JP2843385B2 - Engine air-fuel ratio learning control device - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an air-fuel ratio learning control device for an engine capable of performing air-fuel ratio learning even when the air-fuel ratio greatly deviates from a stoichiometric air-fuel ratio.

[Problems to be solved by conventional technology and invention] Conventionally, in order to improve the exhaust purification performance of an engine,
As shown in JP-A-62-135635, an O2 sensor is arranged upstream of a catalyst provided in an exhaust system of an engine, and a fuel injection amount is corrected according to an air-fuel ratio state based on an output value of the O2 sensor. Thus, the air-fuel ratio is feedback-controlled to a stoichiometric air-fuel ratio at which the catalyst can exhibit the highest purification efficiency. Also,
In this air-fuel ratio control, variations in the production of an intake air system such as an intake air amount sensor, and a fuel system such as an injector,
Alternatively, in order to quickly correct the deviation of the air-fuel ratio due to aging, learning control is adopted so that the state of the stoichiometric air-fuel ratio is always maintained even when the operating state changes significantly.

That is, in the air-fuel ratio learning control, the O2 sensor output value (output voltage) is compared with a slice level (reference voltage) corresponding to the stoichiometric air-fuel ratio (stoichiometric) for determining the air-fuel ratio state, and the air-fuel ratio state is determined. to decide. When the air-fuel ratio shifts from rich to lean or from lean to rich, a proportional constant P for skipping the air-fuel ratio correction amount is added to or subtracted from the previous air-fuel ratio feedback correction coefficient α, and the air-fuel ratio becomes rich or lean. Is set, the current air-fuel ratio feedback correction coefficient α is set by a proportional integration process of subtracting or adding an integration constant I for gradually decreasing or increasing the air-fuel ratio correction amount to the previous air-fuel ratio feedback correction coefficient α. I do. Also,
In order to prevent over-correction of the air-fuel ratio by the air-fuel ratio feedback correction coefficient α, the air-fuel ratio feedback correction coefficient α is
It is regulated by an upper limit (for example, 1.2) and a lower limit (for example, 0.7).

On the other hand, the deviation of the air-fuel ratio feedback correction coefficient α from the reference value (= 1.0) is learned for each engine operation region and stored as a learning value.

Then, the basic fuel injection amount set based on the engine operating state is obtained by interpolating at least the air-fuel ratio feedback correction coefficient α and the learning value learned for each engine operating area in accordance with the current engine operating state. By adjusting the final fuel injection amount with the learning correction coefficient, the air-fuel ratio is maintained at a stoichiometric air-fuel ratio that can improve the exhaust purification efficiency of the catalyst most.

Here, the learning value is stored in a learning value table configured in the backup RAM in accordance with the engine speed and the load (for example,
The basic fuel injection pulse width that determines the basic fuel injection amount) or the engine operation region based on only the load is stored.
The learning of the air-fuel ratio is performed in accordance with the closed-loop correction coefficient based on the O2 sensor output, i.e., the comparison result between the O2 sensor output and the slice level, in the operating region specified by the engine speed and the load. The air-fuel ratio shifts to rich from the state of α> 1.0, in which the air-fuel ratio feedback correction coefficient α set by the proportional integration process corrects the lean air-fuel ratio in the rich direction by the fuel increase correction, and the rich air-fuel ratio is reduced. The air-fuel ratio shifts to lean from the state of α <1.0, which is corrected in the lean direction by the correction, or the state of α <1.0, in which the rich air-fuel ratio is corrected in the lean direction by the fuel reduction correction. When α> 1.0, which is corrected in the rich direction by the increase correction, and crosses the reference value (= 1.0) several times within a predetermined time, That is, when the air-fuel ratio is repeatedly rich / lean a predetermined number of times by the air-fuel ratio feedback correction using the air-fuel ratio feedback correction coefficient α, it is determined that the engine is in a steady operation state, and the average value (local maximum value) of the air-fuel ratio feedback correction coefficient α during this period is determined. The learning value of the corresponding area is updated in accordance with the deviation of the average value between the reference value and the minimum value) from the reference value (= 1.0), and the deviation of the air-fuel ratio feedback correction coefficient α from the reference value is shifted to the learning value. I have.

Then, a learning value table is searched based on the engine operating state, and a learning correction coefficient obtained by interpolation calculation of the learning value is incorporated into an arithmetic expression of a fuel injection amount (fuel injection amount) together with an air-fuel ratio feedback correction coefficient α. This makes it possible to appropriately control the air-fuel ratio even during the open-loop control, and to relatively reduce the air-fuel ratio correction amount with respect to the reference value based on the air-fuel ratio feedback correction coefficient α during the air-fuel ratio feedback control. Air-fuel ratio controllability is improved as much as possible.

Thus, even if the engine operating state changes, the integration speed (generally, several% / second) given by the integration constant I in the proportional integration processing of the air-fuel ratio feedback correction coefficient α immediately follows. The learning correction coefficient based on the learning value is reflected in the fuel injection amount,
The air-fuel ratio can be kept at the stoichiometric air-fuel ratio.

However, if the O2 sensor is deteriorated, the output characteristics and responsiveness of the O2 sensor are deteriorated and a response delay of the air-fuel ratio detection occurs, and at this time, when the deviation amount of the air-fuel ratio with respect to the stoichiometric air-fuel ratio is large, , The O2 sensor output value remains largely deviated from the slice level, and the lean air-fuel ratio is continued by comparing the O2 sensor output value with the slice level, or
The continuation of the rich air-fuel ratio is detected.

Then, by continuing the lean air-fuel ratio or detecting the rich air-fuel ratio, the air-fuel ratio feedback correction coefficient α
Is gradually increased or decreased by the above-mentioned integration constant I, eventually reaches the upper limit or the lower limit, and remains in the state of sticking to the upper limit or the lower limit.

As a condition for updating the learning value, as described above, when the air-fuel ratio is rich and lean repeatedly a predetermined number of times by the air-fuel ratio feedback correction using the air-fuel ratio feedback correction coefficient α within a predetermined time, it is determined that the engine is in a steady operation state. The average value of the air-fuel ratio feedback correction coefficient α during this period (the average value of the local maximum value and the local minimum value) (= 1.
Since the learning value in the corresponding area is updated in accordance with the deviation from 0), the lean air-fuel ratio is maintained, or
With the detection of the rich air-fuel ratio, the learning value is not updated,
As a result, the learning correction of the air-fuel ratio is delayed.

In other words, the O2 sensor deteriorates, the output characteristics and the responsiveness of the O2 sensor deteriorate, causing a response delay in the air-fuel ratio detection, and
At this time, when the deviation amount of the air-fuel ratio from the stoichiometric air-fuel ratio is large, the air-fuel ratio feedback correction coefficient α remains stuck to the upper limit or the lower limit, and at this time, the air-fuel ratio feedback correction coefficient α Deviation from the reference value,
In other words, the deviation of the stoichiometric air-fuel ratio of the air-fuel ratio is not reflected in the learning value, and the basic fuel injection pulse width is corrected by the air-fuel ratio feedback correction coefficient and the learning correction coefficient obtained by interpolation calculation of the learning value. Since the fuel injection pulse width that determines the final fuel injection amount is set, the air-fuel ratio remains deviated from the stoichiometric air-fuel ratio, the air-fuel ratio controllability deteriorates, and the exhaust emissions and fuel efficiency deteriorate. However, there is an inconvenience that running performance is reduced.

To cope with this, Japanese Patent Application Laid-Open No. 62-135635 discloses that the engine steady operation state continues for a predetermined period of time, during which the O2 sensor output does not switch from rich to lean or from lean to rich. It is determined that the output of the O2 sensor is stuck to the rich side or the lean side. If the output of the O2 sensor indicates lean at this time, the learning value stored in the corresponding area is set to the predetermined value (the minimum resolution). The learning value is increased by a predetermined amount. On the other hand, if the O2 sensor output indicates rich, the learning value stored in the corresponding area is increased. A technique is disclosed in which a predetermined value is subtracted from the reference value to reduce the learning value by a predetermined amount.

According to this prior example, the O2 sensor deteriorates, the output characteristics and the responsiveness of the O2 sensor deteriorate, causing a response delay in the air-fuel ratio detection, and at this time, the deviation amount of the air-fuel ratio from the stoichiometric air-fuel ratio is reduced. If the value is large, the O2 sensor output sticks to the rich side or lean side, and the learning value is corrected accordingly.Therefore, the deviation of the air-fuel ratio from the stoichiometric air-fuel ratio is reflected in the learning value, and the air-fuel ratio learning is performed. Control can be resumed.

[Object of the Invention] However, in the above prior art, when the engine steady operation state continues for a predetermined time,
During this time, when the air-fuel ratio detected by the O2 sensor output is not switched from rich to lean or from lean to rich, it is determined that the output of the O2 sensor is stuck to the rich side or the lean side. There is a possibility that a determination may occur.

In other words, with the O2 sensor not deteriorating,
Under the situation where the air-fuel ratio is almost converging to the stoichiometric air-fuel ratio,
By the air-fuel ratio feedback correction by the air-fuel ratio feedback correction coefficient, the time until the air-fuel ratio shifts from rich to lean and then shifts to rich again, or the time until the air-fuel ratio shifts from lean to rich and shifts to lean again, It differs for each engine operating state.

For this reason, the output characteristics and responsiveness of the O2 sensor are deteriorated due to the deterioration of the O2 sensor, causing a response delay in the air-fuel ratio detection, and a state in which the deviation amount of the air-fuel ratio from the stoichiometric air-fuel ratio is large. In order to cover by determining the switching of the air-fuel ratio in a unique predetermined time, the determination threshold value for determining the predetermined time must be set to a slow value, that is, a relatively long time. In the engine state, the output characteristics and responsiveness of the O2 sensor are deteriorated due to the deterioration of the O2 sensor, the response delay of the air-fuel ratio detection is caused, and the deviation amount of the air-fuel ratio from the stoichiometric air-fuel ratio is large. Judgment cannot be made appropriately for each operating state.

Therefore, the learning value is corrected according to the ambiguous determination result, and the fuel injection amount is corrected by the learning correction coefficient obtained by interpolation calculation of the learning value, so that the controllability and reliability of the air-fuel ratio learning control are reduced. It cannot be improved enough.

Further, in this prior example, when the engine steady operation state continues for a predetermined time, the air-fuel ratio detected by the O2 sensor output is not switched from rich to lean or from lean to rich during this period. Sometimes, the learning value is corrected, and the sticking of the air-fuel ratio feedback correction coefficient to the upper limit or the lower limit cannot be directly suppressed.

Therefore, the deviation of the air-fuel ratio from the stoichiometric air-fuel ratio is not always accurately reflected in the learning value, and the basic fuel injection pulse width is determined by the air-fuel ratio feedback correction coefficient and the learning correction coefficient obtained by interpolating the learning value. Since the fuel injection pulse width that determines the final fuel injection amount is set after correction, the air-fuel ratio may remain deviated from the stoichiometric air-fuel ratio, and the air-fuel ratio controllability may deteriorate.

The present invention has been made in view of the above circumstances, and the output characteristics and responsiveness of the O2 sensor have deteriorated due to the deterioration of the O2 sensor, and a response delay of the air-fuel ratio detection has occurred. If the amount of deviation is large, this is accurately determined to prevent the air-fuel ratio feedback correction coefficient from sticking to the upper limit or lower limit, and the deviation between the air-fuel ratio and the stoichiometric air-fuel ratio is accurately determined as a learning value. The air-fuel ratio learning control of the engine that enables the air-fuel ratio learning to be reflected, improves the controllability and reliability of the air-fuel ratio learning control, and improves the exhaust emission, fuel consumption, and drivability It is intended to provide a device.

