JPH0821280A - Air-fuel ratio learning control device for engine - Google Patents

Abstract

PURPOSE:To perform correction of a deviation from a demand correction amount with high response even when a learning value is deviated from the demand correction amount during the stop of an engine. CONSTITUTION:A learning value Lalpha1 on a large region map contained in a memory 34 is updated by a updating means 41 in prior to a learning value Lalpha2 on a small region map contained in a memory 35 and when measurements of the number of up dating times of the learning value Lalpha1 on the large region map exceed a given value, a learning value Lalpha2 on the small region map is updated in such a state that the leaning value Lalpha1 on the large region map is fixed by an updating means 44. It is decided by a deciding means 45 whether a lapse time starting from the starting of an engine is below a given value and from the decision result, when a lapse time beginning from starting the learning value Lalpha1 on the large region map is varied by a updating means 46 without fail.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an engine air-fuel ratio learning control system.

[0002]

2. Description of the Related Art Basic pulse width Tp for determining a base air-fuel ratio
Since it is not possible to settle the air-fuel ratio to the stoichiometric air-fuel ratio with high accuracy under all operating conditions, a correction amount is required to eliminate the deviation from the stoichiometric air-fuel ratio, but this required correction amount varies greatly depending on the operating conditions. . For this reason, all the operating areas on the learning map are divided into a plurality of areas, and the learning value for giving the required correction amount is stored independently for each of the divided areas.

In this case, when the area on the map is enlarged, the learning value converges quickly, but on the other hand, the accuracy of the learning value is not so good due to the error between the required correction amount and the learning value in the same area. . On the other hand, when the area on the map is made smaller, the error with the required correction amount becomes smaller in the same area, but learning does not proceed in the area where there are few opportunities to update the learning value, Uneven learning values can be seen on the map.

Therefore, in Japanese Unexamined Patent Publication No. 4-321741, the operation areas are stored for each area (for example, 16 areas) on a map into which the operation areas are largely divided, and the learning values having a relatively large update ratio are divided into the above-mentioned areas. Each area is further subdivided into areas on the map (for example, if each area subdivided into 16 is subdivided into 16, the number of areas is 16 × 16 = 256), and the update rate is relatively small. A learning value and two types of learning values are prepared, and the sum of these two learning values (learning value Lα1 on the large area map and learning value Lα2 on the small area map) is set as the entire learning value Lα While the basic pulse width Tp is corrected together with the fuel ratio feedback correction coefficient α, the learning value Lα1 on the large area map is prioritized to update the learning value, and the prioritized learning value Lα1 When the new number is equal to or greater than a predetermined value,
While keeping the learning value Lα1 of the priority, the learning value Lα2 is updated on the small area map on the small area map overlapping the area on the map to which the learning value Lα1 belongs. There is. First, the learning value Lα1 is quickly converged by advancing the learning on the large area map, and then the learning value Lα2 is updated on the small area map. The deviation from the required correction amount in the same area on the map is eliminated.

[0005]

By the way, when refueling is performed when the engine is stopped and the fuel property changes, the held learning values Lα1, Lα2 do not match the actual fuel property, and after restarting the air-fuel ratio. Shifts to the rich side or the lean side. The learning values Lα1 and Lα2 do not match the actual engine state not only because of changes in the fuel properties associated with refueling, but when the engine is stopped, the engine parts that affect the air-fuel ratio, such as the air flow meter and fuel injector, are replaced. The same applies when the replacement causes the performance of the component to be different from that before the replacement, or when the environmental conditions such as the atmospheric pressure and the atmospheric temperature are different from those of the previous operation even without the replacement of the component.

The air-fuel ratio deviating from the stoichiometric air-fuel ratio in this way converges to the stoichiometric air-fuel ratio at a constant speed determined by the proportional constant and the integral constant when the feedback control of the air-fuel ratio is started after restarting. However, after that, if the learning condition is satisfied for each area on the map that divides the operating area and the learning value starts to be updated, the speed increases according to the update rate for each area on the map, and it converges. Will speed up.

However, according to the above apparatus, even if the engine is stopped after the learning value Lα2 is updated on the small area map and the learning value and the required correction amount deviate during the engine stop, the Only the update of the learning value Lα2 on the small area map is continued at the time of driving, and the update of the learning value Lα1 on the large area map is not restarted. Since deviations in the air-fuel ratio due to changes in fuel properties due to refueling when the engine is stopped and deviations in air-fuel ratio due to changes in environmental conditions before and after engine stop occur in all operating areas, learning on a large area map By updating the value Lα1, it is less likely that the learning value becomes uneven in the entire driving area and the learning value converges faster if a wide operating area is covered by one update. However, the learning value is updated on the small area map. In this case, since the update can cover only a narrow driving region once, the learning value tends to be uneven in the entire driving region, and the learning value converges slowly.

On the other hand, after shifting to the update of the learning value Lα2 on the small area map, the air-fuel ratio feedback correction coefficient α
And the reference value are compared, and when the difference between the two increases to a predetermined value or more, the update of the learning value Lα2 on the small area map is interrupted, and the update of the learning value Lα1 on the large area map is restarted. It deteriorates while the air cleaner is in use,
Due to the leak, the air-fuel ratio is greatly displaced to the lean side,
There is a measure to deal with the case where a large amount of deviation of the air-fuel ratio suddenly occurs (see Japanese Patent Laid-Open No. 4-145539).

In this system, unlike the deviation of the air-fuel ratio due to the change of the fuel property accompanying refueling and the deviation of the air-fuel ratio due to the change of environmental conditions, the deviation of the air-fuel ratio when it is considerably larger than this is assumed. Therefore, in this case, when there is a deviation of the air-fuel ratio due to changes in fuel properties due to refueling or a deviation of the air-fuel ratio due to changes in environmental conditions, the air-fuel ratio feedback correction coefficient α becomes greater than the reference value. Therefore, the learning value Lα1 on the large area map is not updated after the learning value Lα2 is updated on the small area map.

Therefore, according to the present invention, when the number of times the learning value is updated on the large area map becomes equal to or larger than a predetermined value, the learning value is updated on the small area map. By restarting the update of the learning value on the map, the object is to correct the deviation from the required correction amount with good response even if the learned value deviates from the required correction amount while the engine is stopped.

