DE69627218T2 - Control system for the fuel metering of an internal combustion engine - Google Patents

Description

BACKGROUND OF THE INVENTION 1. Field of the invention

This invention relates to a fuel metering control system for an internal combustion engine.

2. Description of the state of the art

It is known in the prior art to install an air/fuel ratio sensor at an exhaust confluence of a multi-cylinder internal combustion engine to sense the air/fuel ratio of the engine exhaust gas and to control the air/fuel ratio to a target value, as taught by, for example, Japanese Patent Publication No. Sho 62(1987)-20,365. In the prior art system, based on an output of a single air/fuel ratio sensor installed in the exhaust system, the confluence point air/fuel ratio feedback and the individual cylinder air/fuel ratio feedback are controlled at different times. EP 0408 206 discloses another fuel metering control system having an air/fuel ratio sensor and two controllers which receive the output of the air/fuel ratio sensor as a controlled variable and control it by means of a manipulated variable of the fuel injection amount.

EP 0492 431 and US 4990 235 both show the filtering of the output of an exhaust gas oxygen sensor, which is particularly suitable for use in fuel metering control systems for internal combustion engines.

As disclosed in the above-mentioned Japanese Patent Publication No. Sho 62(1987)-20,365, when controllers operate using a single sensor as a common input signal, interference may sometimes occur between the controllers, making the control unstable. More specifically, the problem occurs more frequently when controllers are provided in parallel or in multiple application. In addition, since the output of the air-fuel ratio sensor contains noise, it is preferable to eliminate the noise as much as possible in order to improve the detection accuracy.

SUMMARY OF THE INVENTION

It is therefore an object of the invention to provide a fuel metering control system for an internal combustion engine which can solve the above-mentioned problem and, even when multiple-actuated controllers are provided to operate using an output of a single air-fuel ratio sensor as a common input signal, enables interference between the controllers to be prevented while eliminating noise at the sensor output, thus enabling the controllers to operate stably.

This invention solves this problem by providing a system for controlling fuel metering for a multi-cylinder internal combustion engine, comprising:

an air-fuel ratio sensor installed on the exhaust system of the engine to detect an air-fuel ratio of the engine;

an engine operating condition detecting means for detecting the engine operating conditions including at least the engine speed and the engine load ;

a fuel injection amount determining means for determining the fuel injection amount for the individual cylinders at least on the basis of the detected engine operating conditions;

a plurality of controllers for inputting the output signal of the air/fuel ratio sensor as a controlled variable and for determining the feedback correction coefficients for correcting the fuel injection quantity, which is fed to the motor as a manipulated variable so that the controlled variable is brought to a setpoint;

an initial fuel injection amount determining means for determining the initial fuel injection amount using the feedback correction coefficients; and

a fuel injection device for injecting fuel into each cylinder of the engine based on the initial fuel injection amount;

characterized in that

the plurality of controllers determine the feedback correction coefficients based on setpoints that are determined separately; and

at least one of the plurality of controllers receives the output signal of the air/fuel ratio sensor through a filter such that the input from the output of the air/fuel ratio sensor input to the at least one controller is different from the input from the output of the air/fuel ratio sensor input to another of the plurality of controllers.

BRIEF EXPLANATION OF THE DRAWINGS

These and other objects and advantages of the invention will become more apparent from the following description and drawings in which:

Fig. 1 is an overall schematic view showing a fuel metering control system for an internal combustion engine according to the present invention;

Fig. 2 is a schematic view showing the details of an exhaust gas recirculation (EGR) mechanism shown in Fig. 1;

Fig. 3 is a schematic view showing details of a container flushing mechanism shown in Fig. 1;

Fig. 4 is a graph showing the valve timing characteristics of a variable valve timing mechanism shown in Fig. 1;

Fig. 5 is an explanatory view showing the catalyst and an example of O₂ sensor positioning shown in Fig. 1;

Fig. 6 is a block diagram showing details of the control unit shown in Fig. 1;

Fig. 7 is a graph showing the output of the O2 sensor shown in Fig. 1;

Fig. 8 is a block diagram showing the configuration of the system according to the invention;

Fig. 9 is a flow chart showing the determination or calculation of the basic amount of fuel injection TiM-F shown in Fig. 8;

Fig. 10 is a block diagram showing the determination or calculation of the base value of the fuel injection quantity TiM-F referred to in the calculation of Fig. 9;

Fig. 11 is a block diagram showing the calculation of an effective throttle opening area and its first order lag value used in the calculation related to the calculation of Fig. 9;

Fig. 12 is a graph showing characteristics of map data of a coefficient shown in Fig. 11;

Fig. 13 is a view showing the characteristics of the map data of the fuel injection amount in the steady state engine operation referred to in the calculation of Fig. 9;

Fig. 14 is a view explaining the characteristics of map data of a basic value of a target air-fuel ratio referred to in the calculation of Fig. 9;

Fig. 15 is a graph showing the result of a simulation performed in the Calculation of Fig. 9 is referred to;

Fig. 16 is a timing chart illustrating the transitional engine operating state referred to in the calculation of Fig. 9;

Fig. 17 is a timing chart explaining the first-order delay value of the effective throttle opening area;

Fig. 18 is a view similar to Fig. 10, showing the calculation of Fig. 9;

Fig. 19 is a flow chart showing an estimation of the EGR rate in the calculation of the EGR correction coefficient referred to in the explanation of the configuration shown in Fig. 8;

Fig. 20 is an explanatory view showing the flow rate characteristics of the EGR control valve determined by the valve lift amount and the relationship between upstream pressure (manifold absolute pressure) and downstream pressure (atmospheric pressure);

Fig. 21 is a timing chart showing the operation of the actual valve lift to the target valve lift;

Fig. 22 is an explanatory view showing the characteristics of the map data of a coefficient KEGRMAP;

Fig. 23 is an explanatory view showing the characteristics of map data of a command value for the valve lift amount LCMD;

Fig. 24 is a flow chart showing the subroutine of the flow chart of Fig. 13 for calculating a coefficient KEGRN;

Fig. 25 is an explanatory view showing the configuration of a ring buffer used in the flow chart of Fig. 24;

Fig. 26 is an explanatory view showing the characteristics of the map data of a delay time τ used in the flow chart of Fig. 25;

Fig. 27 is a timing diagram showing a delay in the actual valve lift to a set point and a further delay until the exhaust gas has reached the combustion chamber of the engine;

Fig. 28 is a flow chart showing the determination or calculation of the tank purge correction coefficient KPUG referred to in the explanation of the configuration shown in Fig. 8;

Fig. 29 is a flow chart showing the determination or calculation of the target air-fuel ratio or the target air-fuel ratio correction coefficient referred to in the explanation of the configuration shown in Fig. 8;

Fig. 30 is a graph showing the correction for the charging efficiency referred to in the flow chart of Fig. 29;

Fig. 31 is an explanatory view showing the relationship between the air-fuel ratio at the confluence point of the exhaust system of an engine and the TDC crank position;

Fig. 32 is an explanatory view showing appropriate (best) sampling timings of the air-fuel ratio sensor outputs as opposed to inappropriate sampling timings;

Fig. 33 is a flow chart showing the operation of air-fuel ratio sampling performed by the sampling block shown in Fig. 8;

Fig. 34 is a block diagram showing a model illustrating the behavior of the air/fuel ratio detection referred to in a prior application of the assignee;

Fig. 35 is a block diagram showing the model of Fig. 34 discretized in discrete time series for a period Delta-T;

Fig. 36 is a block diagram showing a real-time air-fuel ratio estimator based on the model of Fig. 35;

Fig. 37 is a block diagram showing a model describing the behavior of the exhaust system of the engine referred to in a prior application of the assignee;

Fig. 38 is a graph showing the preconditions of a simulation in which it is assumed that fuel is supplied to three cylinders of a four-cylinder engine so as to obtain an air/fuel ratio of 14.7:1 and is supplied to one cylinder;

Fig. 39 is a graph showing the result of simulation showing the output of the exhaust system model and the air/fuel ratio at a confluence point when the fuel is supplied in the manner described in Fig. 38;

Fig. 40 is the result of the simulation showing the output of the exhaust system model provided for a sensor detection response delay (time lag) in contrast to the actual sensor output;

Fig. 41 is a block diagram showing the configuration of an ordinary observer;

Fig. 42 is a block diagram showing the configuration of the observer referred to in a prior application of the assignee;

Fig. 43 is an explanatory block diagram showing the configuration achieved by combining the model of Fig. 37 and the observer of Fig. 42;

Fig. 44 is a block diagram showing the overall configuration of the Shows air/fuel ratio feedback loops in the system;

Fig. 45 is an explanatory view showing the characteristics of a timing map referred to in the flow chart of Fig. 33;

Fig. 46 is a timing chart showing the characteristics of the sensor output with respect to the engine speed and a timing chart showing the characteristics of the sensor output with respect to the engine load;

Fig. 47 is a timing diagram showing the sampling of the air/fuel ratio sensor in the system;

Fig. 48 is a timing chart showing the air-fuel ratio detection delay when fuel supply is resumed after a shutdown;

Fig. 49 is a flow chart showing the feedback correction coefficient to be referred to in the explanation of the configuration shown in Fig. 8;

Fig. 50 is a block diagram explaining the calculation of Fig. 49;

Fig. 51 is a subroutine flow chart of Fig. 49 showing the calculation of the feedback correction coefficient in more detail;

Fig. 52 is a subroutine flow chart similar to Fig. 51;

Fig. 53 is a timing chart explaining the calculation of Fig. 51;

Fig. 54 is a subroutine flowchart of Fig. 49 showing the fuel adhesion correction of the initial fuel injection amount;

Fig. 55 is a graph showing the direct ratio and so on referred to in the flow chart of Fig. 54;

Fig. 56 is a graph showing the characteristics of the coefficient referred to in the flow chart of Fig. 54;

Fig. 57 is a subroutine flow chart of Fig. 54; and

Fig. 58 is a view similar to Fig. 8, but showing the configuration of the system according to the second embodiment of the invention.

PREFERRED EMBODIMENTS OF THE INVENTION

Embodiments of the invention are explained below with reference to the drawings.

Fig. 1 is an overview of a fuel metering control system for an internal combustion engine according to the invention.

Reference numeral 10 in this figure denotes an in-line four-cylinder overhead cam (OHC) internal combustion engine. The air drawn into the air intake pipe 12 through an air cleaner 14 mounted at the distal end of the intake pipe 12 is supplied to the first through fourth cylinders via a collector tank 18, an intake manifold 20 and two intake valves (not shown) while its flow rate is adjusted by means of a throttle valve 16. A fuel injector 22 for injecting fuel is installed near the intake valves of each cylinder. The injected fuel mixes with the intake air to form an air-fuel mixture which is ignited in the associated cylinder by means of a spark plug (not shown) in the firing order of cylinders 1, 3, 4 and 2. The resulting combustion of the air-fuel mixture drives a piston (not shown) downward.

The exhaust gas produced by the combustion is discharged through two exhaust valves (not shown) into an exhaust manifold 24, from where it passes through an exhaust pipe 26 to a first catalyst (3-way catalyst) 28 and a second catalyst 30 (also a 3-way catalyst) where harmful components are removed from it before it is discharged into the surrounding atmosphere. The throttle valve 16 is not mechanically connected to the accelerator pedal (not shown) and is controlled to a desired degree of opening by means of a stepping motor M. In addition, the throttle valve 16 is bypassed by means of a bypass 32 which is provided on the air intake line 12 in the vicinity thereof.

The engine 10 is equipped with an exhaust gas recirculation (EGR) mechanism 100 that recirculates a portion of the exhaust gas to the intake side.

More specifically, as shown in Fig. 2, the exhaust gas recirculation mechanism 100 includes an exhaust gas recirculation pipe 121 having one end (port) 121a connected to the exhaust gas pipe 26 upstream of the first catalyst 28 (not shown in Fig. 2) and another end (port) 121b connected to the air intake pipe 12 upstream of the throttle valve 16 (not shown in Fig. 2). For regulating the amount of recirculated exhaust gas, an EGR (Exhaust Gas Recirculation) control valve 122 and a surge tank 121 are provided at an intermediate portion of the exhaust gas recirculation pipe 121. The EGR control valve 122 is a solenoid valve having a solenoid coil 122a connected to a control unit (ECU) 34 (described later). The EGR control valve 122 is controlled to a desired opening degree by an output from the control unit 34 to the solenoid coil 122a. The EGR control valve 122 is provided with a stroke sensor 123 which detects the opening degree of the EGR control valve 122 and sends a corresponding signal to the control unit 34.

The engine 10 is further equipped with a canister purge mechanism 200 connected between the intake system and a fuel tank 36 .

As shown in Fig. 3, the canister purge mechanism 200, which is provided between the top of the sealed fuel tank 36 and a point on the air intake pipe 12 downstream of the throttle valve 16, comprises a vapor supply pipe 221, a container 223 containing an absorbent 231, and a purge pipe 224. The vapor supply pipe 221 is provided with a two-way valve 222, and the purge pipe 224 is provided with a purge control valve 225, a flow meter 226 for measuring the amount of air-fuel mixture containing fuel vapor. mixture flowing through the purge line 224 and a hydrocarbon (HC) concentration sensor 227 for detecting the HC concentration of the air-fuel mixture. The purge control valve (solenoid valve) 225 is connected to the control unit 34 and is linearly controlled to the desired opening degree by a signal from the control unit 34.

When the amount of fuel vapor generated in the fuel tank 36 reaches a prescribed level, it pushes open the relief valve of the two-way valve 222 and flows into the canister 223, where it is stored by absorption on the absorbent 231. Subsequently, when the purge control valve 225 is opened to an extent corresponding to the duty ratio of the on/off signal from the control unit 34, the evaporated fuel temporarily stored in the canister 223 and air sucked through an external air inlet 232 are sucked together into the air intake pipe 12 due to the negative pressure in the air intake pipe 12. On the other hand, when the negative pressure in the fuel tank 36 is e.g. B. due to cooling of the fuel tank by the ambient air temperature, the negative pressure opens the two-way valve 222 to allow the vaporized fuel temporarily stored in the canister 223 to return to the fuel tank 36.

The engine 10 is further equipped with a variable valve timing mechanism 300 (denoted V/T in Fig. 1). For example, as taught by Japanese Patent Application Laid-Open No. Hei 2(1990)-275,043, the variable valve timing mechanism 300 switches the opening/closing timing of the intake and/or exhaust valves between two types of timing characteristics, namely, a low engine speed characteristic, denoted LoV/T, and a high engine speed characteristic, denoted HiV/T, as shown in Fig. 4, in response to the engine speed Ne and the manifold pressure Pb. However, since this is a well-known mechanism, it will not be further described here. (Among the various ways of switching between valve timing characteristics is that of deactivating one of the two intake valves.)

The engine 10 of Fig. 1 is equipped in its ignition distributor (not shown) with a crank angle sensor 40 for detecting the piston crank angles, and is further provided with a throttle position sensor 42 for detecting the opening degree of the throttle valve 16 and a manifold absolute pressure sensor 44 for detecting the pressure Pb of the intake manifold downstream of the throttle valve 16 in absolute value terms. An atmospheric pressure sensor 46 for detecting the atmospheric pressure Pa is provided at an appropriate portion of the engine 10, an intake air temperature sensor 48 for detecting the temperature of the intake air upstream of the throttle valve 16, and a coolant temperature sensor 50 for detecting the temperature of the engine coolant is provided at an appropriate portion of the engine. The engine 10 is further provided with a valve timing (V/T) sensor 52 (not shown in Fig. 1) which detects the valve timing characteristic selected by the variable valve timing mechanism 300 based on the oil pressure.