[Means for Solving the Problems] In order to achieve the above object, the invention described in claim 1 is based on O2
The output value of the sensor is compared with a slice level for judging the air-fuel ratio state to judge the air-fuel ratio state. When the air-fuel ratio shifts from rich to lean or from lean to rich, the air-fuel ratio correction amount is skipped. Add or subtract the proportionality constant from the previous air-fuel ratio feedback correction coefficient,
When the air-fuel ratio continues to be rich or lean, the current air-fuel ratio feedback correction is performed by a proportional integration process of subtracting or adding an integration constant for gradually decreasing or increasing the air-fuel ratio correction amount to the previous air-fuel ratio feedback correction coefficient. The coefficient is set, and the deviation of the air-fuel ratio feedback correction coefficient from the reference value is learned for each engine operation region and stored as a learning value. The basic fuel injection amount set based on the engine operating state is at least the air-fuel ratio. A feedback correction coefficient and a learning correction coefficient obtained by interpolating and calculating a learning value learned for each engine operating area in accordance with the current engine operating state to thereby adjust the final fuel injection amount. In the fuel ratio learning control device, the air / fuel ratio feedback based on the air / fuel ratio feedback correction coefficient is used. Normal-time reference inversion representing at least one of the time from when the air-fuel ratio shifts from rich to lean to rich again by correction, and the time from when the air-fuel ratio shifts from lean to rich and back again. A reference inversion time setting means for setting a time based on the engine speed and the basic fuel injection amount, and switching from rich to lean of the air-fuel ratio within the reference inversion time based on the O2 sensor output value, Reversal determining means for determining whether at least one of switching to rich has been performed,
When there is no air-fuel ratio change within the reference inversion time,
An integration constant correction means for reducing the integration constant for gradually decreasing or increasing the air-fuel ratio feedback correction coefficient when the rich air-fuel ratio or the lean air-fuel ratio state continues, and reducing the integration constant from a normal time; In the case where the air-fuel ratio is switched within the time, and when the number of switching of the air-fuel ratio reaches the set number, the deviation of the air-fuel ratio feedback correction coefficient from the reference value is learned for each engine operating region. When the air-fuel ratio is not switched within the reference inversion time, the change state of the air-fuel ratio feedback correction coefficient is determined. When the air-fuel ratio feedback correction coefficient is increasing, the learning of the corresponding area is determined. When the air-fuel ratio feedback correction coefficient is decreased by increasing the value by a predetermined amount, learning value updating means for decreasing the learning value of the corresponding region by a predetermined amount; The basic fuel injection amount set based on the engine operating state is calculated by interpolating at least the air-fuel ratio feedback correction coefficient and the learning value learned for each engine operating region in accordance with the current engine operating state. And a fuel injection amount setting means for correcting with a correction coefficient and setting a final fuel injection amount.

The invention according to claim 2 determines the air-fuel ratio state by comparing the output value of the O2 sensor with a slice level for determining the air-fuel ratio state, and when the air-fuel ratio shifts from rich to lean or from lean to rich. The proportional constant for skipping the air-fuel ratio correction amount is added or subtracted from the previous air-fuel ratio feedback correction coefficient, and when the air-fuel ratio continues to be rich or lean, the air-fuel ratio correction amount is gradually reduced or increased. The current air-fuel ratio feedback correction coefficient is set by a proportional integration process of subtracting or adding the integration constant to the previous air-fuel ratio feedback correction coefficient, and the deviation of the air-fuel ratio feedback correction coefficient from the reference value is learned for each engine operating region. And stores the basic fuel injection amount set based on the engine operating state at least as a learning value. The air-fuel ratio feedback correction coefficient and the learning value learned for each engine operation area are corrected by a learning correction coefficient obtained by interpolation calculation corresponding to the current engine operation state, and the final fuel injection amount is set. In the air-fuel ratio learning control device of the engine, after the air-fuel ratio shifts from rich to lean by the air-fuel ratio feedback correction by the air-fuel ratio feedback correction coefficient, the time until the air-fuel ratio again shifts to rich, and after the shift from lean to rich,
A reference reversal time setting means for setting a reference reversal time corresponding to a normal time based on the engine speed and the basic fuel injection amount, which represents at least one of a time until the shift to lean again, based on the O2 sensor output value. Reversing determining means for determining whether at least one of the switching of the air-fuel ratio from rich to lean within the reference reversing time and the switching from lean to rich has occurred, and When the fuel ratio is not switched, the change state of the air-fuel ratio feedback correction coefficient is determined. When the air-fuel ratio feedback correction coefficient is increasing, the learning value in the corresponding region is increased by a predetermined amount, and the air-fuel ratio feedback correction coefficient is increased. When the air-fuel ratio is decreasing, the learning value in the corresponding area is reduced by a predetermined amount. When the air-fuel ratio is switched within the reference inversion time, the air-fuel ratio is not switched. From the state to the state with air-fuel ratio switching, and the first time the air-fuel ratio switching number reaches the set number of times, the air-fuel ratio switching number after the air-fuel ratio switching The total amount of deviation of the average value of the air-fuel ratio feedback correction coefficient from the reference value during the period of reaching the reference value is subtracted from the learning value of the corresponding area to update the learning value of the corresponding area, and the air-fuel ratio feedback correction coefficient is set to the reference value. After that, while the air-fuel ratio switching continues, the average value of the air-fuel ratio feedback correction coefficient is
Learning value updating means for subtracting a predetermined ratio of the deviation from the reference value from the learning value in the corresponding area to update the learning value in the corresponding area, and the basic fuel injection amount set based on the engine operating state, at least, A fuel for correcting a fuel ratio feedback correction coefficient and a learning correction coefficient obtained by interpolating and calculating the learning value learned for each engine operating region in accordance with the current engine operating state to set a final fuel injection amount. Injection amount setting means.

According to a third aspect of the present invention, in the first or second aspect, the reference reversal time setting means sets the reference reversal time to a shorter value as the engine rotation speed and the engine load increase. And

[Operation] According to the first aspect of the present invention, based on the engine speed and the load, the time from when the air-fuel ratio shifts from rich to lean by the air-fuel ratio feedback correction using the air-fuel ratio feedback correction coefficient and before the air-fuel ratio shifts to rich again is calculated. , A reference inversion time corresponding to a normal time, which represents at least one of a time until a transition from lean to rich and a transition to lean again. Then, based on the O2 sensor output, it is determined whether or not at least one of the switching of the air-fuel ratio from rich to lean and the switching from lean to rich within the reference inversion time is determined. It is determined that the output characteristics and the response of the O2 sensor are deteriorated due to the deterioration of the O2 sensor, the response delay of the air-fuel ratio detection occurs, and the deviation amount of the air-fuel ratio from the stoichiometric air-fuel ratio is large. Here, the reference reversal time for judging the switching of the air-fuel ratio is set in accordance with the engine operation state based on the engine speed and the basic fuel injection amount. If the air-fuel ratio detection response is delayed due to the deterioration of the output characteristics and responsiveness of the air-fuel ratio and the deviation of the air-fuel ratio from the stoichiometric air-fuel ratio is large, this is appropriately determined for each engine operating state. It is possible to do. Then, when the air-fuel ratio is switched within the reference inversion time, and when the air-fuel ratio switching count reaches the set count, the deviation of the air-fuel ratio feedback correction coefficient from the reference value is determined for each engine operating region. And stored as a learning value. Further, when the air-fuel ratio is not switched within the reference inversion time, the response characteristics of the output characteristics of the O2 sensor deteriorate due to the deterioration of the O2 sensor, and the response delay of the air-fuel ratio detection occurs. It is determined that the deviation amount of the air-fuel ratio with respect to the air-fuel ratio is large, and the integral constant for gradually decreasing or increasing the air-fuel ratio feedback correction coefficient when the rich air-fuel ratio or the lean air-fuel ratio state continues is determined. By reducing the air-fuel ratio feedback correction coefficient to a lower or higher speed than normal, the air-fuel ratio feedback correction coefficient is prevented from sticking to the lower limit or upper limit. Further, when the air-fuel ratio is not switched within the reference inversion time, the change state of the air-fuel ratio feedback correction coefficient is further determined. When the air-fuel ratio feedback correction coefficient is increasing, the learning value of the corresponding area is determined. When the air-fuel ratio feedback correction coefficient is decreased by a fixed amount, the learning value of the corresponding area is decreased by a predetermined amount. Due to the deterioration of the O2 sensor,
Even when the output characteristics and responsiveness of the O2 sensor have deteriorated and the response delay of the air-fuel ratio detection has occurred, and the deviation amount of the air-fuel ratio is large compared to the stoichiometric air-fuel ratio, the air-fuel ratio feedback correction coefficient can be accurately determined. The upper limit value that defines the correction limit of the fuel increase rate by the air-fuel ratio feedback correction coefficient or the lower limit value that defines the correction limit of the fuel reduction rate by the air-fuel ratio feedback correction coefficient by shifting the air-fuel ratio correction amount to the learning value , A margin of the air-fuel ratio feedback correction coefficient is secured, and the air-fuel ratio correction using the air-fuel ratio feedback correction coefficient becomes possible. The air-fuel ratio correction amount for the fuel injection amount is determined by the air-fuel ratio feedback correction coefficient and a learning correction coefficient obtained by interpolation calculation of a learning value obtained by shifting the air-fuel ratio correction amount based on the air-fuel ratio feedback correction coefficient. Therefore, when the output characteristics and responsiveness of the O2 sensor are deteriorated due to the deterioration of the O2 sensor, a response delay of the air-fuel ratio detection occurs, and the deviation amount of the air-fuel ratio from the stoichiometric air-fuel ratio is large. Also, the synergistic effect of the air-fuel ratio feedback correction coefficient and the learning correction coefficient makes it possible to set the air-fuel ratio correction amount large in accordance with a large deviation amount of the air-fuel ratio from the stoichiometric air-fuel ratio. As a result, the output characteristics and responsiveness of the O2 sensor are deteriorated due to the deterioration of the O2 sensor, causing a response delay in the air-fuel ratio detection, and when the deviation amount of the air-fuel ratio from the stoichiometric air-fuel ratio is large. It is also possible to converge the air-fuel ratio to the stoichiometric air-fuel ratio,
It is possible to restart the air-fuel ratio learning based on the air-fuel ratio feedback correction coefficient. Then, the basic fuel injection amount set based on the engine operating state is obtained by interpolating at least the air-fuel ratio feedback correction coefficient and a learning value learned for each engine operating region in accordance with the current engine operating state. The final fuel injection amount is corrected by a learning correction coefficient. Therefore, when the output characteristics and the responsiveness of the O2 sensor are deteriorated due to the deterioration of the O2 sensor, a response delay of the air-fuel ratio detection occurs, and the deviation amount of the air-fuel ratio from the stoichiometric air-fuel ratio is large. It is possible to appropriately judge this, accurately reflect the deviation between the air-fuel ratio and the stoichiometric air-fuel ratio in the learning value, perform the air-fuel ratio learning correction, and appropriately correct the fuel injection amount. Controllability and reliability of control are improved.

According to the second aspect of the present invention, when the air-fuel ratio is switched within the reference reversal time, the state is shifted from the state where the air-fuel ratio is not switched to the state where the air-fuel ratio is switched, and first, the air-fuel ratio is switched. At the first time when the number of times of switching has reached the set number of times, the deviation of the average value of the air-fuel ratio feedback correction coefficient from the reference value during the time when the number of times of switching of the air-fuel ratio has reached the set number of times after the air-fuel ratio has been switched. The total amount is subtracted from the learning value of the corresponding area, and the learning value of the corresponding area is updated at one time according to the deviation. Then, corresponding to this, the air-fuel ratio feedback correction coefficient is corrected to the reference value. And
Thereafter, while the switching of the air-fuel ratio is continued, a predetermined ratio of the deviation of the average value of the air-fuel ratio feedback correction coefficient from the reference value is subtracted from the learning value of the corresponding area to obtain the learning value of the corresponding area. Update.

Further, in setting the reference reversal time based on the engine speed and the basic fuel injection amount, in the invention according to the third aspect, the reference reversal time is set to a shorter value as the engine rotation speed is higher and the load is higher.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.

1 shows an embodiment of the present invention, FIG. 1 is a block diagram showing a functional configuration of a control device, FIG. 2 is an overall schematic diagram of an engine, and FIG. 3 shows a steady state determination matrix and a learning value table. FIG. 4 is an explanatory diagram showing a reference inversion time map, and FIG. 5 is a time chart showing an air-fuel ratio state, an O2 sensor output voltage, an air-fuel ratio feedback correction coefficient setting state, and a learning value update state. , Sixth
FIG. 7 is a flowchart showing a fuel injection control procedure, FIG. 7 is a flowchart showing a procedure for setting an air-fuel ratio feedback correction coefficient, and FIG. 8 is a flowchart showing a learning value updating procedure.

First, the overall configuration of the engine will be described with reference to FIG. In the figure, reference numeral 1 denotes an engine, and in the figure, a horizontally opposed four-cylinder engine is shown. A cylinder head 2 is provided in each of the left and right banks of the engine 1, and an intake port 2 a and an exhaust port 2 b are formed in each cylinder head 2.

