JP3067489B2 - Fuel supply control device for internal combustion engine - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a fuel supply control device for an internal combustion engine, and more particularly, to a learning correction technique for a fuel supply amount.

[0002]

2. Description of the Related Art Conventionally, an exhaust gas recirculation (hereinafter abbreviated as EGR) device for reducing a combustion temperature by recirculating a part of engine exhaust gas to an intake system and thereby reducing NOx emission has been known. ing. On the other hand, by detecting the air-fuel ratio of the engine intake air-fuel mixture via the oxygen concentration in the exhaust gas, the air-fuel ratio for learning the correction value of the fuel injection amount required to obtain the target air-fuel ratio for each engine operating state Learning control is known (Japanese Unexamined Patent Publication No.
No. 1439).

Further, by setting the target air-fuel ratio to an air-fuel ratio that is significantly leaner than the stoichiometric air-fuel ratio (for example, 22 to 25),
An engine aiming at improving fuel efficiency has also been proposed (see Japanese Patent Application Laid-Open No. 2-153243). Here, when EGR is executed when the fuel is burned at the lean air-fuel ratio, E
Since GR becomes a disturbance factor and greatly impairs combustion stability, EGR is executed only when the target air-fuel ratio is set near the stoichiometric air-fuel ratio, and EGR is performed during lean air-fuel ratio combustion.
Is desired to be stopped.

[0004]

In the case of such a configuration in which the EGR is selectively executed in accordance with the switching of the target air-fuel ratio, there is a recirculation of unburned gas and oxygen accompanying the EGR. Accordingly, the required level of the correction value required to obtain the target air-fuel ratio changes depending on the presence or absence of EGR. For this reason, for example, if the result learned during the execution of EGR is directly applied to the lean combustion control in which the EGR is stopped, an error occurs in the air-fuel ratio correction control if the target air-fuel ratio is set near the stoichiometric air-fuel ratio, for example. It will be.

Therefore, as described above, the stoichiometric air-fuel ratio (EG
R during the lean air-fuel ratio control (E
(GR stop state)
Even if an error occurs in the air-fuel ratio control, it is necessary to set the target of the lean air-fuel ratio to be richer than it should be, so that the combustion stability is not significantly impaired. There is a problem that the control range of the lean air-fuel ratio control is narrowed because it is determined from the request of the NOx emission amount.

Here, if a configuration is adopted in which the result learned at the time of EGR execution is not applied as it is at the time of EGR stoppage but is learned in each state (see Japanese Patent Application Laid-Open No. 4-1439), in the state where learning has progressed, Optimal learning correction can be performed regardless of whether EGR is performed or stopped. However, according to such a configuration, even when the engine load, the rotation speed, and the like are the same, the learning is individually performed in accordance with the execution and stop of EGR. Therefore, when the target air-fuel ratio is frequently switched, there is a problem that the learning effect cannot be easily obtained in each air-fuel ratio control state.

SUMMARY OF THE INVENTION The present invention has been made in view of the above-described problems, and can cope with a difference in correction request level depending on the presence or absence of EGR. It is intended to be realized.

[0008]

Therefore, a fuel supply control device for an internal combustion engine according to the present invention is configured as shown in FIG. In FIG. 1, the exhaust gas recirculation means recirculates a part of the engine exhaust gas to the engine intake system at a recirculation rate corresponding to the engine operating state. Further, the exhaust gas recirculation execution control means controls the switching of the execution and stop of the exhaust gas recirculation by the exhaust gas recirculation means in accordance with the engine operating conditions.

On the other hand, the correction value learning means corrects the amount of fuel supplied to the engine based on the combustion state of the engine detected by the combustion state detection means when the exhaust gas recirculation execution control means is performing exhaust gas recirculation. To learn a feedback correction value for performing the correction. The correction value storage means stores the feedback correction value learned by the correction value learning means for each engine operating state.

Further, when the exhaust gas recirculation is stopped by the exhaust gas recirculation execution control means, the stop time correction value setting means sets the exhaust gas recirculation rate at the time of learning the feedback correction value stored in the correction value storage means. And a feedback correction value corresponding to the exhaust gas recirculation stop state is set. Then, the fuel supply correction means corrects and controls the fuel supply amount to the engine based on the feedback correction value.

[0011]

With this configuration, the feedback correction value is learned while the exhaust gas recirculation is being executed, and the fuel supply amount is corrected using the learning result. On the other hand, when the exhaust gas recirculation is stopped, the stored value of the learning result at the time of executing the exhaust gas recirculation is corrected by the exhaust gas recirculation rate at the time of the learning.
A feedback correction value corresponding to the time when the exhaust gas recirculation is stopped is set.