Further, an air-fuel sensor 54, which is an oxygen detector or oxygen sensor, is provided in the exhaust pipe 26 at a confluence point or downstream thereof in the exhaust system downstream of the exhaust manifold 24 and upstream of the first catalyst 28, where it detects the oxygen concentration in the exhaust gas at the confluence point and generates a corresponding signal (explained later). In addition, an O₂ sensor 56 as a second oxygen sensor is provided downstream of the air-fuel ratio sensor 34 (first oxygen sensor). The volumes of the first and second catalysts are appropriately determined taking into account the purification (conversion) efficiency and the temperature characteristics, and are set to about 1 liter for the first catalyst and about 1.7 liters for the second catalyst, for example.

As shown in Figure 5, the first catalyst 28 may be configured to have multiple beds each supporting a catalyst, particularly double beds in the illustration comprising a first catalyst bed and a second catalyst bed. When the first catalyst 28 is configured as shown, the O₂ sensor 56 may be positioned between the first and second beds as shown. In this case, the volume of catalyst supported on the first bed is about 1 liter, that on the second bed being about 1 liter or so. The first catalyst 28 accordingly has a volume of about 2.0 liters when configured as shown. However, since the configuration shown is the same as the case where the O2 sensor is installed downstream of a single catalyst having a capacity of 1.0 liter, the sensor output switching interval becomes shorter than the case where the sensor is positioned downstream of the catalyst having a volume of 2 liters. Thus, when the accurate air-fuel ratio control (explained later) is performed within a catalyst window defined by the outputs of the O2 sensor 56 thus positioned, the control accuracy is improved. The accurate air-fuel ratio control is hereinafter referred to as "MIDO2 control."

The air/fuel ratio sensor 54 is followed by a filter 58, while the O2 sensor 56 is followed by a second filter 60. The outputs of the sensors and filters are sent to the control unit 34.

Details of the control unit 34 are shown in the block diagram of Fig. 6. The output of the air/fuel ratio sensor 54 is received by a first detecting circuit 62 where it is subjected to appropriate linearization processing to produce an output characterized by varying linearly with the oxygen concentration of the exhaust gas over a wide range extending from the lean side to the rich side. (The air/fuel ratio sensor is referred to as the "LAF sensor" in the figure and will be referred to as such in the remainder of this specification.) The output of the O2 sensor is input to a second detection circuit 64 which produces a switching signal indicating that the air/fuel ratio in the exhaust gas emitted from the engine 10 is rich or lean with respect to the stoichiometric air/fuel ratio (= lambda = 1) as shown in Fig. 7.

The output of the first detection circuit 62 is passed to a CPU (central processing unit) via a multiplexer 66 and an A/D converter 68. The CPU has a CPU core 70, a ROM (read only memory) 72 and a RAM (random access memory) 74, the output of the first detecting circuit 62 is A/D converted once for every prescribed crank angle (e.g., 15°) and stored in buffers of RAM 74. As shown in Fig. 47 described later, RAM 74 has twelve buffers numbered 0 to 11, and the A/D converted outputs from the detecting circuit 62 are sequentially stored in the twelve buffers. Similarly, the outputs of the second detecting circuit 64 and the analog outputs of the throttle position sensor 42 and the like are input to the CPU via the multiplexer 66 and the A/D converter 68 and stored in RAM 74.

The output of the crank angle sensor 40 is shaped by the signal converter 76, and its output value is counted by a counter 78. The result of the counting is input to the CPU. According to the instructions stored in the ROM 72, the CPU core 70 calculates a manipulated variable in a manner described later and controls the fuel injectors 22 of the respective cylinders via a driver circuit 82. Via the driver circuits 84, 86 and 88, the CPU core 70 further actuates a solenoid valve (EACV) 90 (for opening and closing the bypass 32 to control the amount of secondary air), the solenoid valve 122 for controlling the above-mentioned exhaust gas recirculation, and the solenoid valve 225 for controlling the above-mentioned canister purge. (The stroke sensor 123, the flow meter 226 and the HC concentration sensor 227 are omitted in Fig. 6.)

Fig. 8 is a block diagram showing the operation of the fuel metering control according to the embodiment.

As shown, the system is provided with an observer (labeled "OBSV" in the figure) that estimates the air/fuel ratios at the individual cylinders from the output of the individual LAF sensor 54 installed on the exhaust system of the engine 10, and an adaptive controller (self-tuning controller; labeled "STR" in the figure) that receives the output of the LAF sensor 54 via a filter 92.

The output of the O₂ sensor 56, labeled "VO₂M", is a target air-fuel ratio correction block (referred to as "KCMD correction" in the figure), where a target air-fuel ratio correction coefficient designated "KCMDM" is determined in accordance with an error between the O₂ sensor output "VO₂M" and a target value (VrefM in Fig. 7). On the other hand, the basic amount of fuel injection TiM-F is determined based on the change in the effective opening area of the throttle valve 16 in the manner explained later. The basic amount of fuel injection TiM-F is multiplied by the target air-fuel ratio correction coefficient KCMDM and another correction coefficient KTOTAL (the product of other correction coefficients including the correction coefficients for the EGR and the canister purge) to determine the amount of fuel injection which is assumed to be required by the engine (referred to as the "required amount of fuel injection Tcyl").

On the other hand, the corrected target air/fuel ratio KCMD is input to the adaptive controller STR and a PID controller (denoted by "PID" in the figure), which respectively determine feedback correction coefficients designated KSTR or KLAF in response to an error from the LAF sensor output. One of the feedback correction coefficients is selected via a switch in response to the operating conditions of the engine and is multiplied by the target fuel injection amount Tcyl to determine the initial fuel injection amount designated Tout. The initial fuel injection amount is then subjected to fuel adhesion correction, with the corrected amount finally being supplied to the engine 10.

Thus, the air/fuel ratio is controlled to the target air/fuel ratio based on the LAF sensor output, with the above-mentioned MIDO₂ control being implemented at or around the target air/fuel ratio, ie, within the catalyst window. The catalyst functions to store O₂ from the exhaust gas of a relatively lean mixture. When the catalyst is saturated with O₂, the purification efficiency decreases. It is therefore necessary to provide exhaust gas of a relatively rich mixture so as to rid the catalyst of the stored O₂, and after the completion of the liberation of the stored O₂, the purification efficiency decreases. the exhaust gas is again provided with a relatively lean mixture. By repeating this process it is possible to maximize the cleaning efficiency. The MIDO₂ control is designed to achieve this.

In order to further improve the purification efficiency in the MID O₂ control, it is necessary to bring the catalyst air/fuel ratio to the catalyst in a shorter time after switching the O₂ sensor output. In other words, it is necessary to bring the actual air/fuel ratio (hereinafter referred to as "KACT") to a target air/fuel ratio KCMD in a shorter time period. When the fuel injection amount determined in the controlled system, i.e., TiM-F, is merely multiplied by the target air/fuel ratio feedback correction coefficient KCMDM, the target air/fuel ratio becomes a smoothed value of the actual air/fuel ratio KACT due to the engine response delay.

Accordingly, to solve the problem, the disclosed system is configured to ensure the response of the detected air-fuel ratio KACT dynamically. More specifically, the fuel injection amount is multiplied by the correction coefficient KSTR (output of the adaptive controller) which ensures the desired behavior of the target air-fuel ratio KCMD. With this arrangement, it becomes possible to allow the actual air-fuel ratio KACT to immediately converge to the target air-fuel ratio KCMD to improve the catalyst purification (conversion) efficiency.

Note that the calculation is facilitated by actually representing the target value KCMD and the actual value KACT as an equivalence ratio, i.e. as Mst/M = 1/Lambda (Mst: stoichiometric air/fuel ratio, M = A/F, A: air mass flow rate, F: fuel mass flow rate, and Lambda = excess air factor).

An explanation of the filters follows.

The configuration shown is designed as a multi-loaded control system in which several feedback loops are provided in parallel which all use a common output of the single LAF sensor 54. More specifically, the system is configured so that the multiple-loaded or multiple feedback loops are connected. The frequency characteristics of the filters are thus determined according to the nature of the feedback loops.

More specifically, it takes 400 ms (milliseconds) for the LAF sensor to obtain a 100% response. Here, the time to obtain the 100% response means a time until the LAF sensor output (which changes with a first order delay) becomes flat when a step input signal of an air-fuel mixture is given. More specifically, a time until the sensor output approaches the stoichiometric air-fuel ratio (lambda = 1) when a stoichiometric mixture is input after a rich mixture (lambda = 1.2) is input. The time is almost the same as the so-called settling time. The sensor output approaches the target value but does not reach it due to a static error.

If the sensor outputs are left unchanged, they contain high frequency noise, deteriorating the control performance. The inventors have found through experiments that if the sensor outputs are passed through a low-pass filter whose cut-off frequency is 500 Hz, the high frequency noise can be removed without significantly deteriorating the response characteristics. If the cut-off frequency of a filter is lowered to 4 Hz, the high frequency noise can be further reduced to a considerable extent, and the time required for 100% response becomes stable. However, the response characteristics of the filter in this case were more delayed than in the case where the sensor output was filtered or passed through a filter with a cut-off frequency of 500 Hz, and required 400 ms or more until 100% response was obtained.

In view of the foregoing, the filter 58 is set as a low-pass filter with a cut-off frequency of 500 Hz, and the sensor output fed to the filter is directly input to the observer. The observer does not cause the actual air/fuel ratio KACT to converge toward the target air/fuel ratio KCMD. Rather, the system is configured so that the air/fuel ratios in each cylinder are estimated by the observer, while the scatter between the individual cylinder air/fuel ratios is absorbed by the PID controller. Consequently, even if the sensor response time is not stable, this does not affect the air/fuel detection. Rather, the shorter response time improves the control performance.

On the other hand, the filter 92 (shown only in Fig. 8) placed before the adaptive controller STR should be a low-pass filter with a corner frequency of 4 Hz. This is because the aperiodic controller such as the STR functions to reliably compensate for the air/fuel ratio detection delay, and any change in noise or response time in the air/fuel detection would affect the control performance. For this reason, the low-pass filter 92 is provided with the corner frequency of 4 Hz. In addition, the filter 93 placed before the input of the PID controller must be a filter whose corner frequency is equal to or greater than that of the filter 92, more precisely 200 Hz, taking the response time into account.

In addition, the filter 60 connected to the O2 sensor 56 is set as a low-pass filter whose corner frequency is equal to 1,600 Hz, since the response of the O2 sensor is much larger than that of the LAF sensor.

With this arrangement, when multiple controllers operate using a common output of the single LAF sensor, since the frequency characteristics of the filters are determined taking into account the purpose and nature of the controllers, it becomes possible not to improve the detection accuracy but to more effectively prevent interference between the controllers and further to achieve an optimal balance between the control response and the control stability.

The operation of the system according to the invention is explained below with reference to the block diagram of Fig. 8.

First, the basic fuel injection quantity TiM-F is determined or calculated.

As mentioned above, the basic fuel injection amount TiM-F is optimally determined in all engine operating conditions including engine transients based on the change in the effective throttle opening area.

Fig. 9 is a flow chart for determining or calculating the basic fuel injection amount TiM-F, while Fig. 10 is a block diagram explaining the operation shown in Fig. 9. Before explaining the figures, however, the estimation of the throttle valve passing air amount and the cylinder intake air amount using a fluid dynamic model on which the invention is based will be explained first. Since the method has been fully described in Japanese Patent Application Laid-Open Hei 6(1994)-197,238 (filed in the United States on July 27, 1995 under No. 081507,999) proposed by the assignee, the explanation will be kept brief.

More specifically, on the basis of the detected throttle opening θTH, the throttle projection area S (which is formed on a plane perpendicular to the longitudinal direction of the air intake pipe 12 when the throttle valve 16 is assumed to be projected in this direction) is determined according to a predetermined characteristic as shown in the block diagram of Fig. 11. At the same time, the discharge coefficient C, which is the product of the flow rate coefficient α and the gas expansion factor epsilon, is retrieved from map data whose characteristics are shown in Fig. 12 using the throttle opening θTH and the manifold pressure Pb as address data, and the throttle projection area S is multiplied by the retrieved coefficient C to obtain the effective throttle opening area A. Since the throttle valve does not function as an orifice in its wide open state (full throttle state), the full-load orifice areas are empirically determined in advance as limited values with respect to engine speeds. If the detected throttle valve opening is found to exceed the relevant limit value, the detected value is limited to the limit value. The value is further subjected to atmospheric correction (explanation omitted).

Next, the chamber-filling air quantity, hereinafter referred to as "Gb", is calculated using equation 1, which is based on the ideal gas law. The term "chamber" is used here to refer not only to the part corresponding to the so-called surge tank, but to all sections extending from immediately downstream of the throttle valve to immediately before the cylinder inlet port:

where:

V: Chamber volume

T: Air temperature

R: Gas constant

P: Chamber pressure

Then, the chamber-filling air quantity in the current control cycle Delta-Gb(k) can be obtained from the pressure change in the chamber Delta-P using Equation 2.

Note that "k" is used throughout the description to denote a discrete variable and is the simple number in the discrete system, more precisely the control or calculation cycle (program loop), or more precisely the current control or calculation cycle (current program loop). "kn" thus denotes the control cycle at a time cycles earlier in the discrete control system. The appending of the suffix "k" is omitted in the description for most variables in the current control cycle:

If it is assumed that the chamber-filling air quantity Delta-Gb(k) has not actually been fed into the cylinder in the current control cycle, the cylinder intake air quantity Gc per unit time Delta-T can be expressed by equation 3:

Gc = Gth·ΔT - ΔGb Eq. 3

On the other hand, the fuel injection amount at the steady state engine operating condition Timap is prepared in advance according to the so-called speed density method and stored in the ROM 72 as map data (the characteristics of which are shown in Fig. 13) with respect to the engine speed Ne and the manifold pressure Pb. Since the fuel injection amount Timap in the map data is corrected by means of a target air-fuel ratio which in turn is determined according to the engine speed Ne and the manifold pressure Pb, the target air-fuel ratio, more precisely its base value KBS, is prepared in advance and stored as map data with respect to the same parameters as shown in Fig. 14. However, since the correction of the value Timap with the target air-fuel ratio relates to the MIDO₂ control, the correction is not carried out here. The correction with the target air/fuel ratio and the MIDO₂ control will be explained later. The fuel injection amount Timap is determined in terms of the opening period of the fuel injection device 22.

Here, when the relationship between the fuel injection amount Timap, which is retrieved from the map data, and the amount of air passing through the throttle valve Gth is considered, the fuel injection amount Timap, which is recovered from the map data, here denoted as Timap1, at a certain aspect under the steady-state engine operating condition defined by the engine speed Ne1 and the manifold pressure Pb1, is expressed by Equation 4:

Timap1 = MATERIAL DATA(Ne1, Pb1) Eq. 4

As described in the above-mentioned prior application (6-197,238) of the assignee, it has been found that the throttle valve passing air amount Gth under the transient engine operating condition can be determined from that under the steady engine operating condition in response to the change in the throttle valve effective opening area. More specifically, it has been found that the throttle valve passing air amount Gc can be determined using a ratio between the throttle valve effective opening area under the steady engine operating condition and that under the transient engine operating condition.