In the intake system of the engine 1, an intake manifold 3 is connected to each intake port 2a, a throttle chamber 5 is connected to an upstream side of the intake manifold 3 via an air chamber 4, and an intake air is provided to an upstream side of the throttle chamber 5. An air cleaner 7 is attached via a pipe 6.

An intake air amount sensor (hot wire type air flow meter in the figure) 8 is interposed immediately upstream of the air cleaner 7 in the intake pipe 6, and a throttle valve 5 a provided in the throttle chamber 5 has a throttle opening 5 a. The degree sensor 9a and an idle switch 9b for detecting the fully closed state of the throttle valve 5a are connected to each other.

Further, an injector 10 is arranged immediately upstream of the intake port 2a of each cylinder of the intake manifold 3. Further, an ignition plug 11 whose tip is exposed to the combustion chamber is attached to each cylinder of the cylinder head 2.

The injector 10 is connected to a fuel tank 13 via a fuel supply path 12, and the fuel supply path 12
From the 13 side, a fuel pump 14 and a fuel filter 15 are interposed.

Further, the injector 10 communicates with the fuel chamber 17a of the pressure regulator 17 via the return passage 16. The downstream side of the fuel chamber 17a is the fuel tank 13
The pressure regulating chamber 17b of the pressure regulator 17 communicates with the intake manifold 3 as shown by a dashed line in FIG.

That is, the fuel pumped from the fuel tank 13 by the fuel pump 14 reaches the injector 10 and the pressure regulator 17 via the fuel filter 15, and the pressure regulator 17 keeps the pressure difference between the pressure in the intake manifold 3 and the fuel pressure constant. Kept in the injector
The fuel pressure is supplied to the intake manifold 10 and the fuel pressure is regulated so that the fuel injection amount from the injector 10 does not fluctuate due to the fluctuation of the pressure in the intake manifold 3.

A crank rotor 18 is mounted on a crank shaft 1b of the engine 1.
A crank angle sensor 19 composed of an electromagnetic pickup or the like is provided to detect the crank angle. Further, a cam rotor 20 is axially mounted on a camshaft 1c which makes a half rotation with respect to the crankshaft 1b, and a cam angle sensor 21 for cylinder identification is provided on the outer periphery of the cam rotor 20.

Further, a cooling water temperature sensor 22 is provided in a cooling water passage (not shown) forming a riser formed in the intake manifold 3.
Is coming.

Further, as an exhaust system of the engine, a catalyst 25 is interposed downstream of a collecting portion of an exhaust pipe 23 communicating with each exhaust port 2b of the cylinder head 2, and an O2 sensor 24 is disposed upstream of the catalyst 25. Have been.

On the other hand, reference numeral 30 denotes a control device comprising a microcomputer or the like, and a CPU (central processing unit) 31, a ROM 32, a RAM 3
3. The backup RAM 34 and the I / O interface 35 are connected to each other via the bus line 36, and a predetermined stabilized voltage is supplied from the constant voltage circuit 37.

The constant voltage circuit 37 is connected to a battery 39 via a relay contact of a power relay 38, and a relay coil of the power relay 38 is connected to the battery 39 via a key switch 40. The constant voltage circuit 37 is directly connected to the battery 39, and supplies control power to each unit of the control device 30 when a key switch is turned on and the relay contact of the power supply relay 38 is closed. ON / O of key switch 40
Regardless of the FF, power for backup is always supplied to the backup RAM 34, and data such as a learning value KLR stored in a later-described learning value table stored in the backup RAM 34 is retained.

The input ports of the I / O interface 35 are connected to the intake air amount sensor 8, the throttle opening sensor 9a, the crank angle sensor 19, the cam angle sensor 21, the cooling water temperature sensor 22, the O2 sensor 24, and the idle switch 9b. At the same time, the battery voltage is monitored by inputting VB.

On the other hand, to the output port of the I / O interface 35, the ignition plug 11 is connected via an igniter 26, and the injector 10 and the fuel pump 14 are connected via a drive circuit 41.

The ROM 32 stores a control program and fixed data such as a reference inversion time map MPt0 described later. The RAM 33 stores the output values from the sensors after the data processing and the data processed by the CPU 31. Also, the backup RAM 34 has
A learning value table TB LR, which will be described later, is stored. Each learning value KLR stored in the learning value table TB LR for each operation area is held regardless of whether the key switch 40 is ON or OFF. .

The CPU 31 calculates the intake air amount Q from the output signal of the intake air amount sensor 8 and calculates the engine speed N from the output signal of the crank angle sensor 19 according to the control program stored in the ROM 32. And a fuel injection pulse width Ti that determines an optimal fuel injection amount for an operation state based on various data stored in the backup RAM 34.
And the ignition timing is calculated. Then, a drive pulse width signal corresponding to the fuel injection pulse width Ti is output to the injector 10 of the corresponding cylinder at a predetermined timing through the drive circuit 41 to inject fuel, and the calculated ignition timing is reached. At this point, an ignition signal is output, and the ignition plug 11 of the corresponding cylinder is sparked via the igniter 26.

As a result, the air-fuel mixture having a predetermined air-fuel ratio supplied to the combustion chamber of the corresponding cylinder ignites, explodes and combusts, and exhaust gas is discharged from the fuel chamber in the exhaust stroke. Then, a predetermined voltage value signal is output from the O2 sensor 24 disposed upstream of the catalyst 25 in the exhaust pipe 23 according to the oxygen concentration contained in the exhaust gas, that is, the air-fuel ratio state. After this voltage value signal is subjected to waveform shaping processing, it is compared as an O2 sensor output voltage (output value) VAF by the CPU 31 with a slice level VS that determines a reference voltage value corresponding to a stoichiometric air-fuel ratio (stoichiometric). It is determined whether the air-fuel ratio state is on the rich side (VAF.gtoreq.VS) or lean (VAF <VS) with respect to the target air-fuel ratio, that is, the stoichiometric air-fuel ratio. If it is, "1" is stored in a predetermined address of the RAM 33. Then, the CPU 31 monitors the air-fuel ratio at a predetermined time or at predetermined intervals, and reflects the result in the calculation of the fuel injection amount.
Control the air-fuel ratio.

Next, a functional configuration of the control device 30 according to the air-fuel ratio learning control of the present invention will be described with reference to FIG.

The control device 30 includes, as functions related to the air-fuel ratio learning control, an intake air amount calculation unit 50, an engine speed calculation unit 51, a basic fuel injection amount setting unit 52, and various increase correction coefficient setting units 5
3, voltage correction pulse width setting means 54, feedback condition determination means 55, learning condition determination means 56, reference inversion time setting means 57, inversion determination means 58, integration constant correction means 59, air-fuel ratio feedback correction coefficient setting means 60, learning Value updating means 61
a, learning correction coefficient setting means 61b, fuel injection amount setting means 62,
And each functional unit of the injector driving unit 63.

The intake air amount calculating means 50 calculates the intake air amount Q from the output signal of the intake air amount sensor 8, and the engine speed calculating means 51 calculates the engine speed N from the output signal of the crank angle sensor 19.

Based on the intake air amount Q and the engine speed N, the basic fuel injection amount calculating means 52 performs a basic fuel injection pulse width Tp to determine the basic fuel injection amount by map search or calculation.
Is calculated. In this embodiment, the basic fuel injection pulse width Tp is calculated by calculation (Tp = K · Q / N, H is a constant).

In addition, the various increase amount correction coefficient setting means 53 includes a throttle opening θ detected by the throttle opening sensor 9a, ON / OFF of the idle switch 9b (representing a throttle valve fully open and a throttle valve open state), a cooling water temperature sensor 22. Is read, and various increase correction coefficients COEF such as well-known acceleration / deceleration correction, full-open increase correction, increase after idling correction, and cooling water temperature correction are set.

The voltage correction pulse width setting means 54 reads the invalid injection time (pulse width) of the injector 10 from a table (not shown) according to the terminal voltage VB of the battery 39, and sets the voltage correction pulse width TS for interpolating the invalid injection time. I do.

The feedback condition determination means 55 determines whether the O2 sensor 24 is in the active state or the inactive state, and
Based on the cooling water temperature Tw, the engine speed N, and the basic fuel injection pulse width Tp representing the engine load, it is determined whether the air-fuel ratio feedback condition for performing the air-fuel ratio feedback control is satisfied. I do.
The determination whether the O2 sensor 24 is in the active state or the inactive state is made based on, for example, whether or not the output voltage VAF of the O2 sensor is equal to or higher than a set value.

Then, when the O2 sensor 24 is in an inactive state, when the cooling water temperature Tw is equal to or lower than a predetermined value (for example, 50 ° C.) and the engine 1 has not been completely warmed up, or when the engine speed N
Is determined that the air-fuel ratio feedback condition is not satisfied when the full-open increase correction is performed at or above the set rotation speed NS (for example, 5200 rpm) or when the basic fuel injection pulse width Tp is at or above the set value TpS (throttle substantially fully open region). The air-fuel ratio open loop control is performed.

Further, the O2 sensor is in an active state, the cooling water temperature Tw exceeds a predetermined value, and the engine has been completely warmed up, the engine speed N is less than the set speed NS, and the basic fuel injection pulse width Tp is equal to the set value TpS. When the full-open increase correction is not performed because the air-fuel ratio is less than the predetermined value, it is determined that the air-fuel ratio feedback condition is satisfied, and the air-fuel ratio feedback control is performed.

In addition, the learning condition determining means 56 determines that the air-fuel ratio feedback condition is satisfied, and that the engine speed N and the basic fuel injection pulse width Tp specify the operating region corresponding to the learning value table TBLR. Within the parameter range of the matrix MT (see FIG. 3) (N0 ≦ N ≦ Nn, Tp0 ≦ Tp
≤ Tpn) and when the section NEW in the matrix specified by the engine speed N and the basic fuel injection pulse width Tp matches the section position OLD selected last time,
When the engine operation region continues in the same region, it is determined that the learning condition is satisfied.

When the learning condition is satisfied (at this time, the air-fuel ratio feedback condition is also satisfied), the reference reversal time setting means 57 determines, based on the engine speed N and the basic fuel injection pulse width Tp representing the load, By the air-fuel ratio feedback control (air-fuel ratio feedback correction using an air-fuel ratio feedback correction coefficient α described later), a time that can be obtained in a normal state, from the time when the air-fuel ratio shifts from rich to lean, and again until the air-fuel ratio shifts to rich again, or the air-fuel ratio Represents the time it takes to transition from lean to rich and back again.
Set the reference inversion time t0.

During the air-fuel ratio feedback control, the reference inversion time t0 is determined by changing the output value of the O2 sensor to the theoretical air-fuel ratio (stoichiometric)
Based on the time from when the air-fuel ratio detected by comparing with the slice level corresponding to the transition from rich to lean to transition to rich again, or after the transition from lean to rich to transition to lean again. The output characteristics and the response of the O2 sensor 24 are deteriorated due to the deterioration of the O2 sensor 24, the response delay of the air-fuel ratio detection is caused, and the deviation amount of the air-fuel ratio with respect to the stoichiometric air-fuel ratio is large. The determination is performed by using the engine speed N and the basic fuel injection pulse width Tp as parameters to search and set a reference inversion time map MPt0 stored in the ROM 32 in advance. Specifically, the reference inversion time t0 is detected by comparing the output value VAF of the O2 sensor 24 with the slice level VS during the air-fuel ratio feedback control in the normal engine operation state where the O2 sensor 24 is normal. The time is set slightly longer than the air-fuel ratio inversion cycle. For example, at the time of idling, the reversal cycle of the air-fuel ratio is about 1 second, and the reference reversal time t0 at this time.
Is set to t = 2 to 3 sec.

The reference reversal time map MPt0 for setting the reference reversal time t0 is composed of a map having parameters of the engine speed N and the basic fuel injection pulse width Tp representing the load, as shown in FIG. At each address of this map, a reference inversion time t0 obtained in advance by an experiment or the like is stored.

Here, the time from when the air-fuel ratio shifts from rich to lean by the air-fuel ratio feedback correction and until the air-fuel ratio shifts to rich again, and the time from when the air-fuel ratio shifts from lean to rich and then shifts back to lean again, are the engine high-speed and high-load. Is shorter, and the lower the rotation speed and the lower the load, the longer the time.

Accordingly, corresponding to this, the reference inversion time map MP
In t0, the reference inversion time t0 of a shorter time value is stored as the engine rotation speed and the engine load increase, and the engine speed N
In setting the reference reversal time t0 based on the basic fuel injection pulse width Tp representing the load and the load, the reference reversal time t0 is set to a shorter value as the engine speed is higher and the load is higher.