[0011]

The first invention, as shown in FIG. 12, is a means 31 for calculating a basic injection amount Tp according to operating conditions, and a sensor 3 for detecting the oxygen concentration in exhaust gas.
2, means 33 for calculating the air-fuel ratio feedback correction amount α so that the air-fuel ratio is controlled close to the stoichiometric air-fuel ratio based on the sensor output, and the learning value for each region on the map into which the operating region is largely divided. A memory 34 for storing Lα1, a memory 35 for storing a learning value Lα2 for each area on the map into which the operating area is divided into small areas, and values of these memories 34, 35 are retained so as not to disappear even after the engine is stopped. Means 36, means 37 for determining which area on the large area map the current operating condition belongs to, and area on the small area map 37, and an area to which the current operating condition belongs based on the determination result. Means 38 for reading the learned values stored in the large area map and the small area map, and the two learned values read out. Means 39 for calculating the fuel injection amount by correcting the basic injection amount Tp with α1, Lα2 and the air-fuel ratio feedback correction amount α, means 40 for supplying the injection amount of fuel to the intake pipe, and the large area map Means 41 for prioritizing the learned value Lα1 on the small area map based on the air-fuel ratio feedback correction amount α, and the learned value Lα on the large area map.
A unit 42 for measuring the number of updates of 1, a unit 43 for judging whether or not the measured value becomes a predetermined value or more, and a unit 42 on the large area map when the measured value becomes a predetermined value or more based on the judgment result. Means 44 for updating the learning value Lα2 on the small area map based on the air-fuel ratio feedback correction amount α while the learning value Lα1 is fixed, and whether the elapsed time from the start of the engine is less than or equal to a predetermined value. And a means 46 for updating the learning value Lα1 on the large area map based on the air-fuel ratio feedback correction amount α when the elapsed time from the start is less than or equal to a predetermined value. Provided.

As shown in FIG. 13, the second aspect of the invention is based on means 31 for calculating the basic injection amount Tp according to the operating conditions, a sensor 32 for detecting the oxygen concentration in the exhaust gas, and the output of this sensor. Means 33 for calculating the air-fuel ratio feedback correction amount α so that the air-fuel ratio is controlled close to the stoichiometric air-fuel ratio, memory 34 for storing the learning value Lα1 for each region on the map into which the operating region is largely divided, and operation A memory 35 for storing the learning value Lα2 for each area on the map obtained by dividing the area into small areas, and these memories 34, 35
Means 36 for holding the value of the value so as not to disappear even after the engine is stopped, and means for determining which region on the large region map the current operating condition belongs to and which region on the small region map. 37, means 38 for reading the learning values stored in the area to which the current driving condition belongs from the determination result, from the large area map and the small area map, respectively, and the two read learning values Lα1 and Lα2. The air-fuel ratio feedback correction amount α
Means 39 for correcting the basic injection amount Tp to calculate the fuel injection amount, means 40 for supplying the injection amount of fuel to the intake pipe, and a learning value Lα1 on the large area map for the small area map. Means 41 for updating based on the air-fuel ratio feedback correction amount α in preference to the learning value Lα2 above
And a means 42 for measuring the number of updates of the learning value Lα1 on the large area map, a means 43 for judging whether or not the measured value is equal to or more than a predetermined value, and a measured value based on the result of the judgment. And a means 44 for updating the learning value Lα2 on the small area map based on the air-fuel ratio feedback correction amount α while the learning value Lα1 on the large area map is fixed. Means 51 for determining whether the total number of cycles or the integrated value of the engine speed is equal to or less than a predetermined value, and the total number of cycles from the start (720 ° for a 4-cylinder engine based on the determination result
(4 cycles in a section) or when the integrated value of the engine speed is less than or equal to a predetermined value, the learning value Lα on the large area map
1 is provided based on the air-fuel ratio feedback correction amount α.

According to a third aspect of the present invention, in the first or second aspect of the invention, the predetermined value to be compared with the elapsed time from the start or the integrated value of the total number of cycles or the number of revolutions from the start is the engine temperature (at the time of start). The water temperature, the oil temperature at the time of starting, the wall temperature of the intake pipe at the time of starting, and the fuel temperature at the time of starting) are higher values as the engine temperature is lower.

According to a fourth aspect of the present invention, in any one of the first to third aspects of the present invention, after shifting to updating of the learning value Lα2 on the small area map, at least on the small area map. When the update value of the learning value Lα2 in the two areas is uniform and at least two areas on the small area map are not in the same area on the large area map, the learning value Lα1 on the large area map is large. To be updated for a predetermined period.

According to a fifth aspect of the invention, in any one of the first aspect to the third aspect, at least after the learning value Lα2 is updated on the small area map, at least on the small area map. The sum of each learning value of the two areas and each learning value of the area on the large area map including the areas is updated at a constant rate, and at least two areas on the small area map are large areas. When they are not in the same area on the map, the learning value Lα1 on the large area map is updated for a predetermined period.

[0016]

In the first aspect of the invention, regardless of the update status of the learning value Lα1 on the large area map during the previous operation, the updating of the learning value Lα1 on the large area map is restarted for a predetermined period at the time of starting. Therefore, after shifting to the learning value update on the small area map, the engine is stopped and refueled, and the refueling changes the fuel properties from before the engine was stopped, and the atmospheric pressure and atmospheric pressure before and after the engine was stopped. Even if environmental conditions such as temperature change, the deviation from the required correction amount due to these changes can be corrected with good response, and unevenness of the learning value between regions on the map can occur. Absent.

Since the number of updates of the learning value Lα1 on the large area map is substantially proportional to the total number of cycles from the start or the integrated value of the engine speed, in the second invention, the learning value Lα1 is actually calculated from the start. Since the number of updates can be estimated, the learning value Lα1 can be reliably updated on the large area map, and then the learning value can be returned to the conventional update of the learning value.

Since it takes time for the learned value to stabilize when the engine temperature is low, in the third invention, it is compared with the elapsed time from the start or the integrated value of the total number of cycles or the number of revolutions from the start. Since the predetermined value becomes larger as the engine temperature becomes lower, an appropriate predetermined value can be set even if the engine temperature is different, and the accuracy of the learned value is improved.

According to the fourth aspect of the invention, during the operation after the shift to the update of the learning value Lα2 on the small area map, the fuel property is changed without stopping the engine to change the fuel property and the fuel injector during operation. Even if a sudden change in the required correction amount occurs during operation, such as a decrease in the flow rate or a change in the spray state due to sudden clogging of the engine, the air-fuel ratio can be quickly returned to the stoichiometric air-fuel ratio. it can.

In the fifth aspect, even when the low speed low load area and the high speed high load area are selected as at least two areas on the small area map, the learning value Lα on the small area map is selected.
It is possible to accurately cope with a sudden change in the required correction amount during the operation after shifting to the update of 2.

[0021]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS In FIG. 1, air is sucked into an engine from an air cleaner 2 through an intake duct 3, a throttle chamber 4 and an intake manifold 5.

At each branch of the intake manifold 5,
A fuel injector 6, which is an electromagnetic on-off valve that opens when the solenoid is energized and closes when the energization is stopped, is provided, and intake is performed by the pressure regulator 7 only while a pulse signal from an ECU 11 described later is at a high level. Fuel adjusted so that the pressure difference with the pipe is constant is intermittently injected and supplied in synchronization with the rotation of the engine.

This injected fuel forms a mixture with air and is ignited and burned by spark ignition in the cylinder.
It is discharged from the exhaust manifold 8 through the three-way catalyst 9.

Reference numeral 12 is a hot-wire type air flow meter provided immediately downstream of the air cleaner 2 and has an intake air flow rate Q.
The signal corresponding to is output. Reference numeral 13 is a crank angle sensor, which is a Ref signal (in the case of four cylinders, the crank angle is 180 °).
Signal for each) and Pos signal (1 at crank angle)
Unit signal every ° or 2 °) is output. 14
Is a water temperature sensor for detecting the cooling water temperature Tw of the water jacket of the engine.