Further, if the current effective throttle opening area is denoted by A (which may be the area under the transient engine operating condition) and the effective throttle opening area under the steady-state engine operating condition is denoted by A1, it is assumed that the value A1 can be determined as the first-order delay of A. This was confirmed by simulation on the computer as shown in Fig. 15. If the first-order delay value of A is denoted by ADELAY, it can be confirmed from the figure that the values A1 and ADELAY are almost equal. Accordingly, it follows that the amount of air passing through the throttle valve Gth can be determined based on the ratio A/first-order delay of A, i.e., A/ADELAY, if the model is approximated using the concept of the fluid dynamic model.

As shown in Fig. 16, under the transient engine operating condition when the throttle valve is opened, due to the large pressure difference across the throttle valve, a large amount of air passes through the throttle valve at once, whereupon the air amount gradually returns to that under the stationary engine operating condition as previously mentioned with reference to the lower portion of Fig. 16. It has been assumed that the ratio A/ADELAY can describe the amount of air passing through the throttle valve Gth under such a transient engine operating condition. Under the steady-state engine operating condition, the ratio becomes equal to 1 as is clear from the lower portion of Fig. 17. The ratio is hereinafter referred to as "RATIO-A".

Further, when the relationship between the effective throttle opening area and the throttle opening θTH is considered, since the effective throttle opening area largely depends on the throttle opening, it can be assumed that the effective throttle opening area changes almost reliably following the change in the throttle opening, as shown in Fig. 17. If this is true, it can be said that the above-mentioned first-order delay value of the throttle opening is, in the sense of the phenomenon, almost equal to the first-order delay value of the effective throttle opening area.

In view of the foregoing, as shown in Fig. 10, it is arranged that the first-order delay value of the effective throttle opening area ADELAY is calculated mainly from the first-order throttle opening. In the figure, (1 - B)/(z - B) is a transfer function of the discrete control system and denotes the value of the first-order delay.

As shown, the projection area S of the throttle valve is more precisely determined from the throttle opening θTH according to a predetermined characteristic, and the discharge coefficient C is determined from the first-order delay value of the throttle opening θTH-D and the manifold pressure Pb according to a characteristic similar to that in Fig. 12. Then, the product of the values is obtained to obtain the first-order delay value of the effective throttle opening area ADELAY. Further, in order to solve the reflection delay of the intake air amount according to the actual chamber-filling air amount Delta-Gb, the first-order delay value of the value Delta-Pb (first order delay of Pb) is used to determine Delta-Ti (which corresponds to Delta-Gb).

The configuration was further verified and it was found that the values of TiM-F and Delta-Ti (which correspond to Gth and Gb, respectively) do not need to be determined separately. Instead, it was found that TiM-F (which corresponds to Gth) can be determined to include Delta-Ti (which corresponds to Delta-Gb). More precisely, the cylinder intake air quantity Gc can be determined only from the air passing through the throttle valve Gth by using a transfer function that includes that of Delta-Ti when calculating ADELAY. This can make the configuration simpler and reduce the calculation workload.

More specifically, the cylinder intake air amount Gc per unit time Delta-T in Equation 1 can be expressed as Equation 5, which is equivalent to Equations 6 and 7. Rewriting Equations 6 and 7 in terms of the transfer function gives Equation 8. Thus, the value Gc can be obtained from the first-order delay value of the throttle-passing air amount Gth, as is clear from Equation 8. This is shown in a block diagram of Fig. 18. In Fig. 18, note that since the transfer function in the figure includes that of Delta-Ti and is different from that in Fig. 10, a symbol "'" is added as in (1 - B')/(z - B').

Gc = Gth(k) - Gb(k - 1) Eq. 5

Gc = α·Gth(k) - β·Gb(k - 1) Eq. 6

Gb = (1 - α)·Gth(k) + (1 - β)·Gb(k - 1) Eq. 7

Finally, the base fuel injection quantity TiM-F is determined or calculated as follows:

TiM-F = fuel injection amount TiM · (current or present effective throttle opening area / effective throttle opening area obtained based on manifold pressure Pb and first-order retardation value of throttle opening θTH-D) = TiM · RATIO-A

Based on the foregoing, the operation of the system is explained below with reference to the flow chart of Fig. 9.

The program starts at S10, where the detected engine speed Ne, manifold pressure Pb, throttle opening θTH, atmospheric pressure Pa, engine cooling water temperature Tb or the like are read. The throttle opening has been subjected to calibration (control learning) in the fully closed state at engine idling, using the detected value based on the calibration. The program then advances to S12, where it is checked whether the engine is running. If not, the program proceeds to S14 where it is checked whether the fuel cut is active and to S16 where the fuel injection amount TiM (equal to the fuel injection amount Timap under the steady state engine operating condition) is retrieved from the map data (the characteristics of which are shown in Fig. 13 and stored in the ROM 72) using the engine speed Ne and the manifold pressure Pb which have been read in as address data. However, although the fuel injection amount TiM may be subsequently subjected to atmospheric pressure correction or the like, the correction itself is not the essence of the invention and therefore will not be explained here.

The program then proceeds to S18 where the first order delay value of the throttle opening θTH-D is calculated, to S22 where the current or actual effective throttle opening area A is calculated using the throttle opening θTH and the manifold pressure Pb, and to S24 where the first order delay value of the effective throttle opening area ADELAY is calculated using the values θTH-D and Pb. The program then proceeds to S26 where the value RATIO-A is calculated as follows:

RATIO-A = (A + ABYPASS)/(A + ABYPASS)DELAY

Here, ABYPASS indicates a value corresponding to the amount of air that bypasses the throttle valve 16, such as that which flows in the secondary path 32 in response to the lift amount of the solenoid valve 74 (denoted as "solenoid valve lift amount" in Fig. 10) and is then introduced into the cylinder. Since it is necessary to take into account the amount of air that bypasses the throttle valve in order to accurately determine the fuel injection amount, the throttle bypass air amount is determined in advance in terms of the effective throttle opening area (denoted as ABYPASS) to be added to the effective throttle opening area A, i.e., A + ADELAY. A first order delay value of the sum (denoted "(A + ABYPASS)DELAY") is calculated, and then a ratio (i.e. RATIO-A) between the sum A + ABYPASS and the first order delay value of the same (A + ABYPASS)DELAY is calculated.

Since the ABYPASS value is added to both the denominator and the numerator in the equation shown in step S26, even if an error occurs in measuring the throttle bypass air amount, the determination of the fuel injection amount is not seriously affected.

The program then advances to S28 where the fuel injection amount Tim is multiplied by the ratio RATIO-A to obtain the fuel injection amount TiM-F corresponding to the throttle bypass air amount Gth.

If S12 determines that the engine is running, the program advances to S30 where the fuel injection amount Ticr during cranking is retrieved from a table (not shown) using the engine cooling water temperature Tw as address data, and to S32 where the basic fuel injection amount TiM-F is determined according to an equation for the engine running (explanation omitted) using the value Ticr, while if S14 determines that the fuel cut is active, the program advances to S34 where the basic fuel injection amount TiM-F is set to 0.

With the arrangement, it thus becomes possible to fully describe the conditions from a steady engine operating state to the transient engine operating state by a simple algorithm. Furthermore, it becomes possible to ensure the fuel injection amount under the steady engine operating state to a considerable extent by retrieving map data, and the fuel injection amount can thus be optimally determined without performing complicated calculations. Furthermore, since the equations are not switched between the steady engine operating state and the transient engine operating state, and since the equations can describe the entire range of engine operating states, no control discontinuity occurs which would otherwise occur in the vicinity of switching when switching between the equations for the steady engine operating state and the transient engine operating state. Furthermore, since the behavior of the air flow is suitably described, the arrangement can improve the convergence and accuracy of the control.

As shown in Fig. 8, the above-mentioned correction coefficient KTOTAL (a general name of various correction coefficients) including the EGR correction coefficient KEGR and the canister purge correction coefficient KPUG is determined or calculated.

The determination of the EGR correction coefficient is explained first.

Fig. 19 is a flowchart showing the operation of the EGR rate estimating system according to the invention.

However, before explaining the flow chart, the EGR rate estimation according to the invention will be briefly described with reference to Fig. 20 and the like.

When the EGR control valve 122 is considered alone, the flow rate of the exhaust gas passing through it is determined from its opening area (the lift amount) and the ratio between the upstream pressure and the downstream pressure at the valve. In other words, the Mass flow rate of exhaust gas passing through the valve is determined from the flow rate characteristics of the valve, ie from the valve design specification.

Thus, when the EGR control valve 122 is observed while it is mounted on the engine, it becomes possible to estimate the exhaust gas recirculation rate to an appropriate extent by detecting the EGR control valve lift amount and the relationship between the manifold pressure Pb (negative pressure) in the intake pipe 12 and the atmospheric pressure Pa, as shown in Fig. 20. (Although in practice the exhaust gas flow rate characteristics vary slightly with the exhaust manifold pressure and the exhaust gas temperature, the change can be absorbed by the relationship between the gas flow rates, as will be explained later.) The invention is based on this concept and estimates the EGR rate based on the flow rate characteristics of the valve.

It should be noted that although the valve opening area is detected by the valve lift amount, this is because the EGR control valve 122 used here has a structure whose lift amount corresponds to the opening area. Therefore, when another valve such as a linear solenoid is used, the valve opening area should be detected in a different way.

The EGR rate is classified into two kinds of rates, namely, one under a steady state and another under a transient state. Here, the steady state is a state in which the EGR operation is stable, while the transient state is a state in which the EGR operation is started or terminated so that the EGR operation is unstable. The EGR rate under a steady state is regarded as a value in which the actual valve lift amount is equal to the command value for the valve lift amount. On the other hand, the transient state is regarded as a state in which the actual valve lift amount is not equal to the command value as shown in Fig. 21, so that the EGR rate deviates from the EGR rate under a steady state (hereinafter referred to as "steady state EGR rate") by an amount equal to the exhaust gas flow rate corresponding to the discrepancy of the actual amount and the command value as shown in Fig. 20. (In the figure, the upstream pressure is indicated by the atmospheric pressure Pa, while the downstream pressure is indicated by the manifold pressure Pb).

More precisely, under a stationary state:

Command value = current valve lift, and

Gas flow rate corresponding to the current valve lift amount /Gas flow rate corresponding to the command value = 1.0

While under a transition state:

Command value ≠ current valve lift, and

Gas flow rate corresponding to the current valve lift amount /Gas flow rate corresponding to the command value + 1.0

As a result, it can be concluded that:

Net EGR rate = (steady state EGR rate) · (ratio between the gas flow rates).

To distinguish the steady-state EGR rate, the EGR rate is sometimes referred to as the "net" EGR rate.

It is thus assumed that it is possible to estimate the exhaust gas recirculation rate by multiplying the steady-state EGR rate by the ratio between the gas flow rates corresponding to the current valve lift amount and the command value.

More specifically, it is assumed that:

Net EGR rate = (steady state EGR rate) · {(gas flow rate QACT which is determined by the current valve lift amount and the ratio between the upstream pressure and the downstream pressure of the valve) / (gas flow rate QCMD which is determined by the command value and the ratio between the upstream pressure and the downstream pressure of the valve)}

Here, the steady-state EGR rate is calculated by obtaining a correction coefficient under a steady state and subtracting it from 1.0. That is, if the correction coefficient under a steady state is denoted by KEGRMAP, the steady-state EGR rate can be calculated as follows.

EGR rate under steady state = (1 - KEGRMAP)

The steady-state EGR rate and the correction coefficient under a steady state are sometimes referred to as the "basic EGR rate" and the "basic correction coefficient", respectively. As mentioned above, the EGR rate is sometimes referred to as the "net EGR rate" to distinguish it from the EGR rate under a steady state. The correction coefficient under a steady state KEGRMAP has been determined in advance by experiments with respect to the engine speed Ne and the manifold pressure Pb, and is prepared in the form of map data as shown in Fig. 22 so that the value can be retrieved based on the parameters.

Here the EGR rate (exhaust gas recirculation rate) is explained again.

The EGR rate is used in various ways in references, such as as:

1) Mass of recirculated exhaust gas/mass of intake air and fuel;

2) Volume of recirculated exhaust gas/volume of intake air and fuel;

3) Mass of the recirculated exhaust gas 1 Mass of the intake air and the recirculated exhaust gas.

The EGR rate is mainly used in the description under the definition of 3). More specifically, the steady-state EGR rate is obtained by (1 - The KEGRMAP coefficient is specifically determined as a value indicating:

Fuel injection amount under EGR operation / Fuel injection amount under non-EGR operation

More specifically, the exhaust gas recirculation rate is determined by multiplying the basic EGR rate (the steady-state EGR rate) by the ratio between the gas flow rates as mentioned previously. As is clear from the description, since the EGR rate is determined as a value with respect to the basic EGR rate, the EGR rate estimation system according to the invention is applied to each EGR rate defined in 1) to 3) when the basic EGR rate is determined in the same manner.

The EGR control is carried out by determining a command value of the EGR control valve lift amount based on the engine speed, manifold pressure and the like, as shown in Fig. 21, the actual behavior of the EGR control valve lags behind the timing at which the command value is output. That is, there is a response delay between the actual valve lift and the output of the command value for executing it. In addition, it requires an additional period of time for the exhaust gas to pass through the valve and enter the combustion chamber.

The assignee has therefore proposed in Japanese Patent Application Hei 6(1994)-100,557 (filed in the United States on April 13, 1995 under No. 081421.191) the technique for determining the net EGR rate using the above-mentioned equation, i.e.

Net EGR rate = (steady state EGR rate) · {(gas flow rate QACT which is determined by the current valve lift amount and the ratio between the upstream pressure and the downstream pressure of the valve) / (gas flow rate QCMD which is determined by the command value and the ratio between the upstream pressure and the downstream pressure of the valve)}

In the art, the delay of the exhaust gas behavior has been assumed to be a first order delay. When the dead time is taken into account, it can be assumed that the exhaust gas flowing through the valve stays for a while in a space (chamber) in front of the combustion chamber and enters the combustion chamber all at once after a pause, i.e. the dead time. Thus, the net EGR rate is continuously estimated and stored in the memory each time the program is activated. Among the stored net EGR rates, one estimated at a previous control cycle according to the delay time is selected and is considered to be the true net EGR rate.

The operation of the system is described below with reference to the flow chart in Fig. 19. The program is activated at each TDC.

The program starts from S200 where the engine speed Ne, the manifold pressure Pb, the atmospheric pressure Pa and the current valve lift amount called LACT (the output of the sensor 123) are read in, and proceeds to S202 where the valve lift amount command value LCMD is retrieved from map data using the engine speed Ne and the manifold pressure Pb as address data. Similar to the above-mentioned correction coefficient, the map data for the command value LCMD is determined in advance with reference to the same parameters as shown in Fig. 23. The program then proceeds to S204 where the basic EGR rate correction coefficient KEGRMAP is retrieved from the map data using at least the engine speed Ne and the manifold pressure Pb as shown in Fig. 22.