The inversion determining means 58 switches the air-fuel ratio from rich to lean within the reference inversion time t0, based on the air-fuel ratio detected by comparing the O2 sensor output voltage VAF and the slice level VS, or By determining whether there has been a change from lean to rich, the O2 sensor 24
It is determined that the output characteristics and the responsiveness of the O2 sensor 24 are deteriorated due to the deterioration of the air-fuel ratio, the response delay of the air-fuel ratio detection occurs, and the deviation amount of the air-fuel ratio from the stoichiometric air-fuel ratio is large.

Here, the reference inversion time t0 for judging the switching of the air-fuel ratio is set according to the engine operating state based on the engine speed N and the basic fuel injection pulse width Tp representing the load, as described above. Because of the deterioration of the O2 sensor 24
If the output characteristics and responsiveness of the O2 sensor 24 have deteriorated and a response delay of the air-fuel ratio detection has occurred, and the deviation amount of the air-fuel ratio from the stoichiometric air-fuel ratio is large, this is determined for each engine operating state. It is possible to judge appropriately.

In addition, as described above, the time from when the air-fuel ratio shifts from rich to lean due to the air-fuel ratio feedback correction and until the air-fuel ratio shifts back to rich, and the time after the air-fuel ratio shifts from lean to rich and then again to lean, The higher the rotation speed and the higher the load, the shorter the time, and the lower the rotation speed and the load, the longer the time, but the reference reversal time t0 is set to a shorter value correspondingly to the higher the engine rotation speed and the high load. Therefore, when the output characteristics and the responsiveness of the O2 sensor 24 are deteriorated due to the deterioration of the O2 sensor 24, a response delay of the air-fuel ratio detection occurs, and the deviation amount of the air-fuel ratio from the stoichiometric air-fuel ratio is large. In this case, it is possible to make an accurate determination without causing an erroneous determination.

As a result of the discrimination by the reversing discriminating means 58, the integration constant correcting means 59 does not switch the air-fuel ratio within the reference reversing time t0, and the output characteristics and the responsiveness of the O2 sensor 24 are deteriorated due to the deterioration of the O2 sensor 24. If it is determined that the air-fuel ratio has deteriorated and a response delay of the air-fuel ratio detection has occurred and the deviation amount of the air-fuel ratio from the stoichiometric air-fuel ratio is large, the air-fuel ratio continues to be rich or lean. Fuel ratio feedback correction coefficient α
, The integration constant I for gradually decreasing or increasing the air-fuel ratio correction amount is reduced. Specifically, the integral constant I when the air-fuel ratio feedback correction coefficient α is set by the proportional integral process is multiplied by a correction constant Ki (0 <Ki <1) to make the integral constant smaller than usual.

When the air-fuel ratio feedback condition is satisfied, the air-fuel ratio feedback correction coefficient setting means 60, based on the air-fuel ratio detected by the comparison between the O2 sensor output value VAF and the slice level VS, when the fifth condition is established, is well known. As shown in the figure, the air-fuel ratio changes from rich (VAF ≧ VS) to lean (VAF <VSF).
VS) or when a transition is made from lean to rich, a proportional constant P for skipping the air-fuel ratio correction amount is added or subtracted from the previous air-fuel ratio feedback correction coefficient, and when the air-fuel ratio continues rich or lean, the air-fuel ratio is increased. The current air-fuel ratio feedback correction coefficient α is set by a proportional integration process of subtracting or adding an integration constant I for gradually decreasing or increasing the correction amount to the previous air-fuel ratio feedback correction coefficient.

That is, the air-fuel ratio feedback correction coefficient α is incorporated in the calculation formula of the fuel injection pulse width Ti that determines the fuel injection amount described later (Ti = Tp × COEF × KBRRC × α + TS),
When α = 1.0 (reference value), the fuel injection amount is not corrected by the air-fuel ratio feedback correction coefficient α.

Then, as shown in FIG. 5, when the air-fuel ratio transitions from lean to rich, the air-fuel ratio feedback correction coefficient α is skipped in the negative direction by the proportionality constant P to perform fuel reduction correction, and while the rich is continued, By gradually decreasing the air-fuel ratio feedback correction coefficient α by the integration constant I, the fuel reduction correction rate is gradually increased, and the air-fuel ratio is made to converge to a stoichiometric air-fuel ratio at which the catalyst 25 can exhibit purification efficiency. On the other hand, when the air-fuel ratio shifts from rich to lean, on the contrary, the air-fuel ratio feedback correction coefficient α is skipped in the positive direction by the proportionality constant P to increase the fuel increase, and gradually the air-fuel ratio is gradually increased while the lean operation is continued. By increasing the feedback correction coefficient α by the integration constant I, the fuel increase correction rate is gradually increased, and the air-fuel ratio is made to converge to a stoichiometric air-fuel ratio at which the catalyst 25 can exhibit purification efficiency.

Here, as described above, during the air-fuel ratio feedback control, there is no switching of the air-fuel ratio within the reference inversion time t0, and the output characteristics and responsiveness of the O2 sensor 24 deteriorate due to the deterioration of the O2 sensor 24. Therefore, when it is determined that the air-fuel ratio detection is delayed and the deviation of the air-fuel ratio from the stoichiometric air-fuel ratio is large, the air-fuel ratio continues to be rich or lean by the integration constant correcting means 59. At this time, the integration constant I for gradually decreasing or increasing the air-fuel ratio correction amount by the air-fuel ratio feedback correction coefficient α described later is corrected to a value Ki · I smaller than usual, so that the air-fuel ratio feedback correction coefficient The decreasing speed or increasing speed of α is reduced, and the sticking of the air-fuel ratio feedback correction coefficient α to the lower limit or upper limit is suppressed.

When the air-fuel ratio feedback condition is not satisfied, the air-fuel ratio feedback correction coefficient α is fixed to α = 1.0, the fuel injection amount is not corrected by the air-fuel ratio feedback correction coefficient α, that is, the air-fuel ratio is not corrected, and the air-fuel ratio is opened. Loop control is used.

The learning value updating means 61a determines that the learning condition is satisfied, and that the air-fuel ratio has been switched within the reference inversion time t0, and the number of times the air-fuel ratio has been switched has reached the set number Cn (for example, Cn = 3). That is, the engine operating region (section) specified by the above-described matrix MT (see FIG. 3) based on the engine speed N and the basic fuel injection pulse width Tp representing the load is in a steady operation state in which the engine operating region (section) continues in the same region. When it is determined that the air-fuel ratio is repeatedly rich and lean with respect to the stoichiometric air-fuel ratio and converges to the stoichiometric air-fuel ratio by the air-fuel ratio feedback control, the reference value α0 (= 1.0), the learning value table TB LR corresponding to the partition position of the matrix MT according to the deviation
The learning value KLR stored in the corresponding region is updated, and the deviation of the air-fuel ratio feedback correction coefficient α from the reference value α0 is learned for each engine operation region and stored as the learning value KLR.

In the present embodiment, as shown in FIG. 3, the matrix MT for specifying the engine operation region is configured using the engine speed N and the basic fuel injection pulse width Tp representing the load as parameters. Learning value table TB LR
Is configured using only the basic fuel injection pulse width Tp as a parameter, and the ranges Tp0Tp1, Tp1Tp of the basic fuel injection pulse width Tp.
A learning value KLR is stored at each address corresponding to 2, Tp2Tp3,..., Tpn-1 Tpn, and KLR = 1.0 is stored as an initial value of the learning value KLR.

Then, an area (address) of the learning value table TBLR is specified corresponding to the section position of the matrix MT. For example, in FIG.
When it is in the section D1 of the MT, the area (address) of the learning value table TB LR correspondingly corresponds to this, as indicated by hatching in the figure.
Is specified. As a result, the use area of the backup RAM 34 by the learning value table TBLR can be reduced, and the learning is advanced at an early stage, thereby improving the effectiveness of the learning. The learning value table is not limited to this embodiment, and may be configured as a table using the engine speed N and the load as parameters.

At this time, in the present embodiment, a deviation Δα (= α0−αav) of the average value αav of the air-fuel ratio feedback correction coefficient α with respect to the reference value α0 is calculated, and the learning value KLR of the corresponding area is used to calculate the deviation Δα. Predetermined ratio Δα / M of deviation Δα
(M is a constant that determines the learning value update ratio) to update the learning value KLR (KLR ← KLR−Δα / M).

That is, when the average value αav of the air-fuel ratio feedback correction coefficient α is larger than the reference value α0 (= 1.0) and during this time the fuel increase is averaged by the air-fuel ratio feedback correction coefficient α, the deviation Δα is minus Since the update amount (Δα / M) with respect to the learning value KLR is given as a negative term, the update amount of the learning value is given as an addition term. At this time, the learning value KLR is | Δα
| / M is increased, and the air-fuel ratio correction amount (fuel increase rate) based on the air-fuel ratio feedback correction coefficient α is transferred to the learning value KLR.

Conversely, the average value αav of the air-fuel ratio feedback correction coefficient α is smaller than the reference value α0 (= 1.0).
When the fuel reduction is averagely performed by the air-fuel ratio feedback correction coefficient α, the deviation Δα becomes a positive value, and the update amount (Δα / M) for the learning value KLR is given by a negative term. The update amount of the learning value is given by a subtraction term, and at this time, the learning value KLR is Δα /
M is decreased, and the air-fuel ratio correction amount (fuel reduction rate) based on the air-fuel ratio feedback correction coefficient α is transferred to the learning value KLR.

That is, the deviation of the air-fuel ratio feedback correction coefficient α from the reference value is shifted to the learning value KLR. Then, as described later, based on the engine operating state, the learning value table TB
LR is searched, and the learning correction coefficient KBLRC obtained by interpolation calculation of the learning value KLR is incorporated into the arithmetic expression of the fuel injection pulse width Ti that determines the fuel injection amount together with the air-fuel ratio feedback correction coefficient α, so that open loop control is performed. In this case, the air-fuel ratio can be appropriately controlled, and the air-fuel ratio correction amount can be relatively reduced with respect to the reference value based on the air-fuel ratio feedback correction coefficient α during the air-fuel ratio feedback control. Is improved.

Thus, even if the engine operating state changes, the integration speed (generally, several% / second) given by the integration constant I in the proportional integration processing of the air-fuel ratio feedback correction coefficient α immediately follows. The learning correction coefficient KBLRC based on the learning value KLR is reflected on the fuel injection amount, and the air-fuel ratio can be maintained at the stoichiometric air-fuel ratio.

On the other hand, when the air-fuel ratio is not switched within the reference inversion time t0, the deviation Δα (αo−α) between the previous air-fuel ratio feedback correction coefficient αo and the current air-fuel ratio feedback correction coefficient α.
The change state of the air-fuel ratio feedback correction coefficient α is determined based on α). When Δα <0, the current air-fuel ratio feedback correction coefficient α is larger than the previous time, and the air-fuel ratio feedback correction coefficient α is increasing, the learning value KLR in the corresponding area is increased by a predetermined amount KLRSET. (KLR ← KLR + KLRSET), update, while Δα> 0
And the air-fuel ratio feedback correction coefficient α
Is smaller and the air-fuel ratio feedback correction coefficient α is decreasing, the learning value of the corresponding area is set to a predetermined amount KLRSET
It is decreased (KLR ← KLR−KLRSET) and updated.

That is, there is no switching of the air-fuel ratio within the reference inversion time t0, and the output characteristics and responsiveness of the O2 sensor 24 deteriorate due to the deterioration of the O2 sensor 24, and a response delay of the air-fuel ratio detection occurs, and If the deviation of the air-fuel ratio from the stoichiometric air-fuel ratio is large, the learning value KLR is corrected and updated in the same direction as the air-fuel ratio feedback correction coefficient α, so that the air-fuel ratio correction amount based on the air-fuel ratio feedback correction coefficient α can be accurately determined. Then, the air-fuel ratio learning correction can be performed by accurately reflecting the deviation of the air-fuel ratio from the stoichiometric air-fuel ratio to the learning value KLR.