The signals from these sensors, together with the signals from the O ₂ sensor 15 and the ignition switch 16 which output signals according to the oxygen concentration in the exhaust gas, are ECUs (Electronic Control Units), which are mainly microcomputers. ) 11 is entered, E
The CU 11 controls fuel supply to the engine while performing feedback control of the air-fuel ratio and learning control of the air-fuel ratio.

The air-fuel ratio control in the ECU 11 is as follows.

The fuel injector 6 is driven in synchronization with the Ref signal. For example, in the sequential injection method, once every two engine revolutions, for each cylinder Ti = 2 × Te + Ts (1) where Te: effective pulse width Ts: injection pulse given by the formula of invalid pulse width according to battery voltage Injector 6 with width Ti
Is activated. Ts is a correction value for compensating the error with the actual injection amount due to the operation of the injector. In the case of the simultaneous injection method, once for each engine revolution,
All cylinders simultaneously Ti = Te + Ts ... Injector 6 with injection pulse width Ti given by the equation (2).
Is activated.

FIG. 2 is a flow chart for calculating the effective pulse width Te of the above equation (1), which has a constant period (for example, 10 ms).
ec).

In step 1, the basic pulse width Tp is calculated from the air flow rate Q detected by the air flow meter 12 and the engine speed N detected by the crank angle sensor 13 as follows: Tp = (Q / N) × K (3) where K : Calculate with a constant formula. The air-fuel ratio determined by this Tp is called the base air-fuel ratio.

In step 2, the effective pulse width Te is calculated by using the basic pulse width Tp: Te = Tp × Co × {(α + Lα-100) / 100} (4) where Co: various correction factors α: air-fuel ratio feedback Correction coefficient [%] Lα: Air-fuel ratio learning value [%] is calculated by the formula.

The various correction coefficients Co in the equation (4) are values for ensuring smooth operation under various conditions. For example, at the time of starting, warming up, high load, etc., the basic pulse width T is based on the signals from the water temperature sensor 15 and other sensors.
Correct p. At this time, the value of the air-fuel ratio feedback correction coefficient α, which will be described later, is clamped at 100% (steps 21 and 22 in FIG. 4).

FIG. 3 is a flow chart of calculation of the fuel injection pulse width Ti and execution of injection, which is executed in synchronization with the Ref signal.

Step 11 for sequential injection
Then, the fuel injection pulse width Ti of the above formula (1) is calculated, and this is transferred to the output register of the CPU (one of the components of the microcomputer) in step 12. With the ignition order in a four-cylinder engine set to # 1- # 3- # 4- # 2, when the Ref signal is input this time, for example, T is set to the first cylinder.
If the fuel corresponding to i is supplied, the third Ref.
When the ef signal is input, the No. 4 cylinder is supplied, and when the Ref signal is input three times after, the Ti fuel is supplied to the No. 2 cylinder.

The air-fuel ratio feedback correction coefficient α in the equation (4) is a value obtained in synchronization with the Ref signal by proportional-plus-integral control (a kind of feedback control) based on the output of the O ₂ sensor 15, and the value of α is 100%. When it exceeds, the air-fuel ratio is returned to the rich side by the equation (4), and when it is less than 100%, the air-fuel ratio is returned to the lean side.

FIG. 4 is a flow chart for calculating the air-fuel ratio feedback correction coefficient α, which is executed in synchronization with the Ref signal.

When the feedback condition (abbreviated as F / B condition in the figure) is satisfied in step 21, step 23
Read the O ₂ sensor output (abbreviated as O ₂ / S in the figure). Since the output of this O ₂ sensor becomes high level (about 1 V) on the rich side and low level (approximately 0 V) on the lean side of the theoretical air-fuel ratio, if it exceeds the slice level provided about 0.5 V, When the actual air-fuel ratio is on the rich side, and when it is smaller than the slice level, it is on the lean side.

In step 24, the O ₂ sensor output is compared with the slice level (approximately 0.5 V). If the O ₂ sensor output is larger than the slice level, the air-fuel ratio is on the rich side of the theoretical air-fuel ratio. Judge, Step 25
Proceed to.

In step 25, it is judged whether it is the rich side for the first time, and if it is the first time, step 26
Then, the value contained in the variable α (representing the air-fuel ratio feedback correction coefficient) α is transferred to the variable a. The value stored in the variable α at the timing when it is judged to be on the rich side for the first time is
It is the maximum value of the past air-fuel ratio feedback correction coefficient (α) in a half cycle starting from that timing, and this is sampled in the variable a.

At step 27, α ← α-P (5) However, the value of the variable α is updated by the equation of P: proportional constant, and at step 28, the flag FP is calculated.
Is set to "1". The fact that it is judged to be the first time in step 25 means that it has been reversed to the rich side for the first time this time, and therefore P, which is a proportional amount, is given by the equation (5). Further, FP = 1 indicates that the inversion is performed. This flag FP is required because the timing of updating the air-fuel ratio learning value, which will be described later, is set to the reverse time.

Also, when the output of the O ₂ sensor is smaller than the slice level in step 24 and the routine proceeds to step 32, it is judged whether it is the lean side for the first time. When it is determined that the variable α is present, the value of the variable α is transferred to the variable b in step 33. The value contained in the variable α at the timing when it is judged to be lean for the first time is the minimum value in the half cycle of α in the past starting from that timing, and this is sampled in the variable b.

In steps 34 and 35, step 27,
Similarly to 28, the value of the variable α is updated by the expression α ← α + P (6) and the flag FP is set to “1”.

At step 29, the count value cnt is set to 0.
When cnt ≠ 0, the count value cnt is decremented by 1 in step 30, and the routine of FIG. 4 is terminated. In addition, when cnt = 0, FIG.
Exit the routine.

The count value cnt is 16 as will be described later.
A value that is set to a predetermined value (for example, 4) when it is not in the same area as the previous time on the area map. Immediately after this setting, the value is decremented by 1 at each inversion by the operations in steps 29 and 30. It becomes 0. Since the count value cnt is not 0, it is determined that it has not stably stopped in one area on the 16-area map.
That is, = 0 is a learning condition on the 16-region map.

On the other hand, if it is not the first time that it is judged to be on the rich side in step 25, the routine proceeds to step 31, where α ← α-I × Ti (7) where I: integration constant Ti: fuel injection pulse The value of the variable α is updated by the expression of width and the routine of FIG. 4 is ended. Similarly, even when it is not the first time that it is judged to be on the lean side in step 32, the value of the variable α is updated by the expression α ← α + I × Ti (8) in step 36, and the routine of FIG. To finish.

In the equations (7) and (8), Ti may basically be omitted.

Thus, the above equations (5) and (7) are obtained.
The actual air-fuel ratio is almost 1 to 1 by repeating a series of updating of the variable α by the equations, and the equations (6) and (8).
It will change in a cycle of 2 Hz, and the average air-fuel ratio will be a window (predetermined air-fuel ratio range centered on the theoretical air-fuel ratio).
It is maintained inside.