The program then proceeds to S206 where it is confirmed that the current valve lift amount LACT is not equal to 0, that is, it is confirmed that the EGR control valve 122 is opened, and to S208 where the retrieved command value LCMD is compared with a predetermined lower limit LCMDLL (a minimum value) to determine whether the retrieved command value is smaller than the lower limit. If S208 determines that the retrieved command value is not smaller than the lower limit, the program proceeds to S210 where the ratio Pb/Pa between the manifold pressure Pb and the atmospheric pressure Pa is calculated using the calculated ratio and the retrieved command value LCMD, the gas flow rate QCMD corresponding thereto is retrieved from map data prepared in advance based on the characteristics shown in Fig. 20. The gas flow rate is the one mentioned in the equation as "gas flow rate QCMD determined by the command value and the ratio between upstream pressure and downstream pressure of the valve".

The program then proceeds to S212, where the gas flow rate QACT is retrieved from map data (whose characteristics are similar to those shown in Fig. 20) prepared in advance. This corresponds to the expression in the equation "gas flow rate QACT determined by the current valve lift amount and the ratio between upstream pressure and downstream pressure of the valve". The program then proceeds to S214, where the retrieved EGR rate correction coefficient KEGRMAP is subtracted from 1.0, and the resulting difference is taken as the steady-state EGR rate (basic EGR rate or steady-state EGR rate). The steady-state EGR rate refers to the EGR rate under which the EGR operation is in a steady state, i.e., the EGR operation is not in a transient state such as a high-pressure valve. B. when the operation is started or ended.

The program then proceeds to S216 where the net exhaust gas recirculation rate is calculated by multiplying the steady-state EGR rate by the ratio QACT/QCMD and to S218 where a fuel injection correction coefficient KEGRN is calculated.

Fig. 24 is a flowchart showing the subroutine for calculating the coefficient KEGRN.

In S300 in the flow chart, the net EGR rate (obtained in S216 of Fig. 19) is subtracted from 1.0, and the resulting difference is taken as the fuel injection correction coefficient KEGRN. The program then advances to S302, where the calculated coefficient KEGRN is stored in a ring buffer set up in the ROM 74. Fig. 25 shows the configuration of the ring buffer. As shown, the ring buffer has n addresses numbered from 1 to n and thus identified. Each time the programs of the flow charts of Figs. 19 and 24 are activated at the respective TDC positions and the fuel injection correction coefficient KEGRN is calculated, the calculated coefficient KEGRN is stored in the ring buffer in sequence from the top.

In the flowchart of Fig. 24, the program then advances to S304, where the delay time τ is retrieved from map data using the engine speed Ne and the engine load such as the manifold pressure Pb as address data. Fig. 26 shows the characteristics of the map data. That is, the delay time τ indicates a dead time during which the gas flowing through the valve remains in the space in front of the combustion chamber. Since the dead time changes with the engine operating conditions including the engine speed and the engine load, the delay time is set to change with the parameters. Here, the delay time τ is set as a ring buffer number.

The program then advances to S306, where, among the stored fuel injection correction coefficients KEGRN, the one corresponding to the recovered delay time τ (ring buffer number) is read out and set as the correction coefficient KEGRN in the current control cycle. Explaining this with reference to Fig. 27, when the current control cycle (or period) is A, for example, the coefficient calculated 12 control cycles earlier is selected as the coefficient to be used in the current control cycle.

Considering this from the EGR control valve operation, the correction coefficient KEGRN corresponding to the EGR rate calculated 12 control cycles earlier was 1.0, which means that the EGR control valve was closed. The value KEGRN then gradually decreases, such as 0.99, 0.98, ..., that is, the EGR control valve is gradually driven in the opening direction and reaches the current position at point A. In this example, it is assumed that the EGR gas has reached the combustion chamber at time A, so no correction is made to correct the On the other hand, when the correction is performed, the basic fuel injection amount TiM-F is multiplied by the correction coefficient KEGRN to decrease it.

As shown in Fig. 19, when S206 determines that the current valve lift amount LACT is equal to 0, it means that no EGR operation is being performed. However, since the correction coefficient KEGRN at this time will be a candidate for selection in a later control cycle, the program advances to S214, etc. to calculate the net EGR rate and the correction coefficient KEGRN. More specifically, in such a case, the net EGR rate is calculated to be 0 in S216, and the fuel injection correction coefficient KEGRN is calculated to be 1.0 at S300 in Fig. 24.

If it is determined in S208 that the valve lift command value LCMD is smaller than the lower limit LCMDLL, the program advances to S222 where the command value LCMDk - 1 from the last control cycle k - 1 is used.

This is because when the valve lift amount command value LCMD is made equal to 0 to terminate the EGR operation, the current valve lift amount LACT does not immediately become equal to 0 due to the delay in the valve response. Thus, when the command value LCMD is smaller than the lower limit, the previous value LCMDk - 1 is held until S206 determines that the current valve lift amount LACT has become equal to 0.

In addition, when the command value LCMD is smaller than the lower limit LCMDLL, the command value may occasionally be 0. When this occurs, the gas flow rate QCMD retrieved in S210 becomes 0, as a result of which division by 0 would occur in the calculation in step S216, making the calculation impossible. However, since the previous value is held in S222, the calculation in S216 can be successfully carried out.

The program then advances to S224, where the basic correction coefficient KEGRMAPk - 1 recovered in the last control cycle is used again in the current control cycle. This is because under In such engine operating conditions where the command value LCMD retrieved in S202 is smaller than the lower limit LCMDLL, the basic EGR rate correction coefficient KGERMAP retrieved in step S14 based on the map data becomes 1.0. As a result, there is a possibility that the steady-state EGR rate is determined to be 0 in S204. Holding the last value in S224 aims to avoid this.

As mentioned above, the net EGR rate is continuously estimated based on the engine speed and engine load such as manifold pressure, and the coefficient is continuously calculated based on this and stored at each control cycle. The delay time during which the exhaust gas passes through the valve but remains before the combustion chamber is determined using the same parameters, and one of the stored coefficients calculated at a previous control cycle corresponding to the delay time is selected as the coefficient in the current control cycle. This system reduces complicated calculations and significantly reduces calculation uncertainties, making its configuration easier, and can accurately estimate the net EGR rate and enable the fuel injection amount to be corrected with high accuracy.

In the foregoing, it should be noted that it is alternatively possible to store the net EGR rate in the ring buffer instead of KEGRN. Furthermore, the dead time can be a fixed value. Since this has been described in more detail in Japanese Patent Application Hei 6(1994)-294,014 (filed in the United States on April 13, 1995 under No. 08/421,182), it will not be discussed further here.

Next, the determination of the tank flushing correction coefficient KPUG (in response to the flushing mass) is explained.

The canister purge is carried out in a program, the flow chart of which is not shown, so that a desired amount of canister purge is determined in response to the engine operating conditions, such as the engine speed and the engine load, according to predetermined characteristics, wherein the above-mentioned flushing control valve 25 is controlled so that the desired degree of container flushing is achieved.

When the canister purge is active, the air/fuel ratio deviates to the rich side because evaporative gas is introduced into the air intake system with fuel. The deviation is corrected in the feedback loop. However, since the air/fuel ratio is expected to deviate to the rich side at the time of the canister purge, the fuel injection amount is preferably corrected in advance by the amount (called KPUG) corresponding to the purge fuel mass so that the correction amount in the feedback system decreases, thereby reducing the calculation load in the calculation loop, improving stability against disturbances, and improving the tracking accuracy.

The correction is performed by calculating the amount of fuel in the purged canister gas based on the flow rate and the HC concentration of the introduced purge gas. Alternatively, it may be performed by determining the correction coefficient KPUG corresponding to the purge mass from the difference of the LAF sensor output with respect to the target air-fuel ratio. The latter method is used in this embodiment.

Fig. 28 is a flow chart showing the coefficient determination.

The program starts at S400 where the flow rate of the purged gas is detected from the output of the above-mentioned flow meter 226, and proceeds to S402 where the HC concentration is detected from the output of the above-mentioned HC concentration sensor, to S404 where the amount (mass) of fuel introduced by the canister purge is determined, and to S406 where the determined amount of fuel is converted to the amount of gasoline fuel. A major part of the fuel component in the canister purge gas is butane, which is a light component of gasoline. Since the stoichiometric air/fuel ratio is different for butane and gasoline, the determined amount is recalculated for the amount of gasoline. The program then proceeds to S408 where the basic fuel amount TiM-F obtained by map retrieval is multiplied by the target air/fuel ratio to determine the amount of air Gc introduced into the cylinders, and on the basis of the value Gc and the amount of gasoline converted, the correction coefficient KPUG is calculated according to the purge mass. Of course, the correction coefficient KPUG is equal to 1.0 if the canister purge is not carried out.

Alternatively, it is possible to set the correction coefficient KPUG in advance, for example, to 0.95 in response to the desired amount of canister purge determined in the engine operating conditions, and to control the purge control valve 225 in response to the correction coefficient.

Alternatively, in the above it is possible to determine the correction coefficient KPUG based on an error between the actual air/fuel ratio and the target air/fuel ratio.

Alternatively, in the foregoing, it is possible to prepare the cylinder intake air quantity Gc as map data in advance to retrieve it based on the engine speed and the engine load.

Alternatively, it is possible to subtract the amount of fuel converted in terms of gasoline (S406) from the required fuel injection amount Tcyl.

The correction coefficient KTOTAL is a general name which is the product of various correction coefficients including KEGR and KPUG. The value additionally includes a correction coefficient KTW for the coolant temperature and a correction coefficient KTA for the intake air temperature and the like. However, since the nature of these corrections is well known, a detailed explanation is omitted.

The basic fuel injection quantity TiM-F is multiplied by the correction coefficient KTOTAL (= KEGR · KPUG · KTW · KTA ...) thus obtained in order to correct it.

Next, the target air/fuel ratio KCMD and the target air/fuel ratio correction coefficient KCMDM are determined or calculated.

Fig. 29 is a flow chart showing the investigation.

The program starts at S500, where the above-mentioned base value KBS is determined. This is accomplished by retrieving the map data (the characteristics of which are shown in Fig. 14) from the detected engine speed Ne and the manifold pressure Pb. The map data includes a base value at engine idling. When the fuel metering control includes the lean burn control, i.e., a lean mixture is supplied at a low engine load to improve fuel economy, the map data includes that for the lean burn control.

The program then advances to S502 where it is determined from a timer value whether lean burn control is being carried out after engine start to determine a lean correction coefficient. The system according to the invention is equipped with the variable control mechanism 300 which allows lean burn control after engine start in which, for a predetermined period after engine start, the target air/fuel ratio is set leaner than the stoichiometric air/fuel ratio while an intake valve is kept at rest during the period. Supplying a rich mixture for a period after engine start during which the catalyst remains inactivated would disadvantageously increase emissions of HG in the exhaust gas. However, lean burn control after engine start can avoid this problem.

If a target air/fuel ratio is set leaner in an engine without the variable valve timing mechanism, combustion becomes unstable and misfires sometimes occur. The engine 10 with the mechanism shown in Fig. 1 is capable of keeping one of two intake valves at rest, which creates a vortex of intake air, referred to as "swirl", and inhibits combustion even immediately after the engine start, making it possible to set a lean target air/fuel ratio for that period. Therefore, the timer value counting the period is read in S502 to determine whether the lean burn control period is present after the engine start and to determine the lean correction coefficient. If the result of the step is positive, the coefficient is determined to be 0.89, while if the result is negative, it is determined to be 1.0, for example.

The program then proceeds to S504 where it is determined whether the throttle opening is equal to the full throttle opening (WOT) and calculates a full throttle opening enrichment correction coefficient, proceeds to S506 where it is determined whether the coolant temperature Tw is high and calculates an enrichment correction coefficient KTWOT. The value KTWOT contains a correction coefficient for protecting the engine when the coolant temperature is high.

The program then advances to S508, where the base value KBS is multiplied by the correction coefficient to correct it, and the target air/fuel ratio KCMD is determined. This is determined by first setting a window (the catalyst window mentioned above) called DKCMD - OFFSET for the accurate air/fuel ratio control (the MIDO₂ control mentioned above) within a range in which the outputs of the O₂ sensor 56 have a linear characteristic in the vicinity of the stoichiometric value, as shown by dashed lines in the ordinate of the graph shown in Fig. 7, and then adding the value DKCMD - OFFSET to the base value KBS. More specifically, the target air/fuel ratio KCMD is determined as follows:

KCMD = KBS + DKCMD - OFFSET

The program then proceeds to S510 where the target air/fuel ratio KCMD(k) is limited to a predetermined range and to S512 where it is determined whether the calculated target air/fuel ratio KCMD(k) is equal to 1.0 or thereabouts. If the result is affirmative, the program proceeds to S514 where it is determined whether the O₂ sensor 56 is activated. This is done in a subroutine not shown by detecting the change in the output voltage called VO₂M of the O₂ sensor 56. The program then proceeds to S516 to calculate a value DKCMD for the MIDO₂ control. This calculation means to make the target air/fuel ratio for the LAF sensor 54 upstream of the O₂ sensor 56 provided downstream of the first catalyst 28 (downstream of the first catalyst bed in the case of the configuration shown in Fig. 5) variable. More specifically, this is accomplished by calculating the value from an error between a predetermined reference voltage VrefM and the O₂ sensor output voltage VO₂M using the PID control as shown in Fig. 7. The reference voltage VrefM is determined in response to the atmospheric pressure Pa, the coolant temperature Tw and the exhaust gas volume (which can be determined in response to the engine speed Ne and the manifold pressure Pb).

Here, the above-mentioned value DKCMD - OFFSET for the window setting is an offset value necessary for the first and second catalysts 28, 30 to maintain optimum purification efficiency. Since the offset values are different depending on the property or characteristics of a catalyst, the values are determined taking into account the property of the first catalyst 28. In addition, since the values vary as the catalyst ages, the values are updated via a learning control by obtaining weighted averages using the periodically calculated value DKCM.

More precisely, the values are calculated as follows:

DKCMD - OFFSET(k) = W DKCMD + (1 - W) DKCMD - OFFSET(k - 1)

where W denotes a weight.

By obtaining a learning control value by calculating the weighted average between the DKCMD calculated in the current cycle and the DKCMD - OFFSET calculated one cycle before, it is thus possible to carry out the control in such a way that the target air/fuel ratio converges to the air/fuel ratio that maximizes the purification efficiency without being influenced by the aging of the catalyst. This learning control can be performed under respective engine operating conditions defined by the engine speed Ne and the manifold pressure Pb and the like.

The program then advances to S518 where the calculated value DKCMD(k) is added to the target air-fuel ratio to update it, and to S520 where a table (the nature of which is shown in Fig. 30) is searched using the updated air-fuel ratio KCMD(k) as address data to retrieve a correction coefficient KETC. Since the charging efficiency of the intake air changes with the heat of vaporization, this is done to compensate for it. More specifically, the target air-fuel ratio KCMD(k) is multiplied by the correction coefficient KETC as shown to obtain the above-mentioned target air-fuel ratio correction coefficient KCMDM(k).

In other words, the target air/fuel ratio is actually expressed by the equivalence ratio, and the target air/fuel ratio correction coefficient is determined by applying the supercharging efficiency correction thereto. If the result in S512 is negative, the program jumps to S520 because this means that the target air/fuel ratio KCMD(k) deviates significantly from the stoichiometric air/fuel ratio such as in the lean-burn control and because it is not necessary to perform the MIDO₂ control. The program finally advances to S522 where the target air/fuel ratio correction coefficient KCMDM(k) is limited to a predetermined range.