In addition, the output characteristics and responsiveness of the O2 sensor 24 are deteriorated due to the deterioration of the O2 sensor 24, and a response delay of the air-fuel ratio detection occurs, and a deviation amount of the air-fuel ratio from the stoichiometric air-fuel ratio is large. When the air-fuel ratio deviates greatly from the stoichiometric air-fuel ratio,
The learning value KLR is corrected and updated in the same manner as the air-fuel ratio feedback correction coefficient α, and the air-fuel ratio correction amount based on the air-fuel ratio feedback correction coefficient α is shifted to the learning value KLR. Upper limit value (for example, 1.2) that defines the correction limit of the fuel increase rate by
Alternatively, the air-fuel ratio feedback correction coefficient α with respect to a lower limit value (for example, 0.7) that defines the correction limit of the fuel reduction rate based on the air-fuel ratio feedback correction coefficient α when the air-fuel ratio is rich.
The output characteristics and responsiveness of the O2 sensor 24 deteriorate due to the deterioration of the O2 sensor 24, causing a response delay in the air-fuel ratio detection, and a deviation of the air-fuel ratio from the stoichiometric air-fuel ratio. Even when the amount is large, the air-fuel ratio can be corrected using the air-fuel ratio feedback correction coefficient α. The air-fuel ratio correction amount with respect to the fuel injection amount is a learning correction coefficient KLRBC calculated by interpolation between the air-fuel ratio feedback correction coefficient α and the learning value KLR obtained by shifting the air-fuel ratio correction amount based on the air-fuel ratio feedback correction coefficient α. And is determined by
Even when the output characteristics and responsiveness of the O2 sensor 24 deteriorate due to the deterioration of the O2 sensor 24, a response delay of the air-fuel ratio detection occurs, and the deviation amount of the air-fuel ratio with respect to the stoichiometric air-fuel ratio is large. By the synergy of the air-fuel ratio feedback correction coefficient α and the learning correction coefficient KBLRC, the air-fuel ratio correction amount can be set to be large in accordance with a large deviation amount of the air-fuel ratio from the stoichiometric air-fuel ratio. As a result, the output characteristics and responsiveness of the O2 sensor 24 deteriorate due to the deterioration of the O2 sensor 24, and a response delay of the air-fuel ratio detection occurs, and the deviation amount of the air-fuel ratio from the stoichiometric air-fuel ratio is large. In this case, it is possible to converge the air-fuel ratio to the stoichiometric air-fuel ratio.
The air-fuel ratio learning based on the air-fuel ratio feedback correction coefficient α can be restarted.

As a result, the O2 sensor 24
If the output characteristics and responsiveness of the air-fuel ratio are degraded due to the deterioration of the air-fuel ratio detection response and the deviation of the air-fuel ratio from the stoichiometric air-fuel ratio is large, this is accurately determined and the air-fuel ratio feedback is performed. It is possible to prevent the correction coefficient α from sticking to the upper limit value or the lower limit value, and to accurately reflect the difference between the air-fuel ratio and the stoichiometric air-fuel ratio in the learning value KLR to perform the air-fuel ratio learning. The controllability and reliability of the vehicle can be improved, and exhaust emission, fuel efficiency, and drivability can be improved.

Further, in the present embodiment, the state where the air-fuel ratio is not switched within the reference inversion time t0 is shifted to the state where the air-fuel ratio is switched, and the number of times the air-fuel ratio is switched is initially set to the set number (for example, 3 times), that is, as described above, the output characteristics and responsiveness of the O2 sensor 24 deteriorate due to the deterioration of the O2 sensor 24, and a response delay of the air-fuel ratio detection occurs. In addition, when the deviation amount of the air-fuel ratio from the stoichiometric air-fuel ratio is large, the learning value KLR is corrected and updated in the same direction as the air-fuel ratio feedback correction coefficient α, and the air-fuel ratio correction amount based on the air-fuel ratio feedback correction coefficient α is used as the learning value. By shifting to KLR, when the air-fuel ratio converges to approximately the stoichiometric air-fuel ratio,
The air-fuel ratio feedback correction coefficient α during which the air-fuel ratio switching frequency reaches the set frequency after the air-fuel ratio switching.
Of the average value αav from the reference value α0, Δα (= α0
-Αav) is subtracted from the learning value KLR of the corresponding area (the corresponding address), and the learning value KLR of the corresponding area is calculated as the deviation Δα.
Are updated at once (KLR ← KLR−Δα). Then, the air-fuel ratio feedback correction coefficient α is corrected to α = 1.0.

That is, when the air-fuel ratio converges to the stoichiometric air-fuel ratio by the above-described process from a state where the amount of deviation of the stoichiometric air-fuel ratio is large, the entire amount of deviation of the air-fuel ratio feedback correction coefficient α from the reference value α0 is learned at a time. The air-fuel ratio feedback correction coefficient α is shifted to the reference value α0
(= 1.0), the air-fuel ratio feedback controllability is improved, and the learning value KLR is converged early to improve and stabilize the convergence of the learning value KLR, thereby stabilizing the air-fuel ratio learning control. Improve the performance.

The learning correction coefficient setting means 61b searches the learning value table TBLR using the basic fuel injection pulse width Tp representing the load as a parameter corresponding to the current engine operating state, and sets the learning correction coefficient KBLRC by a well-known interpolation calculation. I do.

The fuel injection amount setting means 62 corrects the basic fuel injection pulse width Tp, which determines the basic fuel injection amount, by multiplying the increase correction coefficient COEF by various amounts, and multiplies the air-fuel ratio feedback correction coefficient α. In addition to the air-fuel ratio correction, the air-fuel ratio learning correction is performed by multiplying by the learning correction coefficient KBLRC, and the voltage correction pulse width TS is added to interpolate the invalid injection time, and the final fuel injection supplied to the engine is performed. Set the fuel injection pulse width Ti that determines the amount (Ti = Tp x COEF
× KBLRC × α + TS). Then, at a predetermined timing, a driving pulse width signal corresponding to the fuel injection pulse width Ti is output to the injector 10 of the corresponding cylinder via the injector driving means 63, and the predetermined fuel is injected from the injector 10.

As a result, for example, as shown in FIG.
The output characteristics and responsiveness of the O2 sensor 24 deteriorate due to the deterioration of the 24, and the response delay of the air-fuel ratio detection occurs, and the deviation amount of the air-fuel ratio from the stoichiometric air-fuel ratio is large, and the air-fuel ratio becomes large and rich. In the state of shifting, it is determined that the air-fuel ratio has not been changed within the reference inversion time t0.

Until the reference inversion time t0 is reached (until the time t1 in FIG. 5 is reached), as a result of the comparison between the O2 sensor output voltage VAF and the slice level VS, the rich (VA)
F ≧ VS) is detected, the air-fuel ratio feedback correction coefficient α is gradually reduced by the integration constant I.
During this time, the learning value KLR
Is not updated.

Then, when the air-fuel ratio is not switched within the reference inversion time t0 (from time t1 to time t2 in FIG. 5), the output characteristics and response of the O2 sensor 24 due to the deterioration of the O2 sensor 24 It is determined that the response of air-fuel ratio detection is delayed due to deterioration of the air-fuel ratio and that the amount of deviation of the air-fuel ratio from the stoichiometric air-fuel ratio is large, and the rich air-fuel ratio (or lean air-fuel ratio) is determined.
The integral constant for gradually decreasing (or increasing) the air-fuel ratio feedback correction coefficient α while the state is continued is set to a value Ki · I (0 <Ki <1) smaller than the normal time. As a result, the rate of decrease of the air-fuel ratio feedback correction coefficient α is reduced, and sticking of the air-fuel ratio feedback correction coefficient α to the lower limit value is suppressed.

When the air-fuel ratio is not switched within the reference inversion time t0, the change state of the air-fuel ratio feedback correction coefficient α is further determined by the difference Δα between the previous air-fuel ratio feedback correction coefficient α0 and the current air-fuel ratio feedback correction coefficient α. (= Α
0-α). At this time, Δα> 0
And the air-fuel ratio feedback correction coefficient α
Is smaller and the air-fuel ratio feedback correction coefficient α decreases, and in response to this, the learning value KLR in the corresponding area gradually decreases by a predetermined amount KLRSET while the air-fuel ratio feedback correction coefficient α decreases. Reduced (KLR ← KLR−KLRS
ET).

As a result, the O2 sensor 24
When the output characteristics and response of the air-fuel ratio are deteriorated and the response delay of the air-fuel ratio detection occurs, and the deviation amount of the air-fuel ratio is large with respect to the stoichiometric air-fuel ratio, and the air-fuel ratio is largely shifted to rich, Also, accurately, the air-fuel ratio correction amount based on the air-fuel ratio feedback correction coefficient α is shifted to the learning value KLR, and the fuel reduction rate is corrected using the air-fuel ratio feedback correction coefficient α when the air-fuel ratio richness is continuously detected. The margin of the air-fuel ratio feedback correction coefficient α with respect to the lower limit value defining the limit is secured, and the air-fuel ratio correction using the air-fuel ratio feedback correction coefficient α becomes possible.

At this time, the air-fuel ratio correction amount with respect to the fuel injection amount is a learning value obtained by interpolation calculation of the air-fuel ratio feedback correction coefficient α and the learning value KLR obtained by shifting the air-fuel ratio correction amount based on the air-fuel ratio feedback correction coefficient α. Correction coefficient KBLRC
Therefore, the output characteristics and responsiveness of the O2 sensor 24 deteriorate due to the deterioration of the O2 sensor 24, and a response delay of the air-fuel ratio detection occurs. Even when the deviation amount is large and the air-fuel ratio is largely shifted to rich, the synergy between the air-fuel ratio feedback correction coefficient α and the learning correction coefficient KBLRC corresponds to the large deviation amount of the air-fuel ratio with respect to the stoichiometric air-fuel ratio. , The air-fuel ratio correction amount (fuel injection is set large.

Therefore, the output characteristics and the responsiveness of the O2 sensor 24 are deteriorated due to the deterioration of the O2 sensor 24, and the response delay of the air-fuel ratio detection occurs, and the deviation amount of the air-fuel ratio from the stoichiometric air-fuel ratio is large, Even when the air-fuel ratio is largely shifted to rich, the air-fuel ratio can be converged to the stoichiometric air-fuel ratio.

Then, at time t2, the air-fuel ratio converges to the stoichiometric air-fuel ratio,
The air-fuel ratio feedback correction coefficient α is set by a proportional integration process based on the integration constant I and the proportionality constant P in a normal state.
This is incorporated into the arithmetic expression of the fuel injection pulse width Tp that determines the fuel injection amount, so that the air-fuel ratio feedback control is performed. As a result, the air-fuel ratio alternates between rich and lean near the stoichiometric air-fuel ratio.

Then, switching of the air-fuel ratio is determined by alternately repeating the air-fuel ratio between rich and lean. Then, at this time, the learning condition is satisfied, the state shifts to the state where the air-fuel ratio is switched, and the first time the air-fuel ratio switching number first reaches the set number (for example, three times) (FIG. 5). At the time t3), the deviation Δα of the average value αav of the air-fuel ratio feedback correction coefficient α from the reference value α0 during the period when the air-fuel ratio switching number reaches the set number after the air-fuel ratio switching.
(= Α0−αav) the entire amount of the corresponding area (corresponding address)
Is subtracted from the learning value KLR, and the learning value KLR of the corresponding area is updated at one time according to the above-mentioned deviation Δα (KLR ← KLR−Δ
α). Then, the air-fuel ratio feedback correction coefficient α is calculated as α
Modify to 1.0.

That is, when the air-fuel ratio converges to the stoichiometric air-fuel ratio by the above-described process from a state where the amount of deviation of the stoichiometric air-fuel ratio is large, the entire amount of deviation of the air-fuel ratio feedback correction coefficient α from the reference value α0 is learned at a time. The air-fuel ratio feedback correction coefficient α is shifted to the reference value α0
(= 1.0), the air-fuel ratio feedback controllability is improved, and the learning value KLR is converged early to improve and stabilize the convergence of the learning value KLR, thereby stabilizing the air-fuel ratio learning control. Improve the performance.

If the learning condition continues to be satisfied, the deviation Δα (= α0−αav) of the average value αav of the air-fuel ratio feedback correction coefficient α with respect to the reference value α0 from the learning value KLR in the corresponding region is thereafter determined as usual. The learning value KLR in the corresponding area is updated by subtracting a predetermined ratio Δα / M (M is a constant that determines the learning value updating ratio) (KLR ← KLR−Δα / M).