The learning value map is made up of two types of maps, one of which is shown in the upper part of FIG. 5 and has a total operating area of 16 with the rotational speed N and the basic pulse width Tp as parameters.
A map in which the learning value Lα1 having a relatively large update ratio for each area is rewritably stored (hereinafter referred to as a 16-area map). As shown in the lower part of FIG. 5, each area in the 16-area map is further divided into 16 equal areas (thus, the total number of divisions is 16 × 16 = 256). A learning value Lα2 having a relatively small update rate for each area is a rewritable map (hereinafter referred to as 256 area map).

Further, the map value is held by battery backup or the like so that the value does not disappear even after the engine is stopped.

In FIG. 5, the areas on the 16 area map are distinguished as [B, A] and the areas on the 256 area map as [K, I]. In addition, the area [B,
The learning value contained in A] is represented by Lα1 [B, A], and the learning value contained in the region [K, I] is represented by Lα2 [K, I]. The initial values of Lα1 and Lα2 are 100 [%] in all regions.

Here, the area [B, A] on the 16-area map is the rotation speed band N [B] and the basic pulse width band Tp.
Determined from [A]. B and A are integers from 0 to 3, and the number of revolutions is N

[0], N [1], N [2], N
[3] indicates the following rotational speed range, and

[0], Tp
[1], Tp [2], and Tp [3] represent the following respective basic pulse width regions.

N

[0]: N0 ≦ N <N1 (9) N [1]: N1 ≦ N <N2 (10) N [2]: N2 ≦ N <N3 (11) N [3]: N3 ≦ N < N4 (12) Tp

[0]: Tp0 ≦ Tp <Tp1 (13) Tp [1]: Tp1 ≦ Tp <Tp2 (14) Tp [2]: Tp2 ≦ Tp <Tp3 (16) Tp [3]: Tp3 ≦ Tp < Tp4 (17) However, N0, N1, N2, N3, N4 and Tp0, Tp
1, Tp2, Tp3, and Tp4 are boundary values (constant values), N
0 <N1 <N2 <N3 <N4, Tp0 <Tp1 <Tp2
<Tp3 <Tp4.

For example, the current rotational speed N and the basic pulse width T
If N1 ≦ N <N2 and Tp2 ≦ Tp <Tp3 for p, the current rotation speed N is the rotation speed width of N [1] and the current basic pulse width Tp is the basic pulse width band of Tp [2]. It will be included and is determined from N [1] and Tp [2] 16
The area on the area map is [1, 2].

Similarly, I and K of the area [K, I] in the 256 area map are integers from 0 to 15,
Rotation speed band N [K] (N

[0] to N [15]) and the basic pulse width band Tp [I] (Tp

The areas [K, I] on the 256 area map are individually determined from [0] to Tp [15]).

FIG. 6 is a flow chart for calculating the air-fuel ratio learning value Lα (not L × α, which is one symbol for Lα as a whole), which is executed at a constant cycle (for example, 10 ms).

In step 41, the region [B, A] on the 16-region map is calculated from the engine speed N and the basic pulse width (engine load equivalent amount) Tp at that time, and in step 42, the engine speed N at that time. Then, the area [K, I] on the 256 area map is determined from the basic pulse width Tp.

In step 43, the learning value Lα1 [B, A] [%] contained in the area [B, A] and the learning value Lα2 [K, I] [%] contained in the area [K, I] are set. reading,
The air-fuel ratio learning value Lα is calculated by the formula Lα ← Lα1 [B, A] + Lα2 [K, I] -100 (18).

In the equation (18), 100 [%] is a value required for unit matching.

Next, regarding the order in which the two types of learning values Lα1 and Lα2 are updated, the learning value Lα1 on the 16-region map is given priority.

This will be briefly described with reference to FIG. 7. In the upper part of FIG. 7, the horizontal axis represents the basic pulse width band Tp [A], the vertical axis represents the learning value Lα1, and in the lower part, the horizontal axis represents the basic pulse width band Tp.
[I] is the characteristic when the learning value Lα2 is taken on the vertical axis, and when the learning is not performed at all like a new car (that is, both Lα1 and Lα2 are 100% of the initial values), the required correction amount is It is assumed that the value is a step-like value that is out of 100% as shown by the thick line in the upper left. That is, the required correction amount is 1
The fact that it is greater than 00% means that the air-fuel ratio is shifted to the lean side.

In this case, when the learned value Lα2 is fixed (100% when the vehicle is new and fixed at 100%) (see the lower left of FIG. 7), the learned value Lα1 is updated, and the air-fuel ratio becomes richer. The learning value Lα1 is changed to a value larger than 100% in an attempt to restore the value. Here, for simplicity, Tp [1] on the 256 area map is updated a predetermined number of times,
It is assumed that the requested correction amount at Tp [2] is matched (see the upper right of FIG. 7).

However, in this state, the deviation from the required correction amount due to the difference in area on the 256 area map (that is, Tp on the 256 area map)

[0], Tp [3]
Deviation from the required correction amount in) cannot be compensated. At Tp [3] on the 256 area map, the correction by the learning value Lα1 is insufficient and the air-fuel ratio leans to the lean side.

At [0], the learning value Lα1 becomes excessive and the air-fuel ratio deviates to the rich side.

Therefore, this time, with the learning value Lα1 fixed, the learning value shifts to learning on the 256 area map and the learning value Lα2 is converged as shown in the lower right.
At Tp [3] on the area map, the learning value Lα2 changes to a larger value in order to return the air-fuel ratio to the rich side.
Tp on 6 area map

At [0], the learning value Lα2 changes to the smaller side in order to return the air-fuel ratio to the lean side.
The sum Lα1 + Lα2 of the two learning values matches the required correction amount.

Thus, the learning value L on the 16-region map
If the number of updates exceeds a predetermined value while increasing the speed of convergence of the learning value by first updating α1, 16
In order to eliminate the deviation from the required correction amount in the same area on the area map, this time the learning value Lα on the 256 area map is eliminated.
By shifting to the update of 2, the accuracy of the learning value can be improved.

FIG. 8 is a flow chart for updating the learning values Lα1 and Lα2, which is executed at a constant cycle (for example, 10 ms).

In step 51, the value of the flag FP is checked, and when it is "0", the routine of FIG. 8 is terminated as it is, and when the value of the flag FP is "1", when the output of the O ₂ sensor is reversed. (Or at the time of inversion of the air-fuel ratio), the value of FP is returned to "0" in step 52, and the learning value is updated after step 53.

In steps 53 and 54, it is determined which region on the 16-region map the operating condition, which is determined by the rotational speed N and the basic pulse width Tp at that time, and which region on the 256-region map belong.

In step 55, it is checked whether the determined area [B, A] on the 16-area map is the same as the previous time. If they are not the same, the count value c is determined in step 56.
A predetermined value (for example, 4) is set in nt.