As shown in the block diagram of Fig. 8, the basic fuel injection amount TiM-F is multiplied by the target air-fuel ratio correction coefficient KCMDM and the other correction coefficient KTOTAL to obtain the required fuel injection amount Tcyl.

Next, the feedback correction coefficients, such as KSTR, are calculated or determined.

Before explaining the calculation, the sampling of the LAF sensor outputs and the observer will be explained. The sampling block is labeled "Sel-V" in Fig. 8.

The sampling blocks and the observer are explained below.

In an internal combustion engine, during the exhaust stroke in the individual cylinders, the burnt gas is exhausted. Observation of the air/fuel ratio behavior at the exhaust system confluence point clearly shows that it changes in synchronism with the TDC. Thus, the sampling of the air/fuel ratio using the above-mentioned LAF sensor 54 installed in the exhaust system must be carried out in synchronism with the TDC. However, depending on the sampling timing of the control unit (ECU) 34 for processing the detection output, it may become impossible to accurately determine the air/fuel ratio. For example, if the air/fuel ratio at the exhaust system confluence point changes with respect to the TDC as shown in Fig. 31, the air/fuel ratio determined by the control unit may become a completely different value depending on the sampling timing, as shown in Fig. 32. Therefore, sampling is preferably carried out at positions that allow current changes in the output of the air/fuel ratio sensor to be determined as accurately as possible.

In addition, the actual air/fuel ratio varies depending on the time required for the exhaust gas to reach the sensor and the sensor response time (detection delay). The time required for the exhaust gas to reach the sensor varies with the exhaust pressure, the exhaust volume, and the like. In addition, since sampling in synchronism with TDC means that sampling is based on the crank angle, an effect of the engine speed is unavoidable. It is thus clear that air/fuel ratio detection is very dependent on the engine operating condition. In the prior art disclosed in Japanese Patent Application Laid-Open Hei 1 (1989)-313,644, it was therefore practiced to determine the adequacy of detection once after each prescribed crank angle. Since this requires a complex configuration and a requires a long calculation time, it may not be able to keep pace at high engine speeds, and this also presents the problem that the sensor output has already passed its inflection point by the time the decision to sample is made.

Fig. 33 is a flow chart of the operations for sampling the LAF sensor. However, since the accuracy of air-fuel ratio detection has a particularly close relationship with the estimation accuracy of the above-mentioned observer, the estimation of the air-fuel ratio by the observer will be briefly explained before explaining this flow chart.

For highly accurate separation and extraction of the air/fuel ratios of each cylinder at the output of a single LAF sensor, it is first necessary to accurately determine the detection response delay (lag time) of the LAF sensor. This delay was therefore modulated as a first order delay system to obtain the model shown in Fig. 34. Here, defining LAF: LAF sensor output and A/F: input A/F, the equation of state can be written as

LÅF(t) = αLAF(t) - αA/F(t) Eq. 9

Discretizing this equation for the period of Delta-T gives

LAF(k + 1) = LAF(k) + (1 - )A/F(k) Eq. 10

Here is the correction coefficient and is defined as:

= 1 + αΔT + (1/2!)α²ΔT² + (1/3!)α³ΔT³ + (1/4!)α&sup4;ΔT&sup4;

Equation 10 is shown as a block diagram in Fig. 35.

Thus, equation 10 can be used to obtain the current air/fuel ratio from the sensor output. That is, since equation 10 leads to Equation 11 can be rewritten, the value at time k - 1 can be calculated back from the value at time k, as shown in Equation 12.

A/F(k) = {LAF(k + 1} - LAF(k)}/(1 - ) Eq. 11

A/F(k - 1) = {LAF(k) - LAF(k - 1)}/(1 - ) Eq. 12

More specifically, applying the Z-transform to express Equation 10 as a transfer function yields Equation 13, where a real-time estimate of the air/fuel ratio input in the previous cycle can be obtained by multiplying the sensor output LAF of the current cycle by the inverse of this transfer function. Fig. 36 is a block diagram of the real-time A/F estimator.

t(z) = (1 - )/(Z - ) Eq. 13

The separation and extraction of the air/fuel ratios of the individual cylinders using the actual air/fuel ratio obtained in the foregoing manner is explained below. As explained in a previous application proposed by the assignee and filed in the United States on December 24, 1992 under Serial No. 07/997,769, the air/fuel ratio at the exhaust system confluence point can be taken as a weighted average to reflect the temporal distribution of the air/fuel ratios of the individual cylinders. This enables the air/fuel ratio at the confluence point at time k to be expressed in the manner of Fig. 14. (Since F (fuel) is selected as the controlled variable, the air-fuel ratio F/A is used here. However, for ease of understanding, the air-fuel ratio is used in the explanation as long as such use does not lead to confusion. The term "air-fuel ratio" (or "fuel-air ratio") used here is the current value corrected for the response delay calculated according to Equation 13.)

[F/A](k) = C1 [F/A#1 ] + C2 [F/A#3 ] + C3 [F/A#4 ] + C4 [F/A#2 ]

[F/A](k + 1) = C1 [F/A#3 ] + C2 [F/A#4 ] + C3 [F/A#2 ] + C4 [F/A#1 ]

[F/A](k + 2) = C1[F/A#4] + C2[F/A#2] + C3[F/A#1] + C4[F/A#3]... Eq . 14

More precisely, the air/fuel ratio at the confluence point can be expressed as the sum of the products of the past firing histories of the respective cylinders and the weighting coefficients Cm (e.g. 40% for the cylinder that fired last, 30% for the one immediately before it, etc.). This model can be represented as a block diagram as shown in Fig. 37.

Its equation of state can be written as

Furthermore, if the air/fuel ratio at the confluence point is defined as y(k), the output equation can be written as

The following applies:

c1 : 0.05, c2 : 0.15, c3 : 0.30, c4 : 0.50

Since u(k) cannot be observed in this equation, even if an observer is developed from the equation, it is not possible to observe x(k). Thus, assuming a steady state operation in which no sudden change in the air/fuel ratio occurs from that four TDCs earlier (ie, from that of the same cylinder), if x(k + 1) = x(k - 3) is defined, Equation 17 is obtained.

The simulation results for the model obtained in the foregoing manner are shown below. Fig. 38 refers to the case where the fuel is supplied to three cylinders of a four-cylinder internal combustion engine so as to achieve an air/fuel ratio of 14.7:1 and is supplied to one cylinder so as to achieve an air/fuel ratio of 12.0:1. is reached. Fig. 39 shows the air/fuel ratio at this time at the confluence point as obtained using the above-mentioned model. Although Fig. 39 shows that a stepped output is obtained when the response delay of the LAF sensor is taken into account, the sensor output becomes equal to the smoothed wave labeled "delay-adjusted model output" in Fig. 40. The curve labeled "actual sensor output" is based on the actually observed output of the LAF sensor under the same conditions. The close agreement of the model results with this confirms the validity of the model as a model of the exhaust system of a multi-cylinder internal combustion engine.

Thus, the problem reduces to that of an ordinary Kalman filter where x(k) is observed in the state equation (equation 18) and the output equation. When the weighting parameters Q, R are determined, when equation 19 and the Riccati equation are solved, the gain matrix K becomes as shown in equation 20.

The following applies:

Obtaining A-KC from this gives Equation 21.

Fig. 41 shows the configuration of an ordinary observer. However, since there is no input u(k) in the present model, the configuration has only y(k) as input, as shown in Fig. 42. This is expressed mathematically by Equation 22.

The system matrix of the observer whose input is y(k), i.e. the Kalman filter, is

If in the present model the ratio of the element of the weighting parameters R in the Riccati equation to the element of Q is 1 : 1, the system matrix S of the Kalman filter is given by

Fig. 43 shows the above-mentioned model and the observer combined. As a result of the simulation they are shown in the previous application, but are omitted here. Suffice it to say that this enables an accurate estimation of the air/fuel ratios at the individual cylinders from the air/fuel ratio at the confluence point.

Since the observer is capable of estimating the air/fuel ratio cylinder by cylinder from the air/fuel ratio at the confluence point, the air/fuel ratios of the individual cylinders can be separately controlled by the PID control or the like. More specifically, as shown in Fig. 44, in which the feedback section of the observer of Fig. 35 is extracted and shown by itself, a confluence point feedback correction coefficient KLAF is calculated from the sensor output (confluence point air/fuel ratio) and the target air/fuel ratio using the PID control law, and cylinder-by-cylinder feedback correction coefficients #nKLAF (n: affected cylinder) are calculated from the observer estimated air/fuel ratio #nA/F.

More precisely, the cylinder-wise feedback correction coefficients #nKLAF is obtained using the PID law to eliminate the error between the observer-estimated air/fuel ratio #nA/F and the target value obtained by dividing the confluence point air/fuel ratio by the average of the cylinder-wise feedback correction coefficients #nKLAF calculated in the previous cycle.

Due to this convergence of the air/fuel ratios of the individual cylinders toward the confluence point air/fuel ratio and the convergence of the confluence point air/fuel ratio toward the target air/fuel ratio, the air/fuel ratios of all cylinders converge toward the target air/fuel ratio. The initial fuel injection amount #nTout (n: affected cylinder) is determined from the fuel injection device opening period and can be calculated as

#nTout = Tcyl · KCMD · #nKLAF · KLAF.

Since the foregoing is disclosed in Japanese Patent Publication 07-083,138 (filed in the United States on September 13, 1994 under No. 081305.162) proposed by the assignee, no further explanation follows.

Sampling of the LAF sensor output is explained below with reference to the flow chart of Fig. 33. This subroutine is activated at TDC.

The subroutine of the flow chart of Fig. 33 starts at S600 where the engine speed Ne, the manifold pressure Pb and the valve timing V/T are read. The program then proceeds to S604 and S606 where Hi and Lo valve timing maps (which will be explained later) are searched and to S608 where the sensor output is sampled for use in the observer calculation at Hi or Lo valve timing. More specifically, the timing map is searched using the detected engine speed Ne and the manifold pressure Pb as address data, with the number of one of the above-mentioned twelve buffers is selected and the sample value stored in it is selected.

Fig. 45 shows the characteristics of the timing maps. As shown, the characteristics are defined such that the sampling crank angle of the selected value is advanced as the engine speed Ne decreases and the manifold pressure (load) Pb increases. By an "earlier" value, is meant a relatively older value sampled closer to the previous TDC. In contrast, the characteristics are defined such that the sampling crank angle of the selected value is retarded (becomes a newer value closer to the following TDC) as the engine speed Ne increases and the manifold pressure Pb decreases.

It is best to sample the LAF sensor output as close as possible to the inflection point of the actual air/fuel ratio, as shown in Fig. 32. Assuming that the sensor response time (detection lag) is constant, this inflection point, or e.g. its first peak, as shown in Fig. 46, occurs at progressively earlier crank angles as the engine speed decreases. As the engine load increases, the exhaust gas can be expected to increase in pressure and volume, and thus can reach the sensor earlier due to its higher flow rate. This is why the selection of sampled data is determined as shown in Fig. 45.

The valve control is described below. By defining an arbitrary engine speed on the Lo side as Ne1-Lo and on the Hi side as Ne1-Hi and an arbitrary manifold pressure on the Lo side as Pb1-Lo and on the Hi side as Pb1-Hi, the values are mapped so that

Pb1-Lo > Pb1-Hi, and

Ne1-Lo > Ne1-Hi.

In other words, since the time at which the exhaust valve opens is earlier in HiV/T than in LoV/T, the map properties are determined such that an earlier sampling point is selected in HiV/T than in LoV/T, provided the engine speed and manifold pressure are the same.

The program then proceeds to S610 where the observer matrix is calculated for HiV/T and to S612 where the calculation is performed similarly for LoV/T. It then proceeds to S604 where the valve timing is again determined and, depending on the result of the determination, to S616 where the calculation result for HiV/T is selected or to S618 where that for LoV/T is selected. The routine ends.

In other words, since the behavior of the confluence point air/fuel ratio also changes with the valve timing, the observer matrix must change in synchronization with the switching of the valve timing. However, the estimation of the air/fuel ratios at the individual cylinders is not performed immediately. Since the observer calculation requires several cycles to converge, the calculations using the observer matrices before and after the valve timing switching are performed in parallel, with one of the calculation results corresponding to the new valve timing being selected in S614 even if the valve timing is changed. After the estimation is performed for the individual cylinders, the feedback correction coefficient is calculated to eliminate the error from the target value and the fuel injection amount is determined.

The above-mentioned configuration improves the accuracy of air/fuel ratio detection. As shown in Fig. 47, since sampling is performed at relatively short intervals, the sampled values reliably reflect the sensor output, and the values sampled at relatively short intervals are progressively stored in the group of buffers. The inflection point of the sensor is predicted from the engine speed and manifold pressure, and the corresponding value is selected from the group of buffers at the prescribed crank angle. The observer calculation is then performed to estimate the air/fuel ratios at the individual cylinders, thus enabling the cylinder-by-cylinder control to be carried out as with Reference to Fig. 44 has been explained.

The CPU core 70 can thus accurately determine the maximum and minimum values of the sensor output as shown in Fig. 47 below. As a result, the estimation of the air/fuel ratios of the individual cylinders using the above-mentioned observer can be performed using the values that approximate the behavior of the actual air/fuel ratio, thereby enabling an improvement in accuracy when the cylinder-by-cylinder air/fuel ratio control is performed in the manner described with reference to Fig. 44.

In the foregoing, it should be noted that the sampling can be performed for both HiV/T and LoV/T and then the detection can be performed for the first time at which the timing is selected.

It should also be noted that since the LAF sensor response time when the air-fuel ratio is lean is shorter than when the air-fuel ratio is rich, the earlier sampled data is preferably selected when the air-fuel ratio to be detected is lean.

Furthermore, since the exhaust pressure drops due to the reduction in atmospheric pressure at high altitude, the exhaust gas arrives at the LAF sensor in a shorter period of time than at low altitude. As a result, the data sampled earlier are preferentially selected as the altitude of the place where the vehicle is traveling increases.

Furthermore, since the sensor response time becomes longer as the sensor deteriorates, the data sampled earlier are preferentially selected as the sensor deterioration increases.

Since this has been explained in a previous Japanese patent application Hei 6(1994)-243.277 of the assignee, it will not be further described here.

The feedback correction coefficient, such as KSTR, is explained below.

As disclosed in Fig. 44, the PID control law is usually used for fuel metering control for internal combustion engines. The control error between the setpoint and the manipulated variable (control input) is multiplied by a P term (proportional term), an I term (integral term), and a D term (differential or derivative term) to obtain the feedback correction coefficient (feedback gain). In addition, it has recently been proposed to obtain the feedback correction coefficient using modern control theory.

In the MIDO₂ control according to the invention, as mentioned previously, the feedback correction coefficient KSTR is calculated using an adaptive controller (self-tuning controller) instead of the confluence point feedback correction coefficient KLAF calculated using a PID controller as shown in Fig. 44. This dynamically ensures the response of the system from the target air-fuel ratio KCMD to the actual air-fuel ratio KACT, since the value KCMD becomes the smoothed value of KACT due to the engine response delay when the basic fuel injection amount determined in the controlled system is corrected only by the target air-fuel ratio feedback correction coefficient KCMDM. The correction coefficient KSTR is thus multiplied by the basic fuel injection amount and by the correction coefficient KCMDM.

