a predetermined error threshold f₀as determined in block 78, the in-cylinder air charge mapping table is adjusted as indicated in block 80.

[0074] While the best mode for carrying out the present invention has been described in detail, those familiar with the art to which this invention relates will recognize various alternative designs and embodiments for practicing the invention as defined by the following claims. For example, while the invention has been described in connection with a direct injection lean burn engine it is also applicable to the port fuel injection engine.