Therefore, as shown by the broken line in FIG. 5, conventionally, the output characteristics and the responsiveness of the O2 sensor 24 deteriorate due to the deterioration of the O2 sensor 24, and a response delay of the air-fuel ratio detection occurs, and When the deviation of the air-fuel ratio from the stoichiometric air-fuel ratio is large and the air-fuel ratio is largely and richly shifted, the air-fuel ratio feedback correction coefficient α is gradually reduced by the integration constant I corresponding to the normal time. The correction coefficient α sticks to the lower limit, the air-fuel ratio learning is not performed, and the air-fuel ratio remains rich-shifted.According to the present embodiment, the sticking of the air-fuel ratio feedback correction coefficient α to the lower limit is avoided. In addition, there is no switching of the air-fuel ratio within the reference inversion time t0, and the output characteristics and responsiveness of the O2 sensor 24 deteriorate due to the deterioration of the O2 sensor 24, and a response delay of the air-fuel ratio detection occurs, and , Air-fuel against stoichiometric air-fuel ratio If the deviation amount is large, the learning value KLR is corrected and updated in the same direction as the air-fuel ratio feedback correction coefficient α, so that the air-fuel ratio correction amount based on the air-fuel ratio feedback correction coefficient α is accurately shifted to the learning value KLR. The air-fuel ratio learning correction can be performed by accurately reflecting the deviation of the air-fuel ratio from the stoichiometric air-fuel ratio in the learning value KLR.

A learning correction coefficient KBLRC obtained by interpolation calculation of the air-fuel ratio feedback correction coefficient α and the learning value KLR
Are included in the calculation formula of the fuel injection pulse width Ti that determines the fuel injection amount, so that the output characteristics and responsiveness of the O2 sensor 24 are deteriorated due to the deterioration of the O2 sensor 24, and the response delay of the air-fuel ratio detection is reduced. And the air-fuel ratio is largely shifted from the stoichiometric air-fuel ratio to the stoichiometric air-fuel ratio, and the air-fuel ratio is largely shifted to rich, it is possible to converge the air-fuel ratio to the stoichiometric air-fuel ratio.

Conversely, even when the air-fuel ratio is largely shifted to a lean state, the air-fuel ratio can be similarly converged to the stoichiometric air-fuel ratio.

Next, the control procedure executed by the control device 30 will be described in the sixth.
This will be described with reference to the flowcharts shown in FIGS.

FIG. 6 is a flowchart showing a fuel injection control procedure, which is repeated, for example, at predetermined intervals synchronized with engine rotation.

First, in step S101, output signals from the crank angle sensor 19 and the intake air amount sensor 8 are read, and an engine speed N and an intake air amount Q are calculated.

Next, in step S102, a basic fuel injection pulse width Tp that determines the basic fuel injection amount is calculated from the engine speed N and the intake air amount Q (Tp = K × Q / N, where K is a constant).

Then, the process proceeds to step S103, where the throttle opening sensor
Throttle opening θ detected by 9a, ON / OFF of idle switch 9b (represents throttle valve fully closed, throttle valve open state), cooling water temperature detected by cooling water temperature sensor 22
Tw is read, and in the subsequent step S104, well-known acceleration / deceleration correction,
Various increase correction coefficients COEF such as full-open increase correction, post-idle increase correction, and cooling water temperature correction are set.

In step S105, a voltage correction pulse width TS for interpolating the invalid injection time of the injector 10 is set based on the terminal voltage VB of the battery 39.

Next, in step S106, the learning value table TBLR is searched using the basic fuel injection pulse width Tp representing the load as a parameter, and a learning correction coefficient KBLRC is set by well-known interpolation calculation. In the subsequent step S107, air-fuel ratio feedback correction described later is performed. The air-fuel ratio feedback correction coefficient α set in the coefficient setting program and stored in the RAM 33 is read, and the process proceeds to step S108.

In step S108, the basic fuel injection pulse width Tp is
Correction is made by multiplying the above-mentioned various increase correction coefficients COEF, and
The air-fuel ratio is corrected by multiplying the air-fuel ratio feedback correction coefficient α, the air-fuel ratio is corrected by multiplying the learning correction coefficient KBLRC, and the voltage correction pulse TS is added to interpolate the invalid injection time, and the engine Set the fuel injection pulse width Ti that determines the final fuel injection amount to be supplied to the
= Tp × COEF × KBLRC × α + TS).

Then, in step S109, at a predetermined timing, a drive pulse width signal corresponding to the fuel injection pulse width Ti,
The fuel is output to the injector 10 of the corresponding cylinder via the drive circuit 41, and a predetermined amount of fuel is injected from the injector 10.

Next, the procedure for setting the air-fuel ratio feedback correction coefficient α will be described with reference to the flowchart of FIG.

The program of the procedure for setting the air-fuel ratio feedback correction coefficient α is a program repeated every predetermined time or every predetermined cycle. First, in step S201, it is determined whether the air-fuel ratio feedback condition is satisfied. That is, as described above, when the O2 sensor 24 is in the inactive state, when the cooling water temperature Tw is equal to or lower than a predetermined value (for example, 50 ° C.), and the warming-up of the engine 1 is not completed, or It is determined that the air-fuel ratio feedback condition is not satisfied when N is equal to or greater than the set rotational speed NS (for example, 5200 rpm) or when the basic fuel injection pulse width Tp is equal to or greater than the set value TpS (throttle substantially fully open region) and the full-open increase correction is performed. When the air-fuel ratio feedback condition is not satisfied, the process proceeds to step S216,
The air-fuel ratio feedback correction coefficient α is fixed at α = 1.0 (reference value), and the program exits. Therefore, when the air-fuel ratio feedback condition is not satisfied, the fuel injection amount is not corrected by the air-fuel ratio feedback correction coefficient α, that is, the air-fuel ratio is not corrected, and the air-fuel ratio open loop control is performed.

On the other hand, in step S201, the O2 sensor 24 is in the active state, the cooling water temperature Tw has exceeded the predetermined value, the engine has been warmed up, the engine speed N is less than the set speed NS, and the basic fuel injection pulse If the width Tp is less than the set value TpS and the full-open increase correction is not performed, it is determined that the air-fuel ratio feedback condition is satisfied, and the process proceeds to step S202. In step S202 and below, the output voltage (output value) of the O2 sensor 24
The air-fuel ratio feedback correction coefficient α is set by a proportional integration process according to the air-fuel ratio state detected by comparing VAF and the slice level VS.

In step S202, the O2 sensor output value VAF is read, and in subsequent step S203, the O2 sensor output value VAF is compared with the slice level VS to determine whether the air-fuel ratio is rich or lean.

If VAF ≧ VS and the air-fuel ratio is rich, the process proceeds to step S204, and the air-fuel ratio inversion initial determination flag FLAG1 is referred to. The air-fuel ratio inversion initial discrimination flag FLAG1 is a flag for determining the first time when the air-fuel ratio is inverted from lean to rich, or the first time when the air-fuel ratio is inverted from rich to lean, and the air-fuel ratio is inverted from lean to rich. Is changed from 1 to 0 after the change, and 0 to 1 after the change from rich to lean.
It is said.

Therefore, when the air-fuel ratio is rich and FLAG1 = 1, it is the first time that the air-fuel ratio has reversed from lean to rich, so the process proceeds from step S204 to step S205, and the air-fuel ratio feedback correction coefficient α is subtracted by the proportionality constant P. In the direction (α ← α-P), and step S209
To clear the air-fuel ratio reversal first-time determination flag FLAG1 (FLA
G1 ← 0), exit the program.

When the air-fuel ratio is rich and FLAG1 = 0,
Since Rich is connected, the above step S2
From 04, the process proceeds to step S206, where the integration constant correction flag FLAG2
See The integration constant correction flag FLAG2 is set by a learning value update program described later. The air-fuel ratio is switched within the reference inversion time t0, and the air-fuel ratio becomes rich near the logical air-fuel ratio by the air-fuel ratio feedback correction. During the air-fuel ratio feedback control, there is no switching of the air-fuel ratio within the reference reversal time t0, and the output characteristics and responsiveness of the O2 sensor 24 are deteriorated due to the deterioration of the O2 sensor 24. This flag is set when it is determined that the air-fuel ratio has deteriorated and the response delay of the air-fuel ratio detection has occurred, and the deviation of the air-fuel ratio from the stoichiometric air-fuel ratio is large.

Therefore, when FLAG2 = 0 and the predetermined air-fuel ratio is alternately repeated rich and lean by the air-fuel ratio feedback control, the process proceeds from step S206 to step S207, and the air-fuel ratio feedback correction coefficient α is
It is gradually decreased by the integration constant I every time the program is executed (α
← α-I), through the above step S209, exit the program.

On the other hand, in step S203, when VAF <VS and the air-fuel ratio is lean, the process proceeds to step S210, and similarly,
Reference is made to the air-fuel ratio inversion initial determination flag FLAG1. And
When the air-fuel ratio is lean and FLAG1 = 0, since the air-fuel ratio is the first time when the air-fuel ratio is inverted from rich to lean, the process proceeds from step S210 to step S211 and the air-fuel ratio feedback correction coefficient α is skipped in the positive direction by the proportionality constant P. (Α ← α + P), and in step S215, the air-fuel ratio inversion initial determination flag FLAG1 is set (FLAG1 ← 1), and the program exits.

When the air-fuel ratio is lean and FLAG1 = 1,
Since the lean is connected, the above step S2
The process proceeds from step 10 to step S212, where the integration constant correction flag FLAG2 is set.
See

Then, when FLAG2 = 0 and the air-fuel ratio is alternately repeatedly changed between rich and lean by the air-fuel ratio feedback control, the process proceeds from step S212 to step S213, where the air-fuel ratio feedback correction coefficient α is
It is gradually increased by the integration constant I every time the program is executed (α
← α + I), through the above step S215, exit the program.

That is, when the air-fuel ratio is rich, the air-fuel ratio feedback correction coefficient α is caused by the proportional integration process,
When the air-fuel ratio is lean, the air-fuel ratio feedback correction coefficient α is increased. Then, by incorporating the air-fuel ratio feedback correction coefficient α into the equation for calculating the fuel injection pulse width that determines the fuel injection amount, the fuel injection amount is reduced and corrected when the air-fuel ratio is rich, and when the air-fuel ratio is lean. The fuel injection amount is increased and corrected, whereby the air-fuel ratio is controlled to converge on the stoichiometric air-fuel ratio.

Here, as described above, when FLAG2 = 1, during the air-fuel ratio feedback control, there is no switching of the air-fuel ratio within the reference inversion time t0, and the O2 sensor 24
When the output characteristics and responsiveness of the 24 are deteriorated and the response delay of the air-fuel ratio detection occurs, and when it is determined that the deviation amount of the air-fuel ratio from the stoichiometric air-fuel ratio is large, the rich state of the air-fuel ratio continues. If the air-fuel ratio is detected to be large and rich, the process proceeds from step S206 to step S208, where the integral constant Ki · I smaller than the normal value obtained by multiplying the correction coefficient Ki (0 <Ki <1) The air-fuel ratio feedback correction coefficient α is gradually reduced every time the program is executed (α ← α−Ki
-I), through the above step S209, exit the program.

As a result, the decreasing speed of the air-fuel ratio feedback correction coefficient α is reduced, and the sticking of the air-fuel ratio feedback correction coefficient α to the lower limit value is controlled.

If FLAG2 = 1 and the deviation of the air-fuel ratio from the stoichiometric air-fuel ratio is large and the lean state of the air-fuel ratio is continuously detected and the air-fuel ratio is largely shifted to lean, the process proceeds from step S212 to step S212. Proceeding to S214, the correction coefficient Ki (0 <
The air-fuel ratio feedback correction coefficient α is gradually increased every time the program is executed (α ← α + Ki · I) by an integration constant Ki · I smaller than usual multiplied by Ki <1).
After S215, exit the program. Therefore, also in this case, the increasing speed of the air-fuel ratio feedback correction coefficient α is reduced, and the sticking of the air-fuel ratio feedback correction coefficient α to the upper limit is suppressed.

Next, the program of the learning value update procedure will be described with reference to the flowchart of FIG.

The program of the update procedure of the learning value KLR is repeated at a predetermined time or at predetermined intervals. First, in step S301,
It is determined whether the above-described air-fuel ratio feedback condition is satisfied, and when the air-fuel ratio feedback condition is not satisfied,
When the program exits the program without updating the learning value KLR and the air-fuel ratio feedback condition is satisfied, the process proceeds to step S302.
In step S303, the engine speed N and the basic fuel injection pulse width Tp are read from the learning value table TB.
Matrix M for specifying operating area corresponding to LR
Within the parameter range of T (see FIG. 3) (N0 ≦ N ≦ Nn, Tp0
≤ Tp ≤ Tpn).