However, when the feedback correction coefficient is determined using the modern control law such as the adaptive control law, since the control response in such cases is relatively high, it may become unstable under certain engine operating conditions due to a fluctuation or oscillation of the controlled variable, which affects the stability of the control. Furthermore, the fuel supply is cut off during driving and under certain other operating conditions, and it is controlled in an open loop (O/L) during the fuel cut-off period. controlled as shown in Fig. 48.

Subsequently, when fuel supply is continued to obtain a stoichiometric air-fuel ratio (14.7:1), for example, the fuel is supplied based on the fuel injection amount determined according to an empirically obtained characteristic. As a result, the true air-fuel ratio (AIF) jumps from the lean side to 14.7:1. However, a certain amount of time is required until the supplied fuel is burned and the burned gas reaches the air-fuel ratio sensor. In addition, the air-fuel ratio sensor has a detection delay time. As a result, the actual air-fuel ratio is not always the same as the true air-fuel ratio, but has a relatively large error as shown by the dashed line in Fig. 48.

At this time, once the high-control response feedback correction coefficient KSTR is determined based on an adaptive control law, the adaptive controller STR determines the feedback correction coefficient KSTR so that the error between the target value and the actual value is immediately eliminated. However, since this difference is caused by the sensor detection delay and the like, the detected value does not indicate the true air-fuel ratio. Nevertheless, since the adaptive controller absorbs the relatively large difference completely at once, KSTR fluctuates greatly as shown in Fig. 48, further causing the controlled variable to fluctuate or oscillate and deteriorating the control stability.

The occurrence of this problem is not limited to the time of resumption of fuel supply after a shutdown. It also occurs at the time of resumption of control after full load enrichment and when resumption of stoichiometric air-fuel ratio control after lean burn control. It also occurs when switching from perturbation control in which the target air-fuel ratio is intentionally varied to control using a fixed target air-fuel ratio. In other words, the problem occurs whenever a large Change in the target air/fuel ratio occurs.

Therefore, it is preferable to determine a feedback correction coefficient of the high control response using a control law such as the adaptive control law, and another feedback correction coefficient of the low control response using a control law such as the PID control law (denoted by KLAF in the figure), with one or the other of the feedback correction coefficients being selected depending on the engine operating state. However, since the different types of control laws have different characteristics, a sharp difference in level between the two correction coefficients may arise. Therefore, switching between the correction coefficients is responsible for destabilizing the controlled variable and deteriorating the control stability.

The system according to the invention is configured such that the feedback correction coefficients, which are different in control response, are determined using an adaptive control law and a PID control law, switching between them in response to the operating conditions of the engine, and smoothing the switching between the feedback correction coefficients, thereby improving the fuel metering and air/fuel ratio control performance while ensuring control stability.

Fig. 49 is a subroutine flowchart showing the determination or calculation of the feedback correction coefficients including KSTR.

For easier understanding, the adaptive controller STR is first explained with reference to Fig. 50. More specifically, the adaptive controller includes a controller called STR (Self-Tuning Controller) and an adaptation mechanism (controller (system) parameter estimator).

As previously mentioned, the target fuel injection amount Tcyl is determined based on the basic fuel injection amount in the controlled system, wherein the initial fuel injection amount Tout is determined based on the value Tcyl as explained later, and the controlled system (engine 10) via a fuel injector 22. The target air/fuel ratio KCMD and the controlled variable (actual air/air-fuel ratio) KACT (system output y) are input to the STR controller, which calculates the feedback correction coefficient KSTR using a recursion or repetition formula. In other words, the STR controller receives the coefficient vector (controller parameters expressed as a vector) adaptively estimated or identified by the adaptation mechanism and forms a feedback compensator.

An identification or adaptation law (algorithm) available for adaptive control is the one proposed by I. D. Landau et al. The adaptive control system is nonlinear in characteristic, so a stability problem is inherent. In the adaptation law proposed by I. D. Landau et al., the stability of the adaptation law expressed in a recursion formula is ensured at least by using Lyapunov theory or Popov hyperstability theory. This method is described, for example, in Computrol (Corona Publishing Co., Ltd.), No. 27, pp. 28-41; Automaic Control Handbook (Ohm Publishing Co., Ltd.), pp. 703-707; "A Survey of Model Reference Adaptive Techniques - Theory and Applications" by I. D. Landau in Automatica, Vol. 10, pp. 353-379; “Unification of Discrete Time Explicit Model Reference Adaptive Control Designs” by I. D. Landau et al. in Automatica, Vol. 17, No. 4,. pp. 593-611; and "Combining Model Reference Adaptive Controllers and Stochastic Self-tuning Regulators" by I. D. Landau in Automatica, Vol. 18, No. 1, pp. 77-84.

The adaptation or identification algorithm of ID Landau et al. is used in the present system. In this adaptation or identification algorithm, when the polynomials of the denominator and the numerator of the transfer function B(Z⁻¹)/A(Z⁻¹) of the discrete controlled system are defined in the manner of Equation 25 and Equation 26 shown below, the controller parameters or (adaptive) system parameters (k) are formed from the parameters (dynamic engine characteristic parameters) as shown in Equation 27 and expressed as a vector (transposed vector). Further, the input Zeta(k) is equal to that shown by equation 28. Here, as an example, a plant is taken in which m = 1, n = 1 and d = 3, ie the plant model is given in the form of a linear system with three control cycles of dead time.

A(z-1) = 1 + a1 z-1 + ... + anz-n Eq. 25

B(z&supmin;¹) = b&sub0; + b1 z-1 + ... + bmz-m Eq. 26

T(k) = [u(k), ..., u(k - m - d + 1), y(k), ..., y(k - n + 1))

= [u(k), u(k - 1), u(k - 2), u(k - 3), y(k)] Eq. 28

Here, the factors that constitute the STR controller, i.e., the scalar quantity ⊂0⊃min;¹(k) that determines the gain, the control factor R(Z⊃min;¹, k) that uses the manipulated variable, and (Z⊃min;¹, k) that uses the controlled variable, all shown in Equation 27, are expressed as in Equations 29 to 31, respectively. In the equation, "m" and "n" denote the denominator and numerator of the system, and "d" denotes the dead time. As mentioned above, a system in the form of a linear system with three control cycles of dead time is assumed as an example.

0-1(k) = 1/b0 Eq. 29

R(Z-1, k) = r1 z-1 + r2 z-2 + ... + rm+d-1z-(m+d-1) = r1 z-1 + r2 z-2 + r3 z-3 Eq. 30

(Z&supmin;¹, k) = s&sub0; + s1 z-1 + ... + sn-1z-(n-1) = s0 Eq. 31

The adaptation mechanism estimates or identifies each coefficient of the scalar quantity and the control factors and provides them to the STR controller.

When the coefficients in a group are expressed by a vector, the controller parameters are calculated by the following equation 32. In equation 32, 17(k) is a gain matrix (the (m + n + d)-th order square matrix) that determines the estimation/identification rate or speed of the controller parameters, while - Asterisk(k) is a signal indicating the generalized estimation/identification error, an estimation error signal of the controller parameters. These are represented by recursion formulas such as those of equations 33 and 34.

(k) = (k - 1) + Gamma;(k - 1) (k - d)e* (k) Eq. 32

As mentioned previously, the adaptation mechanism estimates or identifies each of the controller parameters (vector) using the manipulated variable u(i) and the controlled variable y(j) of the plant (i, j contain past values) so that an error between the setpoint and the controlled variable becomes equal to 0.

There are different specific algorithms given depending on the selection of lambda 1 and lambda 2 in equation 33. Lambda 1(k) = 1, lambda 2(k) = lambda (0 < lambda < 2) gives the gradually decreasing gain algorithm (least squares method when lambda = 1); and lambda 1(k) = lambda 1 (0 < lambda 1 < 1), lambda 2(k) = lambda 2(0 < lambda 2 < lambda) gives the varying gain algorithm (weighted least squares method when lambda 2 = 1). Further, by defining lambda 1(k)/lambda 2(k) = σ and representing Lambda 3 as in equation 35, the constant track algorithm is obtained by defining Lambda 1(k) = Lambda 3(k). In addition, Lambda 1(k) = 1, Lambda 2(k) = 0 gives the constant gain algorithm. As is clear from equation 33, in this case Γ(k) = Γ(k - 1), which leads to the constant value Γ(k) = Γ. Each of the algorithms is suitable for the time-variant system such as the fuel metering control system according to the invention.

Thus, the adaptive controller (adaptive controller means) is a controller expressed in a recursion formula so that the dynamic behavior of the controlled object (engine) can be ensured. More precisely, it can be defined as the controller provided at its input with the adaptation mechanism (adaptation mechanism means), more precisely the adaptation mechanism expressed in a recursion formula.

The feedback correction coefficient KSTR(k) is specifically calculated as shown in Equation 36:

The adaptive correction coefficient KSTR thus obtained is compared with the target Fuel injection quantity is multiplied by a feedback correction coefficient (general name of the coefficient KSTR and others determined by a PID control law) to determine the initial fuel injection quantity Tout, which is then supplied to the controlled system (engine). More specifically, the initial fuel injection quantity Tout is calculated as follows:

Tout = TiM-F · KCMDM · KFB · KTOTAL + ITOTAL

In the foregoing, TTOTAL denotes the total value of the various corrections for atmospheric pressure and the like performed by addition expressions (but does not include the injector dead time and the like which is separately added at the time of output of the initial fuel injection amount Tout.)

What characterizes Fig. 50 (and Fig. 8) is, first, that the STR controller is placed outside the fuel injection amount calculation system (the above-mentioned controlled system) and that not the fuel injection amount but the air/fuel ratio is defined as the target value. In other words, the manipulated variable is given in terms of the fuel injection amount, and the adaptation mechanism operates to determine the feedback correction coefficient KSTR so as to bring the air/fuel ratio produced as a result of the fuel injection in the exhaust system to the target value, thereby increasing the robustness to disturbances. Since this has been described in the assignee's Japanese Patent Application No. Hei 6(1994)-66,594 (filed in the United States on March 9, 1995 under No. 08/401,430), it will not be explained in detail here.

A second characteristic feature is that the manipulated variable is determined as the product of the feedback correction coefficient and the basic fuel injection quantity. This leads to a significant improvement in the control convergence. On the other hand, the configuration has the disadvantage that the controlled variable tends to fluctuate if the manipulated variable is determined inappropriately. A third characteristic feature is that a conventional PID controller is provided in addition to the STR controller to provide another feedback correction coefficient called KLAF. based on the PID control law, one of which is selected by means of a switch as the final feedback correction coefficient KFB.

More specifically, the actual value KACT(k) and the target value KCMD(k) are also input to the PID controller, which calculates the PID correction coefficient KLAF(k) based on the PID control law so as to eliminate the control error between the actual value at the exhaust system confluence point and the target value. One or the other of the feedback correction coefficient KSTR obtained by means of the adaptive control law and the PID correction coefficient KLAF obtained by using the PID control law is selected by means of a switching mechanism shown in the figure to be used in determining the fuel injection amount.

Next, the calculation of the PID correction coefficient is explained. First, the control error DKAF between the target air/fuel ratio KCMD and the actual air/fuel ratio KACT is calculated as:

DKAF(k) = KCMD(k - d') - KACT(k).

In this equation, KCMD(k - d') is the target air/fuel ratio (where d' indicates the dead time before KCMD is reflected in KACT, and thus denotes the target air/fuel ratio before the dead time control cycle), where KACT(k) is the actual air/fuel ratio (in the current control (program) cycle).

Then, the control error DKAF(k) is multiplied by specific coefficients to obtain variables, i.e. the P-term (proportional term) KLAFP(k), the I-term (integral term) KLAFI(k), and the D-term (differential or derivative term) KLAFD(k), as:

P-expression: KLAFP(k) = DKAF(k) · KP

I-expression: KLAFI(k) = KLAFI(k - 1) + DKAF(k) · KI

D-expression: KLAFD(k) = (DKAF(k) - DKAF(k - 1)) · KD.

Thus, the P term is calculated by multiplying the error by the proportional gain KP; the I term is calculated by adding the value of KLAFI(k - 1), the feedback correction coefficient in the previous control cycle (k - 1), to the product of the error and the integral gain KI; and the D term is calculated by multiplying the difference between the value of DKAF(k), the error in the current control cycle (k), and the value of DKAF(k - 1), the error in the previous control cycle (k - 1), by the differential gain KD. The gains KP, KI and KD are calculated based on the engine speed and the engine load. More specifically, they are taken from a map using the engine speed Ne and the manifold pressure Pb as address data. Finally, KLAF(k), the value of the feedback correction coefficient according to the PID control law in the current control cycle, is calculated by summing the values thus obtained:

KLAF(k) = KLAFP(k) + KLAFI(k) + KLAFD(k).

Note that the offset of 1.0 is assumed to be contained in the I-term KLAFI(k), so that the feedback correction coefficient is a multiplication coefficient (i.e., the I-term KLAFI(k) is given an initial value of 1.0).

It should also be noted that when the PID correction coefficient KLAF is selected for the fuel injection amount calculation, the STR controller holds the controller parameters so that the adaptive correction coefficient KSTR is equal to 1.0 (initial value) or close to 1.

Based on the foregoing, the determination or calculation of the feedback correction coefficient will be explained with reference to Fig. 49. The program is activated at every TDC.

In Fig. 49, the program starts at S700 where the detected engine speed Ne and the manifold pressure Pb and the like are read, and advances to S704 where it is checked whether the fuel supply has been cut off.

A fuel cut is implemented under specific engine operating conditions, such as when the throttle is fully closed and the engine speed is higher than a prescribed value, at which time the fuel supply is stopped and control is performed.

If it is determined in S704 that the fuel cut is not implemented, the program advances to S706 where it is determined whether the activation of the LAF sensor 54 is complete. This is accomplished by comparing the difference between the output voltage and the center voltage of the LAF sensor 54 with a prescribed value (e.g., 1.0 V) and determining that the activation is complete if the difference is less than the prescribed value.

When S708 determines that activation is complete, the program goes to S710 where it checks whether the engine operating condition is within the control range. This is done using a separate routine (not shown in the drawing). For example, if the engine operating condition has suddenly changed, such as during full load enrichment, high engine speed, EGR or the like, the fuel metering is not controlled with a closed loop but is controlled with an open loop.

If the result is positive, the program goes to S712 where the output of the LAF sensor is read, to S714 where the air/fuel ratio KACT(k) is determined or calculated from the output, and to S716 where the feedback correction coefficient KFB (the general name for KSTR and KLAF) is calculated. As mentioned previously, is used to denote a discrete variable in the description and the sample number in the discrete-time system.

The subroutine for this calculation is shown in the flow chart of Fig. 51.

First, S800 checks whether during the previous cycle (during the last control (calculation) cycle, ie during the previous routine activation time). If control was carried out during the fuel cut or the like in the previous cycle, the result in S800 is positive. In this case, a counter value C is reset to 0 in S102, the bit of a flag FKSTR is reset to 0 in S804, and the feedback correction coefficient KFB is calculated in S106. Resetting the bit of the flag FKSTR to 0 in S804 indicates that the feedback correction coefficient is to be determined by the PID control law. Furthermore, as will be explained below, setting the bit of the flag FKSTR to 1 indicates that the feedback correction coefficient is to be determined by the adaptive control law.