At step 57, the value of the flag F [B, A] is checked. The flag F [B, A] = 1 indicates that the learning value Lα1 in the region [B, A] on the 16-region map has been updated a predetermined number of times, and F [B, A] = 0 indicates the number of updates. Represents less than a predetermined number of times. Therefore, F [B,
When the value of [A] is “0”, it is determined that the number of updates of the learning value Lα1 in the area [B, A] has not reached the predetermined value, and the process proceeds to step 58. In step 58, the count value cnt is checked. If it is not zero, the learning value Lα1
When it is determined that the update condition of [B, A] is not satisfied, the routine of FIG. 8 ends. The count value cnt is set to 4 when the area [B, A] on the 16-area map is not the same as the previous time, and then decreases by 1 every time the output of the O ₂ sensor is inverted, and becomes 0. Even when the driving condition deviates from the same region on the 16-region map, the learning value Lα1
When [B, A] is updated, the learning value Lα1 [B,
Since the accuracy of [A] deteriorates, if it deviates from the same area on the 16-area map, it must be after a predetermined time elapses.
The learning value Lα1 [B, A] is not updated.

At step 59, the count value CNT1
[B, A] is compared with a predetermined value (integer of 1 or more) C1.
If you proceed to step 59 for the first time, CNT1
[B, A] <C1, and in step 60, the learning value Lα1 contained in the region [B, A] is Lα1 [B, A] ← Lα1 [B, A] + X1 · {(a + b) / 2− 100} (19) However, X1 is updated by the formula of update ratio (0 <X1 <1), and the updated value is stored again in the same area [B, A]. In the equation (19), (a + b) / 2 [%] is the average value of the air-fuel ratio feedback correction coefficient α in the past half cycle, and the difference between this and 100 [%] represents the deviation from the theoretical air-fuel ratio, The learning value Lα1 [B, A] is rewritten by adding some percentage of this deviation.

At step 61, the count value CNT1
Increment [B, A] by 1. Count value C
NT1 [B, A] is the learning value Lα in the area [B, A]
The number of updates of 1 is counted, and the initial value is 0.

With this count, CNT1
When [B, A] ≧ C1, it is determined that the learning in the area [B, A] on the 16-area map has progressed sufficiently, and the process proceeds from step 59 to step 62 to set the flag F [B, A].
A] is set to "1". The initial value of F [B, A] is "0".

When the flag F [B, A] is set to "1", the process proceeds from step 57 to step 63.

In step 63, the half-cycle average value (a + b) / 2 of the air-fuel ratio feedback correction coefficient α is compared with 100. If this half-cycle average value does not substantially match 100, in step 64 the region [K , I] learning value Lα2
Lα2 [K, I] ← Lα2 [K, I] + X2 · {(a + b) / 2-100} (20) where X2: update ratio (0 <X2 <1) The later value is restored in the same area [K, I].

As can be seen by comparing equation (20) with equation (19), the method of updating the learning value Lα2 on the 256 area map is the same as the updating of the learning value Lα1 on the 16 area map. The difference is that the value of the update rate X2 is made relatively smaller than the update rate X1.
On the 16-region map, X1 is set to a relatively large value so that the learning value Lα1 converges quickly. On the other hand, on the 256-region map, X2 needs to be set to a relatively small value to improve the accuracy of the entire learning value Lα. Because there is.

Needless to say, this area [K, I] on the 256 area map is included in the area [B, A] in which the flag F is set to "1".

When the half-cycle average value of the air-fuel ratio feedback correction coefficient α is substantially equal to 100 in step 63, it is determined that the learning value Lα2 in the region [K, I] has converged, and step 65 Flag FF [K, I] to "1"
Set to. FF [K, I] = 0 indicates that the learning value Lα2 in the area [K, I] has converged when FF [K, I] = 1.
Indicates that the learning value Lα2 [K, I] has not yet converged. The initial value of the flag FF [K, I] is also "0".
Is.

In this way, learning is first performed for each region on the 16-region map, and for the region where the learning value Lα1 on the 16-region map has converged, the region is further divided by 1.
By updating the learning value Lα2 for each of the six regions, it is possible to improve the accuracy of the learning value Lα while quickly converging the average air-fuel ratio to the theoretical air-fuel ratio.

The learning value Lα1 is updated on the 16-region map and the learning value Lα2 on the 256-region map is updated until the number of updates of all learning values Lα1 on the 16-region map exceeds a predetermined value. Updates will be mixed.

The control up to this point is the same as the conventional one.

Now, the learning value Lα on the 256 area map
When the engine is stopped after shifting to the update of 2 and restarted after the fuel property changes due to refueling, or after environmental conditions such as atmospheric temperature and atmospheric pressure have changed significantly while the engine was stopped However, if the learning value Lα2 is continuously updated on the 256 area map after that, the held learning value does not match the actual fuel property or the required correction amount in the environmental conditions, and the air-fuel ratio Is shifted to the rich side and the lean side again.

To deal with this, the ECU 11 always advances learning on the 16-region map for a predetermined period at the time of starting.

FIG. 9 is a flow chart for updating the learning values Lα1 and Lα2, which is executed at a constant cycle (for example, 10 ms).

The routine of FIG. 8 corresponding to the conventional example is constructed as a part of the routine of FIG. However, the step (step 53 in FIG. 8) that is duplicated due to the relationship with FIG. 9 is unnecessary.

In FIG. 9, in step 71, the starter switch is checked, and if it is ON, step 72
Sets all flags F (F [0,0] to F [3,3]) to "0". Regardless of the state of progress of the learning value Lα1 during the previous operation, first, the learning value L on the 16 area map
This is to start updating α1.

When the starter switch is in the OFF state, the ignition switch is checked in step 73. If the ignition switch is in the OFF state, it is determined that the engine is stopped and the routine of FIG. 9 is terminated.

If the starter switch is in the OFF state and the ignition switch is in the ON state, it is determined that the operation is continued, and in step 74, the engine speed N and the basic pulse width Tp are in any region [B, In step 74, the count value CNT2 [B, A] provided corresponding to the area [B, A] is compared with the predetermined value C2. The initial value of the count value CNT2 [B, A] is 0.

Immediately after starting, CNT2 [B, A] <C
2 and proceeds to step 76 and subsequent steps.

Steps 76, 77, 78, 79, 80,
81 is steps 51, 52, 55, 56, 5 of FIG.
8, 60, 61, <1> FP = 1 (step 76), <2> area [B, A] is the same as the previous time (step 78), <3> cnt = If all of 0 (step 80) are satisfied,
Proceeding to step 81, the learning value L on the 16 area map
Update α1 and count value CNT2 in step 82
Increment [B, A] by 1.

The count value CNT2 [B, A] represents the number of updates of the learning value Lα1 from the start in the area [B, A] on the 16 area map, and the learning in the area [B, A] progresses. When CNT2 [B, A] ≧ C2, step 7
From 5 to step 83, the routine of FIG. 8 is executed as in the conventional case.

In this way, the learning value on the 16-region map is updated until the number of updates from the start of the learning value on any one region on the 16-region map exceeds a predetermined value. It will restart. Therefore, theoretically the earliest step proceeds to step 83 when the learning value in any one of the 16 region maps is continuously updated a predetermined number of times. Further, the process proceeds to the latest step 83 when the learning value is updated in any one area after the learning values in all areas are updated (C2-1) times on the 16 area map. .