When it is outside the range of the matrix, the program exits the program without updating the learning value. When the engine speed N and the basic fuel injection pulse width Tp are within the parameter range of the matrix MT (N0 ≦ N ≦ Nn, Tp0 ≤Tp≤T
pn) Further, in step S304, it is determined whether or not the current section NEW in the matrix specified by the engine speed N and the basic fuel injection pulse width Tp matches the section position OLD selected last time.

If the current section NEW does not match the previous section OLD and the engine is in a transient operation state, the section identified this time is determined in step S305 without updating the learning value KLR.
NEW is the data of the previous section OLD, and the following step S306
And a count value C for counting the number of times the air-fuel ratio is switched.
Is cleared (C ← 0), and the program exits. When the first program is executed, since there is no previous section data, the process jumps from step S303 to step S305, and exits the program via step S306.

On the other hand, when the previous and current sections match (NEW = OLD) and the engine operation area continues in the same area, it is determined that the learning condition is satisfied, and the process proceeds from step S304 to step S307, where the engine speed N The reference inversion time map MPt0 is searched using the basic fuel injection pulse width Tp representing the load as a parameter, and the output characteristics and response of the O2 sensor 24 are deteriorated due to the deterioration of the O2 sensor 24, and the response of the air-fuel ratio detection is reduced. A reference reversal time t0 is set for determining whether or not a delay has occurred and the deviation of the air-fuel ratio from the stoichiometric air-fuel ratio is large based on whether or not the air-fuel ratio has been switched.

Then, in the subsequent step S308, it is determined whether the air-fuel ratio has been switched from rich to lean or from rich to lean within the reference inversion time T0, and when there is no air-fuel ratio switching within the reference inversion time t0, The output characteristics and responsiveness of the O2 sensor 24 are deteriorated due to the deterioration of the O2 sensor 24, causing a response delay of the air-fuel ratio detection, and the deviation amount of the air-fuel ratio from the stoichiometric air-fuel ratio is large. Then, the process proceeds to step S309, and the learning value KLR of the corresponding area is updated according to the change state of the air-fuel ratio feedback correction coefficient α by the processing of steps S309 to S317.

In step S309, an integral constant correction flag FLAG2 for instructing the above-described integral constant decrease correction is set (FLAG2 ←
1) In the following step S310, a deviation Δα between the previous air-fuel ratio feedback correction coefficient αo and the current air-fuel ratio feedback correction coefficient α is calculated (Δα = αo−α).

Then, based on the deviation Δα, steps S311, S31
In step 4, the change state of the air-fuel ratio feedback correction coefficient α is determined. Then, in step S311, when Δα <0,
When the current air-fuel ratio feedback correction coefficient α is larger than the previous time and the air-fuel ratio feedback correction coefficient α is increasing, the process proceeds to step S312, and the learning value table TB LR is searched using the basic fuel injection pulse width Tp as a parameter. Then, in the following step S313, the learning value KLR of the corresponding area (corresponding address) is updated by increasing the predetermined value KLRSET (KLR ←
(KLR + KLRSET), and in step S317, an initial discrimination flag FLAG3 for judging the first transition from a state in which the air-fuel ratio has not been switched within the reference inversion time t0 to a state in which the air-fuel ratio has been switched is set (FLAG3). ← 1), exit the program.

On the other hand, if Δα ≧ 0 in step S311, a decrease in the air-fuel ratio feedback correction coefficient is determined in step S314. When Δα> 0 and the current air-fuel ratio feedback correction coefficient α is smaller than the previous time and the air-fuel ratio feedback correction coefficient α is smaller,
Proceeding to step S315, the learning value table TBLR is searched using the basic fuel injection pulse width Tp as a parameter, and in the following step S316, the learning value KLR of the corresponding area (corresponding address) is
Update by decreasing the predetermined amount KLRSET (KLR ← KLR−KLRSE
T) After the above step S317, the program exits.

When Δα = 0 and there is no change in the air-fuel ratio feedback correction coefficient α, the learning value KLR is not updated, and
The program exits from step S314.

As a result, there is no switching of the air-fuel ratio within the reference inversion time t0, and the output characteristics and responsiveness of the O2 sensor 24 deteriorate due to the deterioration of the O2 sensor 24, and a response delay of the air-fuel ratio detection occurs. When the deviation of the air-fuel ratio from the stoichiometric air-fuel ratio is large, the learning value KLR is corrected and updated in the same direction as the air-fuel ratio feedback correction coefficient α, so that the air-fuel ratio correction amount based on the air-fuel ratio feedback correction coefficient α is calculated. By appropriately shifting the learning value KLR, the deviation of the air-fuel ratio from the stoichiometric air-fuel ratio can be accurately reflected on the learning value KLR to perform the air-fuel ratio learning correction.

On the other hand, if the air-fuel ratio has been switched within the reference reversal time t0 in step S308, step S3
Proceeding to 18, the process of step S318 and subsequent steps learns the deviation of the air-fuel ratio feedback correction coefficient α from the reference value α0 for each engine operation region and stores it as a learned value.

In step S318, the integral constant correction flag FLAG2 for instructing the decrease correction of the integral constant is cleared (FLAG ← 0), and in the following step S319, the count value C for counting the number of air-fuel ratio switching is incremented (C ← C + 1).

Next, in step S320, the count value C is set to the set value Cn.
(For example, Cn = 3), the air-fuel ratio switching circuit is a setting circuit determined by the set value Cn (ie, three times)
Is determined, and if C <Cn and the air-fuel ratio switching count has not reached the set count, the program exits without updating the learned value KLR.

In step S320, when C ≧ Cn and the number of switching of the air-fuel ratio reaches the set number, that is, based on the engine speed N and the basic fuel injection pulse width Tp representing the load, the above matrix MT (FIG. 3) The engine operating area (section) specified by the reference air-fuel ratio is a steady operating state in which the engine operating area (section) continues in the same area. When it is determined that convergence has been achieved, the process proceeds to step S321, and the reference value α0 of the air-fuel ratio feedback correction coefficient α (=
The learning value KLR stored in the corresponding area of the learning value table TBLR is updated according to the deviation from 1.0).

In step S321, the count value C is cleared in preparation for the next learning value update (C ← 0), and in the following step S322, the average value of the air-fuel ratio feedback correction coefficient α during the time when the number of air-fuel ratio switching reaches the set number. The deviation Δα of αav from the reference value α0 is calculated (Δα = α0−αav), and step S3
At 23, the learning value table TBLR is searched using the basic fuel injection pulse width Tp as a parameter.

Then, in a succeeding step S324, an initial determination flag FLAG3
From FLAG3 = 1 to a state in which there is no air-fuel ratio switching within the reference inversion time t0, and a state in which there is air-fuel ratio switching, and the first air-fuel ratio switching count is the set number (C
≧ Cn), the process proceeds to step S326 for the first time,
The total amount of the deviation Δα calculated in step S322 is
The learning value KLR of the corresponding area (corresponding address) is subtracted from the learning value KLR, and the learning value KLR of the corresponding area is updated at one time according to the above-mentioned deviation Δα (KLR ← KLR−Δα).

Then, in a succeeding step S327, the air-fuel ratio feedback correction coefficient α is corrected to α = 1.0, and in a step S328, the first-time determination flag FLAG3 is cleared (FLAG ← 0), and the program exits.

That is, when the air-fuel ratio converges to the stoichiometric air-fuel ratio by the above-described process from a state in which the amount of deviation of the stoichiometric air-fuel ratio from the air-fuel ratio is large, the total amount of deviation of the air-fuel ratio feedback correction coefficient α from the reference value is determined at a time by the learning value. KLR, the air-fuel ratio feedback correction coefficient α is set to the reference value α0 (=
1.0), the air-fuel ratio feedback controllability is improved, and the learning value KLR is converged early.
The convergence of the learning value KLR is improved and stabilized, and the stability of the air-fuel ratio learning control is improved.

Then, in the above step S324, FLAG3 = 0,
When the air-fuel ratio switching state continues,
The process proceeds from step S324 to step S325, and subtracts a predetermined ratio Δα / M (M is a constant for determining a learning value update ratio) of the deviation Δα calculated in step S322 from the learning value KLR of the corresponding area, The learning value KLR is updated (KLR ←
KLR−Δα / M), and the program exits through the step S328.

Then, the air-fuel ratio feedback correction coefficient α set as described above and the learning correction coefficient KBLRC obtained by interpolation calculation of the learning value KLR are used to calculate the fuel injection pulse width Ti in step S108 of the above-described fuel injection control procedure. By incorporating it into the equation, the fuel injection amount can be appropriately corrected, the controllability and reliability of the air-fuel ratio learning control can be improved, and the exhaust emission, the fuel consumption, and the drivability can be improved. Effect] As described above, according to the first aspect of the present invention,
It has the following effects.

(1) Based on the engine speed and the basic fuel injection amount, the time from when the fuel ratio shifts from rich to lean by the air-fuel ratio feedback correction using the air-fuel ratio feedback correction coefficient until the fuel ratio shifts to rich again, and from lean to rich. After the shift, a reference inversion time corresponding to the normal time, which represents at least one of the time until the shift to the lean again, is set.
Then, based on the O2 sensor output, it is determined whether or not at least one of the switching of the air-fuel ratio from rich to lean and the switching from lean to rich within the reference inversion time is determined. Due to deterioration of O2 sensor
Since it is determined whether the output characteristics and response of the O2 sensor have deteriorated and the response delay of the air-fuel ratio detection has occurred and the deviation of the air-fuel ratio from the stoichiometric air-fuel ratio is large, the air-fuel ratio switching is performed. Since the reference reversal time for judging the difference is set according to the engine operating state based on the engine speed and the basic fuel injection amount, the output characteristics and responsiveness of the O2 sensor due to the deterioration of the O2 sensor are reduced. When the response is delayed due to the deterioration of the air-fuel ratio and the deviation of the air-fuel ratio from the stoichiometric air-fuel ratio is large, this can be appropriately determined for each engine operating state.

(2) When the air-fuel ratio is switched within the reference reversal time, and when the air-fuel ratio switching frequency reaches the set frequency, it is determined that the air condition is in the normal state. The deviation of the air-fuel ratio feedback correction coefficient from the reference value is learned for each engine operating region and stored as a learned value. On the other hand, when there is no air-fuel ratio change within the reference reversal time,
Due to the deterioration of the O2 sensor, the output characteristics and responsiveness of the O2 sensor have deteriorated, causing a response delay in the air-fuel ratio detection, and
It is determined that the deviation of the air-fuel ratio from the stoichiometric air-fuel ratio is large, and the air-fuel ratio feedback correction coefficient is gradually reduced or increased when the rich air-fuel ratio or the lean air-fuel ratio state is continued. The lowering value of the air-fuel ratio feedback correction coefficient that defines the limit of the fuel reduction correction rate due to the air-fuel ratio feedback correction because the integration constant is made smaller than normal and the reduction rate or increase rate of the air-fuel ratio feedback correction coefficient is reduced. Sticking to the upper limit of the air-fuel ratio feedback correction coefficient that regulates the limit of the fuel increase correction rate, or sticking to the lower limit or upper limit of the air-fuel ratio. It is possible to prevent the learning from being stopped, thereby improving the controllability of the air-fuel ratio learning control.