A subroutine showing the specific procedures for calculating the feedback correction coefficient KFB is shown in the flow chart of Fig. 52. In S900, it is checked whether the bit of the flag FKSTR is set to 1, i.e., whether the operating state is in the STR (controller) operating range. Since this flag was reset to 0 in S804 of the subroutine of Fig. 51, the result in this step is No, and in S902, it is checked whether the bit of the flag FKSTR was set to 1 in the previous control cycle, i.e., whether the operating state was in the STR (controller) operating range in the previous cycle.

Since the result here is of course No, the program goes to S904, where the PID correction coefficient KLAF(k) is calculated by the PID controller based on the PID control law in the manner described previously. More specifically, the PID correction coefficient KLAF(k) calculated by the PID controller is selected. Returning to the subroutine of Fig. 51 again, KFB is set to KLAF(k) in S808.

In the subroutine of Fig. 51, when it is determined in S800 that no control was carried out in the previous control cycle, ie, the control was continued after the control, the difference between KCMD(k - 1), the value of the set value in the previous control cycle, and the value of KCMD(k), the set value in the current control cycle, is calculated and compared with a reference value DKCMDref in S810. If it is determined that the difference DKCMD exceeds the reference value DKCMDref, the PID correction coefficient is calculated using the PID control law in S802 and the following steps.

This is because when the change in the target air/fuel ratio is large, a situation similar to that when the fuel cut is released occurs. More specifically, the actual value is likely to not indicate the true value due to an air/fuel ratio detection delay or the like, so that the controlled variable may become unstable similarly. A large change occurs in the target equivalence ratio, for example, when normal fuel supply is continued after a full load enrichment, when stoichiometric air/fuel ratio control is continued after a lean burn control (e.g., at an air/fuel ratio of 20:1 or leaner), and when stoichiometric air/fuel ratio control using a fixed target air/fuel ratio is continued after a perturbation control in which the target air/fuel ratio fluctuates.

On the other hand, if S810 determines that the difference DKCMD is equal to or smaller than the reference value DKCMDref, the counter value C is incremented in S812, and it is checked in S814 whether the engine coolant temperature Tw is smaller than a prescribed value TWSTR.ON. The prescribed value TWSTR.ON is set to a relatively low coolant temperature, and if the detected engine coolant temperature Tw is below the prescribed value TWSTR.ON, the program advances to S804, where the PID correction coefficient is calculated using the PID control law. The reason for this is that combustion is unstable at low coolant temperatures, making it impossible to obtain stable detection of the value KACT due to misfire and the like. Although not shown, the same applies even if the coolant temperature is abnormally high for the same reason.

If S814 determines that the engine coolant temperature Tw is not lower than the specified value TWSTR.ON, the program proceeds to S816 where it is checked whether the detected engine speed Ne is equal to or greater than a specified value NESTRLMT. The specified value NESTRLMT is set to a relatively high engine speed. When S816 determines that the detected engine speed Ne is equal to or greater than the specified value NESTRLMT, the program proceeds to S804 where the PID correction coefficient is calculated. This is because during high-speed engine operation, there tends to be insufficient time for calculation and, in addition, combustion is unstable.

If S816 determines that the detected engine speed Ne is less than the prescribed value NESTRLMT, the program advances to S818 where it is checked which valve timing is selected in the variable valve timing mechanism. If HiV/T is selected, the program advances to S804 where the PID correction coefficient is calculated. This is because the large amount of valve timing overlap that exists when the valve timing characteristic for high engine speeds has been selected causes intake air blow-by (leakage of intake air through the exhaust valve), in which case the actual value KACT is hardly stable. In addition, the detection delay of the LAF sensor during high speed operation cannot be ignored.

If S818 determines that LoV/T has been selected (this includes the condition in which one of the two intake valves is at rest), the program advances to S820 where it is checked whether the engine is idling. If the result is yes, the program proceeds to S804 where the PID correction coefficient is calculated. This is because the generally stable operating condition during idling eliminates the need for a high gain such as that according to the adaptive control law. Furthermore, the above-mentioned electric air control valve (EACV) 53 is controlled to control the intake air quantity. There is a possibility that the intake air control and the subject fuel metering control, if carried out, interfere with each other. This is another reason why the gain is set low during use of the PID correction coefficient.

If S820 determines that the engine is not idling, the program advances to S822 where it is judged whether the engine load is low. If the If the result is Yes, the program goes to S804 where the PID correction coefficient is calculated. This is because the combustion is not stable in the low engine load range.

If S822 determines that the engine load is not low, the counter value C is compared with a preset value, e.g. 5, in S824. As long as the counter value C is determined to be equal to or smaller than the preset value, the PID correction coefficient KLAF(k) calculated with the PID control law is selected by the procedures S804, S806, S900, S902 (S916), S904 and S908.

In other words, during the period from time T1 when the fuel cut is interrupted in Fig. 48 and the control is continued after a control (when C = 1 as mentioned in connection with Fig. 51) to time T2 (counter value C = 5), the feedback correction coefficient is set to the value KLAF determined by the PID controller using the PID control law. Unlike the feedback correction coefficients KSTR determined by the STR controller, the PID correction coefficient KLAF according to the PID control law does not absorb the control error DKAF between the set value and the actual value completely at once, but has a relatively gradual absorption characteristic.

Thus, as shown in Fig. 48, even if a relatively large difference occurs due to the delay until the fuel is completely burned after fuel supply is resumed and due to the LAF sensor detection delay, the correction coefficient does not become unstable as in the case of the STR controller, and therefore no instability of the controlled variable (system output) is caused. The predetermined value is set to 5 (i.e., five control cycles or TDCs (TDC = top dead center)) in this embodiment because this period is considered sufficient to absorb the combustion delay and the detection delay. Alternatively, the period (predetermined value) may be determined based on the engine speed, the engine load and other such factors that affect the exhaust gas transport delay parameters. For example, the predetermined value may be set small when the engine speed and manifold pressure produce a small exhaust gas transport delay parameter, and can be set large if they produce a large exhaust gas transport delay parameter.

Next, when S824 in the subroutine of Fig. 51 determines that the counter value C exceeds the above-described value, i.e., is equal to 6 or more, the bit of the flag FKSTR is set to 1 in S826, and the feedback correction coefficient KFB is calculated in accordance with the subroutine of Fig. 52 in S828. In this case, the result of the check in S900 of the subroutine of Fig. 52 becomes Yes, and it is checked in S906 whether the bit of the flag FKSTR was reset to 0 in the previous control cycle, i.e., whether the operating state in the previous cycle was in the PID operating range.

If this is the first time that the counter value exceeds the specified value, the result of this check is Yes, in which case the actual value KACT(k) is compared with a lower limit value of, for example, 0.95 in S908. If the actual value is equal to or greater than the lower limit value, the actual value is compared with an upper limit value of, for example, 1.0 in S910. If it is found to be equal to or less than the upper limit value, the program advances via S912 (explained later) to S914, where the adaptive correction coefficient KSTR(k) is calculated using the STR controller.

In other words, when S908 determines that the feedback value is below the lower limit value or S910 determines that the feedback value exceeds the upper limit value, the program goes to S904 where the feedback correction coefficient is calculated based on the PID control. In other words, switching from PID control to STR (adaptive control) control is performed when the motor operating condition is in the STR controller operating range and the feedback value KACT is equal to or close to 1. This enables switching from PID control to STR (adaptive control) control to be performed smoothly and prevents fluctuation of the controlled variable.

If S910 determines that the actual air/fuel ratio KACT(k) is equal to or smaller than the upper limit value, the program advances to S912 where, as shown, the above-mentioned scalar quantity b₀, the value determining the gain of the STR controller, is set to or replaced by the value obtained by dividing it by KLAF(k - 1), the value of the PID correction coefficient from the PID control in the previous control cycle, whereupon the feedback correction coefficient KSTR(k) determined by the STR controller is calculated in S914.

In other words, the STR controller basically calculates the feedback correction coefficient KSTR(k) according to Equation 36 as previously explained. However, if the result in S906 is positive and the program advances to S908 and the subsequent steps, it means that the feedback correction coefficient has been determined based on the PID control in the previous control cycle. In addition, as previously explained with reference to the configuration of Fig. 50, the feedback correction coefficient KSTR is fixed to 1, and the STR controller operation is kept suspended when the feedback correction coefficient is determined by means of PID control. In other words, the vector of controller parameters to be used in the STR controller is determined such that KSTR = 1.0. Therefore, when the STR controller continues to determine the feedback correction coefficient KSTR, the controlled variable becomes unstable when the value of KSTR significantly deviates from 1, thereby making the controlled variable unstable.

In view of this, the scalar quantity b₀ (in the controller parameters held by the STR controller so that the adaptive correction coefficient KSTR is fixed at 1.0 (initial value) or around it) is divided by the value of the feedback correction coefficient from the PID controller in the previous control cycle. Thus, as is clear from Equation 37, since the first term is equal to 1, the value of the second term KLAF(k - 1) becomes equal to the correction coefficient KSTR(k) of the current control cycle, provided that the controller parameters are held such that KSTR = 1.0 as just mentioned:

As a result, the actual value KACT in S908 and S910 is equal to 1 or close to 1, and in addition, switching from PID control to STR control can be carried out smoothly.

In the subroutine of Fig. 52, when S902 determines that the engine operating condition was in the STR (controller) operating range in the previous control cycle, the value of KSTR(k - 1), the adaptive correction coefficient in the previous control cycle, is set to or replaced by the value of KLAFI(k - 1), the I term of the PID correction coefficient in the previous cycle in S916. As a result, when KLAF(k) is calculated in S904, the I term KLAFI thereof becomes:

KLAFI(k) = KSTR(k - 1) + DKAF(k) · KI

where the calculated I-term is added to the P-term and the D-term to obtain KLAF(k).

This method is adopted due to the rapid change that may occur in the integral term when the feedback correction coefficient is calculated after switching from adaptive control to PID control. By using the value of KSTR to determine the initial value of the PID correction coefficient in the foregoing manner, the difference between the correction coefficient KSTR(k - 1) and the correction coefficient KLAF(k) can be kept small. Thus, at the same time as switching from STR control to PID control, the difference in the gain of the feedback correction coefficient kept small, while the transition can be made smooth and continuous, thus preventing a sudden change in the controlled variable.

If S900 in the subroutine of Fig. 52 determines that the engine operating condition is in the STR (controller) operating region and S906 determines that the operating condition was not in the PID operating region in the previous control cycle, the feedback correction coefficient KSTR(k) is calculated based on the STR controller in S914. This calculation is carried out according to Equation 36 as previously explained.

Next, in S830 of the subroutine of Fig. 51, it is checked whether the correction coefficient calculated by the subroutine of Fig. 52 is equal to KSTR, and if so, the difference between 1.0 and KSTR(k) is calculated and this absolute value is compared with a threshold value KSTKref in S832.

This is partly related to what was said in the previous explanation. A wild fluctuation of the feedback correction coefficient causes sudden changes in the controlled variable and deteriorates the control stability. The absolute value of the difference between 1.0 and the feedback correction coefficient is therefore compared with a threshold value, and if it exceeds the threshold value, a new feedback correction coefficient is determined based on the PID control in S804 and the following steps. As a result, the controlled variable does not change suddenly, and stable control can be realized. Alternatively, it is possible to compare the coefficient, instead of the absolute value, with two threshold values based on the amplitude, with 1.0 set as the center thereof. This is shown in Fig. 53.

If S832 determines that the absolute value of the difference between 1.0 and the calculated feedback correction coefficient KSTR(k) does not exceed the threshold, the value determined by the STR controller is set as the feedback correction coefficient KFB in S834. If the result in S830 is no, the bit of the flag FKSTR is reset to 0 in S836, and the value determined by the PID controller is set as the feedback correction coefficient KFB in S838.

Next, in S718 of the routine of Fig. 49, the required fuel injection amount Tcyl is multiplied by the calculated feedback correction coefficient KFB and the like, and the addition term TTOTAL is added to the result to obtain the initial fuel injection amount Tout. The program then proceeds to S720 where the fuel adhesion correction is performed (explained later) and to S722 where the corrected initial fuel injection amount Tout is output to the fuel injector 22 via the driver circuit 72 as a manipulated variable.

If S704 determines that fuel cut is active, the output fuel injection amount Tout is set to 0 in S728. If the result in S708 or S710 is negative, since this means that open loop control is being performed, the program goes to S726 where KFB is set to 1.0 and to S718 where Tout is calculated. If the result in S704 is positive, control is performed and Tout is set to a predetermined value in S728.

In the foregoing, when the control is interrupted and the feedback control is continued, as in the case where the fuel supply is continued after it has been interrupted once, the feedback correction coefficient is determined based on the PID control law for a predetermined period of time. As a result, the feedback correction coefficient determined by the STR controller is not used during the periods during which the difference between the actual air/power ratio and the true air/fuel ratio is large due to the time required for the supplied fuel to be burned and the detection delay of the sensor itself. The controlled variable (actual value) therefore does not become unstable and does not affect the stability of the control.

On the other hand, since a predetermined value is set during this period, the control convergence can be improved after the feedback value has stabilized by using the adaptive correction coefficient obtained by the STR controller for the operation of the system, thus to absorb the control error completely at once. A particularly noteworthy feature of the embodiment is that an optimum balance between the control stability and the control convergence is achieved due to the fact that the control convergence is determined by determining the manipulated variable as a product of the feedback correction coefficient and the manipulated variable.

Note that since the actual air-fuel ratio is not stable immediately after the activation of the LAF sensor, the feedback correction coefficient can be determined using the PID control law for a predetermined period of time after the completion of the activation of the LAF sensor.

In addition, when a fluctuation of the target air-fuel ratio is large, the feedback correction coefficient is determined based on the PID control even after the elapse of the predetermined period, so that an optimum balance between control stability and convergence is achieved when the control is continued after a control such as at the time of interruption of fuel cut, full load enrichment or the like.

In addition, since the feedback control coefficient is determined by the PID control law when the adaptive control coefficient determined by the STR controller becomes unstable, a better balance between control stability and convergence is achieved.

Furthermore, when switching from STR control to PID control, the I term of KLAF is calculated using the feedback correction coefficient obtained from the STR controller, while when continuing STR control after PID control, a time point at which the actual value KACT is equal to or close to 1 is selected and the initial value of the feedback correction coefficient is set to be almost equal to the PID correction coefficient from the PID control law based on the adaptive control law (STR controller). In other words, the system ensures a smooth transition back and forth between PID control and adaptive control. Since the manipulated variable does not change suddenly, changes, the controlled variable does not become unstable.

In addition, since the feedback correction coefficient is determined based on the PID control law during engine idling, no conflict occurs between the fuel metering control and the intake air quantity control which is carried out during engine idling.

The following explains the fuel adhesion correction of the initial fuel injection amount Tout. The fuel adhesion correction is performed for each cylinder value as mentioned previously, with the values for each cylinder being identified by assigning a cylinder number n (n = 1, 2, 3, 4).

In the illustrated configuration, a fuel adhesion correction compensator having a transfer function inverse to that of the system is inserted in series before the fuel adhesion system. The fuel adhesion correction parameters are retrieved from map data prepared in advance in accordance with engine operating conditions such as the coolant temperature Tw, the engine speed Ne, the manifold pressure Pb, and the like.

When the recovered parameters and the actual parameters of the engine are identical, the product of the transfer functions of the system and the compensator becomes equal to 1.0, which means that the target fuel injection quantity is equal to the actual cylinder fuel intake quantity and the correction is perfect.