That is, based on the learning result when the exhaust gas recirculation is performed and the exhaust gas recirculation rate when the learning is performed, the correction level required when the exhaust gas recirculation is stopped is estimated.
The fuel supply amount when the exhaust gas recirculation is stopped is corrected using the estimation result.

[0013]

Embodiments of the present invention will be described below. FIG. 2 is a diagram illustrating a system configuration of the embodiment. In FIG. 2, fuel injection valves 3a to 3d are provided for each cylinder in a branch portion of an intake manifold 2 of the engine 1, and intake air adjusted by a throttle valve 4 and the fuel injection valve 3a are provided.
A fuel-air mixture is formed by the fuel injected and supplied from -3d and is sucked into the cylinder.

The opening operation of the fuel injection valves 3a to 3d is controlled by a control unit 5 containing a microcomputer. The microcomputer incorporated in the control unit 5 includes a CPU 5a for performing various arithmetic processes, and a ROM 5b for storing various data necessary for the arithmetic processes. Signals from various sensors are input to the control unit 5 for the fuel injection control.

As the various sensors, there is provided a crank angle sensor 6 built in a distributor.
The crank angle sensor 6 outputs a signal for each unit angle and a signal for each reference piston position. Here, the engine speed Ne is calculated based on the detection signal from the crank angle sensor 6. An air flow meter 7 for detecting an engine intake air amount Qa is provided upstream of the throttle valve 4, and a water temperature sensor 8 for detecting a cooling water temperature Tw in a cooling jacket is provided. Is provided with a potentiometer type throttle sensor 9 for detecting the opening TVO.

Further, a sensor whose output changes in response to the oxygen concentration in the exhaust gas, which is correlated with the air-fuel ratio of the engine intake air-fuel mixture, is provided at the collecting portion of the exhaust manifold 10 of the engine 1. Wide-range air-fuel ratio sensor 11 (Sankaido Co., Ltd.) that can linearly detect lean to rich regions
(See "Internal combustion engine" issued on May 1, 1992) is provided as combustion state detecting means.

Further, a vehicle speed sensor 12 for extracting a rotation signal from an output shaft of a transmission (not shown) combined with the engine 1 is provided. On the other hand, the engine 1 of the present embodiment is provided with an exhaust gas recirculation device (EGR device) 13 as exhaust gas recirculation means. The EGR device 13 includes an EGR passage 14 that communicates with an exhaust system on the downstream side of the wide area air-fuel ratio sensor 10 and an intake system on the downstream side of the throttle valve 4, and an EGR control valve disposed in the EGR passage 14. It consists of 15. The EGR control valve 15 is controlled by the control unit 5. The control unit 5 adjusts the opening degree of the EGR control valve 15 based on EGR rate data set in advance for each engine operating state. Variable control.

Further, the control unit 5 sets the target air-fuel ratio of the engine intake air-fuel mixture to an air-fuel ratio near the stoichiometric air-fuel ratio and an air-fuel ratio substantially leaner than the stoichiometric air-fuel ratio (hereinafter simply referred to as “the air-fuel ratio”) according to the engine operating conditions. Lean air-fuel ratio). Then, a final fuel injection pulse width Ti is calculated in order to form an air-fuel mixture of the set target air-fuel ratio, and an injection pulse signal of the pulse width Ti is generated at a predetermined timing synchronized with engine rotation by the fuel injection valve. 3a-3d
Output to

The control unit 5 controls the EG based on the exhaust gas recirculation rate set in advance for each engine operating state.
Although the opening of the R control valve 15 is controlled, when the lean air-fuel ratio is set as the target air-fuel ratio, EGR becomes a disturbance factor and impairs combustion stability. The engine is stopped and EGR is executed only when the target air-fuel ratio is set near the stoichiometric air-fuel ratio. The EGR execution / stop switching control function of the control unit 5 corresponding to the target air-fuel ratio corresponds to the exhaust gas recirculation execution control means in the present embodiment.

FIG. 3 is a block diagram showing the fuel control function of the control unit 5 in this embodiment. In FIG. 3, the intake air amount Qa detected by the intake air amount detecting means A corresponding to the air flow meter 7
The basic fuel pulse width Tp is calculated by the basic fuel pulse width calculating means C based on the engine speed Ne detected by the engine speed detecting means B corresponding to the crank angle sensor 6.