Now, the operation of this example will be described.

The two learning values Lα1 and Lα2 are retained as before so that the values do not disappear after the engine is stopped. Therefore, before the engine is stopped using heavy gasoline, In 10 for example 16
The number of times the learning value has been updated in the area [0, 0] on the area map is equal to or greater than the predetermined value C1, and the area [0, 0] on the 16 area map is further subdivided into all areas on the 256 area map. Learning value Lα on 256 area map
2 is being updated, the learning value Lα1 [0,
0] and all learning values Lα2 in the area on the 256 area map that overlaps with the area [0, 0] on the 16 area map are held.

When the engine is restarted after being stopped, the area [0,0] on the 16-area map is used unless the fuel properties and environmental conditions are the same as before the engine was stopped.
When the engine is operated at, the air-fuel ratio is accurately controlled to the stoichiometric air-fuel ratio even after restarting.

However, when light gasoline with different fuel properties is supplied when the engine is stopped, after the restart, the learning value Lα is the required correction amount for the light gasoline in all the regions [0, 0] on the 16 region map. And the air-fuel ratio shifts to the rich side.

In this case, in this example, immediately after the start, regardless of the update status of the learning value Lα1 on the 16-region map during the previous operation, the learning value Lα1 on the 16-region map is always required.
Is restarted, the air-fuel ratio feedback control is started by the satisfaction of the air-fuel ratio feedback condition after the engine is restarted, and then the learning condition is satisfied at point C in FIG. 10, the learning value Lα1 [0, 0] becomes smaller (the direction in which the air-fuel ratio is returned to the lean side), the update ratio X1
Will be updated with. After that, when the operating condition moves to the point D or the point E and the learning condition is satisfied, the update of the learning value Lα1 [0,0] is continued at the same update rate X1, and the learning value Lα1 [0, 0] converges to the required correction amount when using light gasoline with good response.

Under the same conditions, C point and D in the conventional example
When the learning condition is satisfied at the points E and E, the learning value Lα2 is continuously updated on the 256 area map. Therefore, the learning values Lα2 [1,1], Lα2 [2,1], Lα2 [1,
0] is only updated at the update rate X2, the areas [1,1], [2,1], [1,0 on the 256 area map among the areas [0,0] on the 16 area map are updated. ], The learning value is left as it is deviating from the required correction amount when using light gasoline.

As described above, according to the conventional example, it is already 25
After shifting to the update of the learning value Lα2 on the 6-region map, the update of the learning value Lα1 on the region on the 16-region map overlapping the region containing the learning value Lα2 is not restarted. However, in this example, regardless of the update status of the learning value Lα1 on the 16-area map during the previous operation, the learning value Lα1 on the 16-area map is newly set.
The engine is restarted, so that the engine is stopped and refueled after shifting to the learning value update on the 256 area map, and the fuel property changes from before the engine was stopped due to the refueling, Even if the environmental conditions change before and after the stop, the deviation from the required correction amount due to these changes can be corrected with good response, and unevenness of the learning value between regions on the map occurs. Nothing.

FIG. 11 shows the second embodiment.

During the operation after the learning value Lα2 is updated on the 256 area map, the fuel properties are changed by refueling without stopping the engine, or the fuel injector is suddenly clogged during operation. Even if the flow rate decreases or the spray state changes due to
α and the required correction amount do not match, and the air-fuel ratio shifts to the rich side or the lean side. Since it is not possible to respond to the sudden change in the required correction amount by updating the learning value Lα2 on the 256 area map, in this example, it is possible to update the learning value Lα2 on the 256 area map. Even while driving
When the required correction amount changes, the update of the learning value Lα2 on the 256 area map is once interrupted, and the update of the learning value Lα1 on the 16 area map is restarted.

The necessity of refueling without stopping the engine can be considered in the following cases. Overseas, especially in the US and other self-service gas stations, fuel may be supplied while the engine is still running, and gasoline may become thin in remote areas, and the engine may be stopped when the fuel in the spare plastic tank is added. There is no doubt about it. In Japan as well as overseas, when an old car has a bad engine performance, it may be refueled without turning off the engine.

In FIG. 11, in steps 91 and 92, it is checked whether the following conditions are satisfied. If both conditions are satisfied, the process proceeds to step 93 and thereafter to update the learning value Lα2 on the 16 area map. Resume.

<1> The update amount of the learning value Lα2 in at least two areas on the 256 area map is uniform.
For example, if the learning values in the two regions on the 256 region map are Lα2a and Lα2b, then ΔLα2a (Lα
2a) and ΔLα2b (which is the update fee of Lα2b) are ΔLα2a = new Lα2a-old Lα2a (21) where new Lα2a: current Lα2a old Lα2a: previous Lα2a ΔLα2b = new Lα2b-old Lα2b (22) However, new Lα2b: current Lα2b old Lα2b: previous Lα2b, and ΔLα2a≈ΔLα2b is satisfied (step 91).

<2> The above-mentioned at least two areas on the 256 area map are not in the same area on the 16 area map. Speaking of the two learning values Lα2a and Lα2b, the area on the 16-region map to which one learning value Lα2a belongs is ARALα2a, and the region on the 16-region map to which another learning value Lα2b belongs is ARALα2b. It is to satisfy ARALα2b (step 92). ARALα2a and ARALα2b
When and are in the same area on the 16-area map,
This is because it cannot be determined whether the uniform updating of the two learning values Lα2a and Lα2b is normal or is accompanied by a sudden change in the required correction amount.

In step 93, it is judged whether or not the process has proceeded to step 93 for the first time. If it is the first time, all flags F are set to "0" in step 94. This is because the update of the learning value Lα1 on the 16-region map is restarted regardless of the update status of the learning value Lα1 on the 16-region map.

Next time, the process proceeds from step 93 to 95, but from step 95 to 103, steps 75 to 8 in FIG.
The same as up to 3.

In this example, after the shift to the update of the learning value Lα2 on the 256 area map, fuel is supplied without stopping the engine to change the fuel property, and the fuel injector during operation is suddenly clogged. Even if a sudden change in the required correction amount occurs during operation, such as a decrease in the flow rate or a change in the spray state due to, the air-fuel ratio can be quickly returned to the stoichiometric air-fuel ratio.

The above condition <1> (step 9 in FIG. 11).
1 condition), the sum of each learning value of at least two areas on the 256 area map and each learning value of the area on the 16 area map including each area is updated at a constant rate. It may be that.