Based on the foregoing, the fuel adhesion correction of the initial fuel injection amount Tout in S720 of the flow chart of Fig. 49 will be explained with reference to a subroutine flow chart shown in Fig. 54. The program shown in Fig. 54 is activated at a crank angle position synchronized at every TDC and is repeatedly executed until the values Tout(n) for all cylinders are determined. The suffix (k - 1) indicates a value calculated in the last control cycle (last program loop). The value calculated in the current control cycle (current program loop) is not provided with the suffix (k).

The program starts at S1000 where the various parameters are read in and advances to S1002 where a direct ratio A and a discharge ratio B are determined. This is done by retrieving from mapped data (the characteristics of which are shown in Fig. 55) using the detected engine speed Ne and the manifold pressure Pb as address data. Note that the mapped data is set up separately for the HiV/T and dis LoV/T characteristics of the variable valve timing characteristics, and retrieval is carried out by actually selecting the mapped data corresponding to the valve timing characteristics. At the same time, a table (the characteristics of which are shown in Fig. 56) is searched using the detected coolant temperature as address data to retrieve a correction coefficient KATW and KBTW.

The ratios A, B are multiplied by the coefficients KATW and KBTW and corrected. Similarly, the other correction coefficients KA, KB are determined in response to the presence/absence of the EGR and canister purge operation and to the target air/power ratio KCMD, although the determination is not shown in the figure. More specifically, if the corrected ratios are referred to as Ae, Be, they are corrected as follows:

Ae = A · KATW · KA

Be = B · KBTW · KB

The program then proceeds to S1004 where it is determined whether the supply is cut off, and if the result is negative, it proceeds to S1006 where the initial fuel injection amount Tout is corrected in the manner shown to determine the initial fuel injection amount for the individual cylinders Tout(n)-F. If the result in S1104 is positive, the program proceeds to S1008 where the value Tout(n)-F is made equal to 0. Here, the displayed value TWP(n) indicates the amount of fuel adhering to the wall of the intake pipe 12.

Fig. 57 is a subroutine flow chart for determining or calculating the value TWP(n). The program shown is activated at a predetermined crank angle position.

The program starts at S1100, where it is determined whether the current program loop is within a period that begins at a time when the Tout calculation starts and ends at a time when the fuel injection to any cylinder ends. The period is hereinafter referred to as a "fuel metering control period". If the result is positive, the program advances to S1102, where the bit of a first flag FCTWP(n) indicating the end of the TWP(n) calculation for the cylinder n is set to 0 to allow the calculation, whereupon the program immediately ends.

If the result in S1101 is negative, the program proceeds to S1104, where it is confirmed whether the bit of the flag FCTWP(n) is equal to 1, and if so, since this means that the value of TWP(n) for the cylinder in question is completed, the program proceeds to S1106. If, on the other hand, the result is negative, the program proceeds to S1108, where it is determined whether the fuel supply is shut off. If the response in S1108 is no, the program proceeds to S1110, where the value of TWP(n) is calculated in the manner shown.

Here, the value TWP(k - 1) is a value calculated in the last control cycle. The first term on the right side of the equation denotes the amount of fuel that adhered to the wall at the last injection and still remains there without being discharged, and its second term denotes the amount of fuel that adheres to the wall at the current injection.

The program then advances to S1112 where the bit of a flag TTWPR(n) (indicating that the fuel adhesion amount is 0) is set to 0, and to S1106 where the bit of the first flag FCTWP(n) is set to 1, whereupon the program ends.

If S1108 determines that the fuel supply is cut off, the program proceeds to S1114 where it is determined whether the bit of the second flag FTWPR(n) (indicating that the remaining fuel adhesion amount is 0) is 1, and if so, it proceeds to S1106 since TWP(n) = 0. If the result at S1114 is negative, the program proceeds to S1116 where the value TWP(n) is calculated according to the equation shown. The equation is the same as that shown at S1110 except that the right second term is deleted. This is because the fuel supply is cut off and no fuel adhesion occurs.

The program then proceeds to S1118 where it is determined whether the value TWP(n) is greater than a predetermined small value TWPLG, and then to S1112 if this is affirmative. If this is not affirmative, since this means that the remaining fuel adhesion amount is small enough to be ignored, the program proceeds to S1120 where the value is set to 0 and to S1122 where the bit of the second flag is set to 1 and then to 31106.

With the arrangement, it becomes possible to accurately determine the amount of fuel adhering to the intake manifold wall for each cylinder (the value TWP(n)), and using the value in determining the initial fuel injection amount Tout in the configuration shown in Fig. 54, it becomes possible to supply the optimum amount of fuel to each cylinder combustion chamber taking into account the amount of fuel remaining on the intake manifold wall and that taken out therefrom. Note that the foregoing adhesion correction including the calculation of the ratios A, B is carried out when the engine starts. Furthermore, the foregoing description should be applied regardless of whether the fuel injection for each cylinder is carried out simultaneously or is carried out sequentially in the firing order.

In the foregoing, it should be noted that in the configuration shown in Fig. 8, it is alternatively possible to provide a third catalyst 94, as shown in a block 400 shown with a dashed line. The The third catalyst 94 is preferably a so-called "lighting" catalyst which stimulates the activation of the catalysts in a shorter period, such as a catalyst which is a so-called "electrically heated catalyst" with a heater to promote the activation. In this sense, the volume of the third catalyst should be sufficiently smaller than the catalysts installed downstream thereof.

The third catalyst 94 may be a 3-way catalyst similar to the others. The third catalyst 94 may be provided if necessary. However, when the engine is a V-type engine and the fuel metering control system is formed according to the invention (for each bank of the V-type engine), the provision of the third catalyst 94 is effective because the volume or amount of exhaust gas at each bank is relatively small. The provision of the third catalyst would affect the dead time in the system, as a result of which the control amount and the like become different.

In the foregoing, it should be noted that the configuration shown in Fig. 8 alternatively allows for providing a filter 96 in front of the observer, as shown by dashed lines. The detection response delay of the LAF sensor is adjusted by the observer calculation as mentioned above, where the delay can alternatively be adjusted in hardware fashion by providing such a filter 96 as is possible to compensate for the first order delay.

It should also be noted that in the configuration disclosed in the block diagram of Fig. 8, not all elements are necessary. Rather, one or some of the elements may be omitted.

Fig. 58 is a block diagram similar to Fig. 8, but showing the configuration of the system according to a second embodiment of the invention.

In the second embodiment, a second O₂ sensor 98 is installed downstream of the second catalyst 30. The outputs of the second O₂ sensor 98 are used to correct the target air-fuel ratio KCMD as shown. The target air-fuel ratio KCMD can thus be determined more optimally, which improves the control performance. Since By detecting the air/fuel ratio of the exhaust gas that is ultimately released into the air, the emission efficiency is improved. The configuration also makes it possible to monitor whether the catalysts positioned upstream of the O₂ sensor 98 are deteriorating.

The second O2 sensor 98 may be used as a replacement for the first O2 sensor 56. The second catalyst 30 may have the same configuration as disclosed in Figure 5, wherein the second O2 sensor may be located at a position between the catalyst beds.

The second O₂ sensor 98 is followed by a low-pass filter 500 with a corner frequency of 1,000 Hz. Since the filter 500 and the filter 60 do not have linear characteristics, they may be of the type called a linearizer which can compensate for the deficiency.

Furthermore, in the second embodiment, the filter 58 (shown in the figure of the first embodiment) is no longer provided between the observer input and the LAF sensor 54. This is because, as previously mentioned, the observer does not function to converge the actual air/fuel ratio to the target air/fuel ratio. Rather, the system is configured to obtain the dispersion between the individual cylinder air/fuel ratios based on the air/fuel ratios estimated by the observer. Accordingly, even if the LAF sensor response (time) is not stable, the estimation is not greatly affected. Rather, the observer estimation accuracy improves with decreasing response time, i.e., with a faster response of the LAF sensor, resulting in improved PID controller performance. Thus, when priority is given to the detection response, the filter 58 is removed. This makes it possible to make the LAF sensor response time as short as possible.

In the second embodiment, however, it should be noted that the filters 92, 93 are still left in front of the inputs of the adaptive controller and the PID controller. When the same type of manipulated variables (one manipulated variable in the disclosure, ie the fuel injection quantity) are determined by means of several controllers using a single sensor output, then, if the response (time) of the sensor was set uniformly, interference between the controllers would occur, causing a problem such as generation of control oscillations in the manipulated variables and thus in the controlled variables. However, in the second and first embodiments, since the LAF sensor outputs are input to the controllers through filters of different corner frequencies, this configuration can not only eliminate the noise in the sensor output but also make the sensor response (time), that is, the sensor phase characteristics, different, and can consequently avoid interference between the controllers.

Generally speaking, in a multiple feedback control system with several controllers arranged in parallel, such as the one disclosed in the figure, when the controllers, whose control objects are the same but whose control laws are different from each other, use a single sensor output for calculating a same type of manipulated variables, it is possible to vary the sensor response (time), i.e. the sensor phase characteristics, by installing at least one filter in front of at least one of the controllers. With the arrangement, it becomes possible to avoid interference between the controllers and to improve the controller stability while achieving the desired control accuracy.

This is effective not only for the multiple controller system as disclosed, but also for a conventional system in which controllers are switched.

In addition, although the LAF sensor output sampling block Sel-V is left only for the observer in the second embodiment, similarly to the first embodiment, the block is removed before the STR and PID controllers. Since the filters 92 and 93 are provided for the controllers, it is possible to achieve the desired control stability by setting the corner frequency of these filters appropriately. For this reason, the sampling block is omitted from the inputs of the controllers.

Further, in the second embodiment, it is noted that the third catalyst 94 may be installed as in the first embodiment.

Furthermore, in the second and first embodiments, it should be noted that the filters may be formed using a software technique or a hardware technique.

While in the previous embodiments the throttle valve has been actuated by a stepper motor, it may instead be mechanically connected to the accelerator pedal and actuated directly in response to depression of an accelerator pedal.

Although a motor-driven type EGR control valve is used in the EGR mechanism, it is alternatively possible to use a valve with a diaphragm that is actuated by the negative pressure in the intake pipe.

The second catalyst 30 may be omitted, although this depends on the performance of the first catalyst.

Although a low-pass filter is used, it is alternatively possible to use an equivalent band-pass filter.

Although the foregoing embodiments as described use the output of a single air/fuel ratio sensor installed at the exhaust system confluence point, the invention is not limited to this arrangement, and instead it is possible to perform the air/fuel ratio control based on air/fuel ratios detected by air/fuel ratio sensors installed for the individual cylinders.

Although the air/fuel ratio is actually expressed as an equivalence ratio, the air/fuel ratio and the equivalence ratio may instead be determined separately.

Although the feedback correction coefficients RSTR, #nKLAF and KLAF were calculated as multiplication coefficients (expressions) in the foregoing embodiments, they may be calculated as addition expressions instead.

Although the foregoing embodiments have been described with reference to examples that use STRs, MRACS (Model Reference Adaptive Control Systems) may be used instead.

Claims (8)

1. A system for controlling the fuel metering for a combustion engine (10) with multiple cylinders, comprising: an air/fuel ratio sensor (54) installed on the exhaust system (26) of the engine (10) for detecting an air/fuel ratio (KACT) of the engine (10); an engine operating state detecting means (40, 44, 34, S10) for detecting the engine operating states including at least the engine speed (Ne) and the engine load (Pb); a fuel injection quantity determining means (34, S16-S28, S706) for determining the fuel injection quantity (TiM-F; Tcyl) for the individual cylinders at least on the basis of the detected engine operating conditions; a plurality of controllers (PID, PID(STR)) for inputting the output signal of the air/fuel ratio sensor (54) as a controlled variable and for determining the feedback correction coefficients (#nKLAF, KFB) for correcting the fuel injection quantity (TiM-F; Tcyl) supplied to the engine as a manipulated variable so that the controlled variable (KACT, Gfuel) is brought to a setpoint value (KCMD, Gfuel-str); an initial fuel injection amount determining means (34, S720) for determining the initial fuel injection amount (Tout) using the feedback correction coefficients (#nKLAF; KFB); and a fuel injection device for injecting fuel into the individual cylinders of the engine (10) based on the initial fuel injection amount (Tout); characterized in that the multiple controllers (PID, PID(STR)) the feedback correction coefficients (#nKLAF; KSTR, KLAF(KFB) based on setpoints which are determined separately; and at least one of the plurality of controllers (PID, PID(STR)) receives the output signal of the air/fuel ratio sensor (54) via a filter so that the input from the output of the air/fuel ratio sensor (54) input to the at least one controller is different from the input from the output of the air/fuel ratio sensor (54) input to another of the plurality of controllers (PID, PID(STR)). 2. A system according to claim 1, wherein the output of the air/fuel ratio sensor (54) is input to the plurality of controllers (PID, PID(STR)) via respective filters whose characteristics are different from each other. 3. The system of claim 1 or 2, further comprising: individual cylinder air/fuel ratio estimation means (34, S600- S618, S712, S714), comprising: a model (34) describing the behavior of the exhaust system (26) of the engine (10) and receiving the output signal from the air/fuel ratio sensor (54); an observation means (34, S610-S612) for observing an internal state of the exhaust system (26) described by the model; and an estimating means (34, S614-S618, S712, S714) for estimating the air/fuel ratios of the individual cylinders based on an output of the observing means; wherein at least one of the plurality of controllers (PID, PID(STR)) determines one of the feedback correction coefficients (#nKLAF) such that the estimated air/fuel ratios of the individual cylinders are brought to one of the target values. 4. System according to any one of the preceding claims 1 to 3, wherein at least one of the controllers (PID, PID(STR)) is a controller based on a control law expressed in a recursion formula and determines one of the feedback correction coefficients according to the recursion formula so that the actual air/fuel ratio (KACT) is adjusted to a target Air/fuel ratio (KCMD). 5. System according to any one of the preceding claims 2 to 4, wherein at least one of the filters is a low-pass filter (58). 6. System according to any one of the preceding claims 2 to 5, further comprising: individual cylinder air/fuel ratio estimation means (34, S600- S618, S712, S714), comprising: a model (34) describing the behavior of the exhaust system (26) of the engine (10) and receiving an output signal of the air/fuel ratio sensor (54) via a first filter (58, 96); an observation means (34, S610-S612) for observing an internal state of the exhaust system described by the model; and an estimating means (34, S614-S618, S712, S714) for estimating the air/fuel ratios of the individual cylinders based on an output signal of the observing means; wherein at least one of the plurality of controllers (PID, PID(STR)) receives the output signal of the air/fuel ratio sensor (54) via a second filter (92) and determines one of the feedback correction coefficients (#nKLAF) so that the estimated air/fuel ratios of the individual cylinders are brought to one of the target values, and wherein another of the plurality of controllers (PID, PID(STR)) is a controller which uses a control law expressed by a recursion formula and which receives the output signal of the air/fuel ratio sensor (54) via a second filter (92) and determines another of the feedback correction coefficients (KSTR(KFB)) according to the recursion formula so that the actual air/fuel ratio (KACT) is brought to a target air/fuel ratio (KCMD). 7. The system of claim 6, wherein the first filter (58, 96) and the second filter (92) are low-pass filters. 8. The system of claim 7, wherein the second filter (92) has a corner frequency that is lower than that of the first filter (58, 96).