(3) When the air-fuel ratio is not switched within the reference reversal time, the change state of the air-fuel ratio feedback correction coefficient is further determined. When the air-fuel ratio feedback correction coefficient is increasing, the learning value of the corresponding area is determined. Is increased by a predetermined amount,
On the other hand, when the air-fuel ratio feedback correction coefficient is decreasing, the learning value of the corresponding area is decreased by a predetermined amount. Then, the basic fuel injection amount set based on the engine operating state is calculated by interpolating at least the air-fuel ratio feedback correction coefficient and a learning value learned for each engine operating region in accordance with the current engine operating state. Since the correction is performed using the correction coefficient and the final fuel injection amount is set, the output characteristics and responsiveness of the O2 sensor deteriorate due to the deterioration of the O2 sensor, and a response delay of the air-fuel ratio detection occurs, and If the deviation of the air-fuel ratio from the stoichiometric air-fuel ratio is large, this is accurately determined, and the air-fuel ratio correction amount based on the air-fuel ratio feedback correction coefficient is accurately shifted to the learning value.
The air-fuel ratio learning correction can be performed by accurately reflecting the deviation of the air-fuel ratio from the stoichiometric air-fuel ratio in the learning value. In addition, the output characteristics and responsiveness of the O2 sensor deteriorate due to the deterioration of the O2 sensor, causing a response delay in the air-fuel ratio detection, and a large deviation amount of the air-fuel ratio from the stoichiometric air-fuel ratio. Is greatly deviated from the stoichiometric air-fuel ratio, the learning value is corrected and updated in the same direction as the air-fuel ratio feedback correction coefficient, and the air-fuel ratio correction amount based on the air-fuel ratio feedback correction coefficient is shifted to the learning value, whereby the air-fuel ratio feedback The margin of the air-fuel ratio feedback correction coefficient is secured with respect to the upper limit value that defines the correction limit of the fuel increase rate by the correction coefficient or the lower limit value that defines the correction limit of the fuel reduction rate by the air-fuel ratio feedback correction coefficient, and the O2 sensor is deteriorated. As a result, the output characteristics and responsiveness of the O2 sensor deteriorate, causing a delay in air-fuel ratio detection response, and the deviation of the air-fuel ratio from the stoichiometric air-fuel ratio is large. In case, it is possible to air-fuel ratio correction by the air-fuel ratio feedback correction coefficient. The air-fuel ratio correction amount for the fuel injection amount is determined by the air-fuel ratio feedback correction coefficient and a learning correction coefficient obtained by interpolation calculation of a learning value obtained by shifting the air-fuel ratio correction amount based on the air-fuel ratio feedback correction coefficient. Therefore, when the output characteristics and responsiveness of the O2 sensor are deteriorated due to the deterioration of the O2 sensor, a response delay of the air-fuel ratio detection occurs, and the deviation amount of the air-fuel ratio with respect to the stoichiometric air-fuel ratio is large. Also, by the synergy of the air-fuel ratio feedback correction coefficient and the learning correction coefficient,
In response to the large deviation of the air-fuel ratio from the stoichiometric air-fuel ratio,
The air-fuel ratio correction amount can be set large. As a result, when the output characteristics and responsiveness of the O2 sensor are deteriorated due to the deterioration of the O2 sensor, a response delay of the air-fuel ratio detection occurs, and when the deviation amount of the air-fuel ratio with respect to the stoichiometric air-fuel ratio is large, Also, the air-fuel ratio can be converged to the stoichiometric air-fuel ratio, and the air-fuel ratio learning based on the air-fuel ratio feedback correction coefficient can be surely restarted. This allows O2
If the output characteristics and responsiveness of the O2 sensor deteriorate due to the deterioration of the sensor and the response delay of the air-fuel ratio detection occurs, and the deviation amount of the air-fuel ratio from the stoichiometric air-fuel ratio is large, Precisely determine and prevent the air-fuel ratio feedback correction coefficient α from sticking to the upper limit or lower limit, and accurately reflect the deviation between the air-fuel ratio and the stoichiometric air-fuel ratio in the learning value to perform air-fuel ratio learning. It is possible to appropriately correct the fuel injection amount, improve the controllability and reliability of the air-fuel ratio learning control, and improve the exhaust emission, fuel consumption, and drivability.

According to the second aspect of the present invention, it has the same effects as the effects (1) and (3) of the first aspect of the invention, and replaces the effect (2) of the first aspect of the invention. It has the following effects.

(4) When the air-fuel ratio is switched within the reference reversal time, the state shifts from the state where the air-fuel ratio is not switched to the state where the air-fuel ratio is switched, and the number of times the air-fuel ratio is switched is set first. In the first time when the number of times the air-fuel ratio has been switched has been reached, the total amount of deviation of the average value of the air-fuel ratio feedback correction coefficient from the reference value during the period in which the number of times the air-fuel ratio has been switched reaches the set number is calculated in the corresponding region. The learning value of the corresponding area is subtracted from the learning value of the corresponding region, and the learning value of the corresponding area is updated at a time in accordance with the deviation, and in response to this, the air-fuel ratio feedback correction coefficient is corrected to the reference value. Can reduce the air-fuel ratio correction amount, thereby improving the air-fuel ratio feedback controllability and improving the convergence of the learning value. Thereafter, while the switching of the air-fuel ratio is connected, the predetermined ratio of the deviation of the average value of the air-fuel ratio feedback correction coefficient from the reference value is subtracted from the learning value of the corresponding region, and the learning of the corresponding region is performed. Since the value is updated, the learning value can be stabilized. Therefore, it is possible to achieve both improvement of the convergence of the learning value by converging the learning value at an early stage and stabilization of the learning value, thereby improving both the air-fuel ratio feedback controllability and the air-fuel ratio learning controllability. can do.

According to the third aspect of the invention, the following effects are obtained in addition to the effects of the first or second aspect of the invention.

(5) The time from when the air-fuel ratio shifts from rich to lean by the air-fuel ratio feedback correction using the air-fuel ratio feedback correction coefficient until the air-fuel ratio shifts back to rich, and the time from when the air-fuel ratio shifts from lean to rich and back to lean again. ,
The shorter the engine speed and the higher the load, the shorter the time. Accordingly, in response to this, when setting the reference inversion time based on the engine speed and the basic fuel injection amount, the reference inversion time is set to a shorter value as the engine rotation speed is higher and the load is higher. The output characteristics and responsiveness of the O2 sensor are deteriorated due to the deterioration of the air-fuel ratio, and the response delay of the air-fuel ratio detection has occurred, and the deviation of the air-fuel ratio from the stoichiometric air-fuel ratio is large. Without this, it is possible to make an accurate determination, and there is an effect that the reliability of the air-fuel ratio learning control can be further improved.

[Brief description of the drawings]

1 shows an embodiment of the present invention, FIG. 1 is a block diagram showing a functional configuration of a control device, FIG. 2 is an overall schematic diagram of an engine, and FIG. 3 shows a steady state determination matrix and a learning value table. FIG. 4 is an explanatory diagram showing a reference inversion time map, and FIG. 5 is an air-fuel ratio state and an O2 sensor output voltage.
A time chart showing a setting state of an air-fuel ratio feedback correction coefficient and an update state of a learning value; FIG. 6 is a flowchart showing a fuel injection control procedure; FIG. 7 is a flowchart showing a setting procedure of an air-fuel ratio feedback correction coefficient; FIG. 8 is a flowchart showing a learning value updating procedure. 1 ... Engine 8 ... Intake air amount sensor 10 ... Injector 19 ... Crank angle sensor 24 ... O2 sensor 30 ... Control device 57 ... Reference inversion time setting means 58 ... Inversion determination means 59 ... Integration constant Correcting means 60 Air-fuel ratio feedback correction coefficient setting means 61a Learning value updating means 62 Fuel injection amount setting means

Claims (3)

(57) [Claims] An output value of an O2 sensor is compared with a slice level for determining an air-fuel ratio state to determine an air-fuel ratio state,
When the air-fuel ratio shifts from rich to lean or from lean to rich, a proportional constant for skipping the air-fuel ratio correction amount is added or subtracted from the previous air-fuel ratio feedback correction coefficient, and the air-fuel ratio continues to be rich or lean. The current air-fuel ratio feedback correction coefficient is set by a proportional integration process of subtracting or adding an integration constant for gradually decreasing or increasing the air-fuel ratio correction amount to the previous air-fuel ratio feedback correction coefficient, and setting the air-fuel ratio feedback. The deviation of the correction coefficient from the reference value is learned for each engine operating region and stored as a learning value, and the basic fuel injection amount set based on the engine operating state is calculated based on at least the air-fuel ratio feedback correction coefficient and each engine operating region. An interpolator is used to calculate the learned values according to the current engine operating conditions. In the air-fuel ratio learning control device for an engine, which sets the final fuel injection amount, the air-fuel ratio shifts from rich to lean by the air-fuel ratio feedback correction using the air-fuel ratio feedback correction coefficient. After that, the reference judgment time corresponding to the normal time, which represents at least one of the time until the transition to the rich again and the time from the transition from the lean to the rich to the transition to the lean again, is defined as the engine speed and the basic fuel injection amount. Based on the O2 sensor output value, at least one of the switching of the air-fuel ratio from rich to lean, and the switching from lean to rich, based on the O2 sensor output value. Reversal determining means for determining whether or not the air-fuel ratio has been switched within the reference reversal time.
An integration constant correction means for reducing the integration constant for gradually decreasing or increasing the air-fuel ratio feedback correction coefficient when the rich air-fuel ratio or the lean air-fuel ratio state is continued, to a value smaller than a normal time; In the case where the air-fuel ratio is switched within the time, and when the number of switching of the air-fuel ratio reaches the set number, the deviation of the air-fuel ratio feedback correction coefficient from the reference value is learned for each engine operating region. When the air-fuel ratio is not switched within the reference inversion time, the change state of the air-fuel ratio feedback correction coefficient is determined. When the air-fuel ratio feedback correction coefficient is increasing, the learning of the corresponding area is determined. Learning value updating means for increasing the value by a predetermined amount and reducing the learning value of the corresponding area by a predetermined amount when the air-fuel ratio feedback correction coefficient is decreasing; The basic fuel injection quantity set based on the engine operating state, at least the air-fuel ratio feedback correction coefficient,
A fuel injection amount setting means for correcting the learning value learned for each engine operation region with a learning correction coefficient obtained by interpolation calculation corresponding to the current engine operation state, and setting a final fuel injection amount. An air-fuel ratio learning control device for an engine, comprising: 2. An air-fuel ratio state is determined by comparing an output value of the O2 sensor with a slice level for determining an air-fuel ratio state.
When the air-fuel ratio shifts from rich to lean or from lean to rich, a proportional constant for skipping the air-fuel ratio correction amount is added or subtracted from the previous air-fuel ratio feedback correction coefficient, and the air-fuel ratio continues to be rich or lean. The current air-fuel ratio feedback correction coefficient is set by a proportional integration process of subtracting or adding an integration constant for gradually decreasing or increasing the air-fuel ratio correction amount to the previous air-fuel ratio feedback correction coefficient, and setting the air-fuel ratio feedback. The deviation of the correction coefficient from the reference value is learned for each engine operating region and stored as a learning value, and the basic fuel injection amount set based on the engine operating state is calculated based on at least the air-fuel ratio feedback correction coefficient and each engine operating region. An interpolator is used to calculate the learned values according to the current engine operating conditions. In the air-fuel ratio learning control device for an engine, which sets the final fuel injection amount, the air-fuel ratio shifts from rich to lean by the air-fuel ratio feedback correction using the air-fuel ratio feedback correction coefficient. After that, the reference reversal time corresponding to the normal time, which represents at least one of the time until the transition to the rich again and the time from the transition from the lean to the rich to the transition to the lean again, is defined as the engine speed and the basic fuel injection amount. Based on the O2 sensor output value, at least one of the switching of the air-fuel ratio from rich to lean, and the switching from lean to rich, based on the O2 sensor output value. A reversal determining means for determining whether or not there is an air-fuel ratio feedback supplement when there is no switching of the air-fuel ratio within the reference reversal time. The change state of the positive coefficient is determined, and when the air-fuel ratio feedback correction coefficient is increasing, the learning value of the corresponding area is increased by a predetermined amount. When the air-fuel ratio feedback correction coefficient is decreasing, the learning value of the corresponding area is reduced. Is reduced by a predetermined amount, and when the air-fuel ratio is switched within the reference inversion time, the state is shifted from the state where the air-fuel ratio is not switched to the state where the air-fuel ratio is switched, and the air-fuel ratio is switched first. In the first time the number of changes reaches the set number, the total amount of deviation of the average value of the air-fuel ratio feedback correction coefficient from the reference value during the period when the number of air-fuel ratio changes reaches the set number after the air-fuel ratio has changed Is subtracted from the learning value of the corresponding area, the learning value of the corresponding area is updated, and the air-fuel ratio feedback correction coefficient is corrected to the reference value. Thereafter, while the switching of the air-fuel ratio is continued, A learning value update unit for subtracting a predetermined ratio of a deviation of the average value of the fuel ratio feedback correction coefficient from the reference value from the learning value of the corresponding region to update the learning value of the corresponding region, and is set based on the engine operating state. Basic fuel injection amount, at least the air-fuel ratio feedback correction coefficient,
A fuel injection amount setting means for correcting the learning value learned for each engine operation region with a learning correction coefficient obtained by interpolation according to the current engine operation state, and setting a final fuel injection amount. An air-fuel ratio learning control device for an engine, comprising: 3. The engine idle engine according to claim 1, wherein the reference inversion time setting means sets the reference inversion time to a shorter value as the engine speed increases and the load increases. Fuel ratio learning control device.