For example, the above learning values Lα2a and Lα2
Speaking of b, the learning value in the area on the same 16 area map as the area containing one learning value Lα2a is Lα1A,
Same as the area containing another learning value Lα2b 16
If the learning value in the area on the area map is Lα1B,
Sum (Lα2a + Lα1A) and sum (Lα2b + Lα1B)
Is to be updated at the same rate. At this time, the sum (L
Each update rate of (α2a + Lα1A) and the sum (Lα2b + Lα1B) is the update rate of (Lα2a + Lα1A) = ΔLα2a / old (Lα2a + Lα1A) = ΔLα2a / (old Lα2a-old Lα1A) (23) (Lα2b + Lα1B) ... (23) (Lα2b + Lα1B) Lα2b + Lα1B) = ΔLα2b / (old Lα2b + old Lα1B) (24)

Here, the learning value Lα on the 16-region map
1A is before and after updating the learning value Lα2a on the 256 area map, and the learning value Lα1B on the 16 area map is 2
Since the learning value Lα2b does not change before and after the learning value Lα2b is updated on the 56 area map, if the old Lα1A and the old Lα1B are simply replaced by Lα1A and Lα1B, the formulas (23) and (24) are expressed as (Lα2a + Lα1A). Update rate = ΔLα2a / (old Lα2a−Lα1A) (25) Update rate of (Lα2b + Lα1B) = ΔLα2b / (old Lα2b + Lα1B) (26) When the update rate of (Lα2a + Lα1A) in equation (25) and the update rate of (Lα2b + Lα1B) in equation (26) substantially match, (Lα2a + Lα1A) and (Lα2a + Lα1A)
It is determined that (α2b + Lα1B) has been updated at a constant rate.

Since both learning values Lα1 and Lα2 tend to increase as the speed and load increase, the same 2
Even if the learning value on the 56 area map is 25
When one area on the 6-area map is selected as a low-speed low-load area and the other area is selected as a high-speed high-load area, the update allowances of the learning values in the two selected areas cannot be said to be uniform.
Although it becomes impossible to proceed to step 92 in FIG. 11, the learning values of at least two areas on the 256 area map and the learning values of the area on the 16 area map including each area are When the sum is updated at a constant rate, and the at least two areas on the 256 area map are not in the same area on the 16 area map,
By shifting to the update of the learning value on the 16 area map, even when the low speed low load area and the high speed high load area are selected as at least two areas on the 256 area map, the learning on the 256 area map is performed. It is possible to accurately cope with a sudden change in the required correction amount during operation after shifting to the value update.

In step 75 of FIG. 9, step 76 is executed when the elapsed time from the start is less than the predetermined value t1, and step 8 is executed when the elapsed time from the start becomes the predetermined value t1 or more.
It doesn't matter if you go to 3. Further, when the number of cycles from the start or the integrated value of the engine speed from the start is less than the predetermined value S1, the process proceeds to step 76, and the integrated value of the number of cycles from the start or the engine speed from the start becomes the predetermined value S1 or more. Then, it is possible to proceed to step 83.

Further, the predetermined values t1 and S1 in this case can be set to different values depending on the parameter representing the engine state. For example, the lower the engine temperature (for example, cooling water temperature, oil temperature, wall temperature, fuel temperature, etc.), the more unstable the engine state becomes, and it takes time for the learning value to stabilize. By setting the temperature to be lower as the temperature is lower, an appropriate predetermined value can be set even if the engine state is different, and the accuracy of the learned value can be improved.

[0113]

According to the first aspect of the present invention, the means for calculating the basic injection amount according to the operating condition, the sensor for detecting the oxygen concentration in the exhaust gas, and the air-fuel ratio close to the stoichiometric air-fuel ratio based on the sensor output. Control means to calculate the air-fuel ratio feedback correction amount, a memory to store learning values for each area on the map into which the operating area is largely divided, and learning for each area on the map into which the operating area is divided into smaller areas. Memories for storing values, means for retaining the values of these memories so as not to be lost even after the engine is stopped, and current operation conditions in any of the areas on the large area map and on the small area map. And a learning value stored in the area to which the current driving condition belongs from the determination result. And means for calculating the fuel injection amount by correcting the basic injection amount with the read two learning values and the air-fuel ratio feedback correction amount, and supplying fuel of this injection amount to the intake pipe. Means for updating the learning value on the large area map on the basis of the air-fuel ratio feedback correction amount prior to the learning value on the small area map, and the learning value on the large area map. Means for measuring the number of updates, a means for determining whether or not the measured value is equal to or greater than a predetermined value, and a learning value on the large area map when the measured value is equal to or greater than the predetermined value based on the determination result. Means for updating the learning value on the small area map based on the air-fuel ratio feedback correction amount in a fixed state, and determining whether the elapsed time from engine start is less than or equal to a predetermined value And a means for updating the learning value on the large area map based on the air-fuel ratio feedback correction amount whenever the elapsed time from the start is less than or equal to a predetermined value based on the result of this determination. After shifting to updating the learning value above, stop the engine and refuel, the fuel property changes due to the refueling, the atmospheric pressure before and after the engine stops,
Even if environmental conditions such as atmospheric temperature change, it is possible to correct the deviation from the required correction amount due to these changes with good response, and to cause uneven learning values between regions on the map. Nor.

The second aspect of the invention is to control the air-fuel ratio close to the stoichiometric air-fuel ratio based on the means for calculating the basic injection amount according to the operating conditions, the sensor for detecting the oxygen concentration in the exhaust gas, and the sensor output. As described above, a means for calculating the air-fuel ratio feedback correction amount, a memory for storing a learning value for each area on the map into which the operating area is largely divided, and a learning value for each area on the map into which the operating area is divided into smaller areas. Memories to be stored, means for retaining the values of these memories so as not to be lost even after the engine is stopped, and current operation conditions in any area on the large area map and any area on the small area map. And a learning value stored in a region to which the current driving condition belongs from the determination result. And a means for calculating the fuel injection amount by correcting the basic injection amount with the read two learning values and the air-fuel ratio feedback correction amount, and supplying fuel of this injection amount to the intake pipe. Means to do
Means for updating the learning value on the large area map based on the air-fuel ratio feedback correction amount in preference to the learning value on the small area map, and the number of times of updating the learning value on the large area map, A means for measuring, a means for determining whether or not the measured value becomes a predetermined value or more, and a state in which the learning value on the large area map is fixed when the measured value becomes a predetermined value or more based on the determination result. Means for updating the learning value on the small area map based on the air-fuel ratio feedback correction amount, and determining whether the total number of cycles from the start of the engine or the integrated value of the engine speed is less than or equal to a predetermined value. And the air-fuel ratio feedback correction of the learning value on the large region map when the total number of cycles from the start or the integrated value of the engine speed is less than a predetermined value Since the means for updating based on the above is provided, it is possible to count the number of times the learning value on the large area map is actually updated from the start, and it is possible to reliably update the learning value on the large area map. Later, the conventional learning value update can be reverted to.

According to a third aspect of the present invention, in the first or second aspect of the invention, the predetermined value to be compared with an elapsed time from start or an integrated value of the total number of cycles or the number of revolutions from start depends on the engine temperature. Since the lower the engine temperature is, the larger the value becomes, the appropriate predetermined value can be set even if the engine temperature is different, and the accuracy of the learned value is improved.

A fourth invention is the method according to any one of the first invention to the third invention, wherein after the learning value is updated on the small area map, at least 2 pieces on the small area map are updated. When there is a uniform update fee for the learning value in one area and the at least two areas on the small area map are not in the same area on the large area map, the learning value is updated on the large area map. Because the engine is refueled for a predetermined period of time, the fuel is refueled without stopping the engine and the fuel properties change during the operation after the learning value is updated on the small area map. Even when a sudden change in the required correction amount occurs during operation, such as when the flow rate is reduced due to clogging or the spray state changes,
The air-fuel ratio can be quickly returned to the stoichiometric air-fuel ratio.

In a fifth aspect of the present invention, in any one of the first to third aspects, at least 2 on the small area map is set after the learning value is updated on the small area map. Sum of each learning value of one area and each learning value of the area on the large area map including each area is updated at a constant rate, and at least two areas on the small area map are the large area map. If you don't see them in the same area,
Since the learning value is updated for a predetermined period on the large area map, even when the low speed low load area and the high speed high load area are selected as at least two areas on the small area map,
It is possible to accurately cope with a sudden change in the required correction amount during operation after shifting to the update of the learning value on the small area map.

[Brief description of drawings]

FIG. 1 is a system diagram of an embodiment.

FIG. 2 is a flowchart for explaining calculation of an effective pulse width Te.

FIG. 3 is a flowchart for explaining calculation of a fuel injection pulse width Ti and fuel injection.

FIG. 4 is a flowchart for explaining calculation of an air-fuel ratio feedback correction coefficient α.

FIG. 5 is a characteristic diagram showing a region in which two types of learning values Lα1 and Lα2 enter.

FIG. 6 is a flowchart for explaining reading of a learning value.

FIG. 7 is an explanatory diagram showing how two types of learning values Lα1 and Lα2 are updated.

FIG. 8 is a flowchart for explaining updating of learning values Lα1 and Lα2.

FIG. 9 is a flow chart for explaining a learning value update from the start.

FIG. 10 is a region diagram for explaining the operation of the first embodiment.

FIG. 11 is a flowchart for explaining updating of learning values Lα1 and Lα2 according to the second embodiment.

FIG. 12 is a diagram corresponding to claims of the first invention.

FIG. 13 is a diagram corresponding to a claim of the second invention.

[Explanation of symbols]

6 Fuel injector (fuel supply means) 11 ECU 12 Air flow meter 13 Crank angle sensor 14 Water temperature sensor 15 O ₂ sensor (oxygen concentration sensor) 31 Basic injection amount calculation means 32 Oxygen concentration sensor 33 Air-fuel ratio feedback correction amount calculation means 34 Large area On-map learning value memory 35 Small area on-map learning value memory 36 Memory value holding means 37 Area determination means 38 Learning value reading means 39 Fuel injection amount calculating means 40 Fuel supply means 41 Large area on-map learning value priority updating means 42 Number of updates Measuring means 43 Judging means 44 Small area map learning value updating means 45 Judging means 46 Large area map learning value updating means 51 Judging means 52 Large area map learning value updating means

Claims (5)

[Claims] 1. A means for calculating a basic injection amount according to operating conditions, a sensor for detecting an oxygen concentration in exhaust gas, and an air-fuel ratio controlled to be close to a stoichiometric air-fuel ratio on the basis of the sensor output. A means for calculating the air-fuel ratio feedback correction amount, a memory for storing a learning value for each region on the map into which the operating region is largely divided, and a memory for storing a learning value for each region on the map in which the operating region is divided into a small region. , A means for holding the values of these memories so as not to disappear even after the engine is stopped, and to which area on the large area map and which area on the small area map the current operating condition belongs. Based on the determination result and the determination result, the learning value stored in the region to which the current driving condition belongs is selected from the large area map and the small area map, respectively. Read out means, means for calculating the fuel injection quantity by correcting the basic injection quantity with the read-out two learning values and the air-fuel ratio feedback correction quantity, and fuel of this injection quantity is supplied to the intake pipe. Means, means for updating the learning value on the large area map based on the air-fuel ratio feedback correction amount in preference to the learning value on the small area map, and the learning value on the large area map A means for measuring the number of updates, a means for judging whether or not the measured value becomes a predetermined value or more, and a learning value on the large area map fixed when the measured value becomes a predetermined value or more as a result of this judgment. A means for updating the learning value on the small area map based on the air-fuel ratio feedback correction amount in a state, and means for determining whether the elapsed time from the start of the engine is a predetermined value or less, According to this determination result, when the elapsed time from the start is less than or equal to a predetermined value, a means for updating the learned value on the large area map based on the air-fuel ratio feedback correction amount is provided. Fuel ratio learning control device. 2. A means for calculating a basic injection amount according to an operating condition, a sensor for detecting an oxygen concentration in exhaust gas, and an air-fuel ratio controlled to be close to a stoichiometric air-fuel ratio based on the sensor output. A means for calculating the air-fuel ratio feedback correction amount, a memory for storing a learning value for each region on the map into which the operating region is largely divided, and a memory for storing a learning value for each region on the map in which the operating region is divided into a small region. , A means for holding the values of these memories so as not to disappear even after the engine is stopped, and to which area on the large area map and which area on the small area map the current operating condition belongs. Based on the determination result and the determination result, the learning value stored in the region to which the current driving condition belongs is selected from the large area map and the small area map, respectively. Read out means, means for calculating the fuel injection quantity by correcting the basic injection quantity with the read-out two learning values and the air-fuel ratio feedback correction quantity, and fuel of this injection quantity is supplied to the intake pipe. Means, means for updating the learning value on the large area map based on the air-fuel ratio feedback correction amount in preference to the learning value on the small area map, and the learning value on the large area map A means for measuring the number of updates, a means for judging whether or not the measured value becomes a predetermined value or more, and a learning value on the large area map fixed when the measured value becomes a predetermined value or more as a result of this judgment. Means for updating the learned value on the small area map based on the air-fuel ratio feedback correction amount in a state of being kept, and the total number of cycles from the start of the engine or the integrated value of the engine speed is equal to or less than a predetermined value Means for determining whether or not the total cycle number from start or the integrated value of engine speed is less than or equal to a predetermined value based on the determination result, the learning value on the large area map is always used as the air-fuel ratio feedback correction amount. An air-fuel ratio learning control device for an engine, characterized in that it is provided with means for updating based on the above. 3. The predetermined value to be compared with the elapsed time from the start or the integrated value of the total number of cycles or the number of revolutions from the start is a value that increases as the engine temperature decreases, depending on the engine temperature. The air-fuel ratio learning control device for the engine according to claim 1. 4. After shifting to the update of the learning value on the small area map, the update amount of the learning value on at least two areas on the small area map is uniform, and the update value on the small area map is the same. 4. When at least two areas are not in the same area on the large area map, the learning value on the large area map is updated for a predetermined period of time.
An air-fuel ratio learning control device for an engine according to any one of items 1 to 3. 5. After shifting to the updating of the learning value on the small area map, each learning value of at least two areas on the small area map and the area on the large area map including each learning value When the sum of each learning value is updated at a constant rate and at least two areas on the small area map are not in the same area on the large area map, the learning value is updated on the large area map. The air-fuel ratio learning control device for an engine according to any one of claims 1 to 3, wherein the control is performed for a predetermined period.