On the other hand, based on the detection result of the vehicle operating state detecting means D such as the water temperature sensor 8 and the throttle sensor 9, the operating state correction coefficient calculating means E calculates an operating state for correcting the basic fuel pulse width Tp. A correction coefficient Coef is calculated. Further, based on the detection result of the combustion state detecting means F corresponding to the wide area air-fuel ratio sensor 10, a feedback correction coefficient L for bringing the actual air-fuel ratio close to the target air-fuel ratio.
alpha (feedback correction value) is the correction coefficient Lalpha stored at the corresponding grid point in the correction coefficient storage map H described later.
Is used as a reference value by the feedback correction coefficient calculating means G, and the calculated feedback correction coefficient Lalpha is updated and stored in the feedback correction coefficient storage map H (see FIG. 5), and the correction request level is set for each operating state. It has been learned. The function of storing the calculation result of the feedback correction coefficient calculating means G in the map H corresponds to the correction value learning means in the present embodiment.

Here, the feedback correction coefficient Lal
pha is learned and stored in the map H only when EGR is performed by the EGR device 13 under the condition that an air-fuel ratio near the stoichiometric air-fuel ratio is set as the target air-fuel ratio. When the lean air-fuel ratio is set as the target air-fuel ratio (the condition for stopping the EGR), the EGR
EGR correction rate REGR stored in correction rate storage means I
Based on (see FIG. 6), the map data learned at the time of performing the EGR is corrected by the lean operation correction coefficient calculating means J (stop-time correction value setting means), and the correction result is used as the feedback for the lean air-fuel ratio. It is set as the correction coefficient LLalpha.

In the fuel injection pulse width calculating means K (fuel supply correcting means), the basic fuel pulse width Tp, the operating state correction coefficient Coef, and the feedback correction coefficient Lalph are calculated.
The final fuel injection pulse width Ti is calculated based on a (or LLalpha). Next, according to the flowchart of FIG. 4, the state of the arithmetic processing performed according to the control block diagram shown in FIG. 3 will be described in detail.

The routine shown in the flowchart of FIG. 4 is executed at regular intervals (for example, 4 msec). As shown in the flowchart of FIG. 4, in the present embodiment, the functions of the correction value learning unit, the stop-time correction value setting unit, and the fuel supply correction unit are provided by the control unit 5 as software. It is assumed that the correction value storage means corresponds to a memory built in the control unit 5.

In the flowchart of FIG. 4, first, P
In step 1, the engine speed Ne calculated based on the detection signal of the crank angle sensor 6 is read. Next, at P2, the intake air amount Qa detected by the air flow meter 7 is read. In P3, the basic fuel pulse width Tp is calculated based on the engine speed Ne and the intake air amount Qa read in P1 and P2, respectively.

At P4, parameters indicating the operating state of the engine such as the cooling water temperature Tw detected by the water temperature sensor 8 are read. In P5, the operating state correction coefficient Coef is determined based on the information such as the cooling water temperature Tw read in P4.
Is calculated. At P6, the output value of the wide area air-fuel ratio sensor 10 is read, and at the next P7, a feedback correction coefficient Lalpha for correcting the basic fuel pulse width Tp is determined based on the read output value of the wide area air-fuel ratio sensor 10. The calculation is performed so that the detected actual air-fuel ratio approaches the target air-fuel ratio.

In P8, the feedback correction coefficient L
It is determined whether or not the driving condition allows learning of alpha. The operating conditions that can be learned are that at least feedback control is being performed on the stoichiometric air-fuel ratio (EGR is being executed),
It is assumed that the engine is in a steady state, for example. When it is determined in P8 that the learning condition is satisfied, the routine proceeds to P9, where the calculated feedback correction coefficient Lalpha is used as correction request data corresponding to the current operating state (engine load, rotation speed) and the engine load (basic fuel A table in which grid points are determined using the pulse width Tp) and the rotation speed Ne as parameters (feedback correction coefficient storage map H: see FIG. 5).
To learn by updating.

The learning result is referred to for each operation state, and the feedback correction coefficient Lalpha is calculated based on the reference value. In P10, it is determined whether or not the operating range is such that the lean air-fuel ratio should be set to the target air-fuel ratio. In particular,
As shown in FIG. 7, the engine load (basic fuel pulse width T
p) and the engine speed Ne correspond to a lean air-fuel ratio control region set as a parameter, and the cooling water temperature Tw
Is equal to or higher than a predetermined temperature, and is still in a steady operation (the change rate of the throttle valve opening TVO is small) as a condition of the lean air-fuel ratio control.

When it is determined in P10 that the lean air-fuel ratio control condition is not satisfied, that is, when the operating condition is to control the air-fuel ratio to a target air-fuel ratio near the stoichiometric air-fuel ratio, P10
Proceeding to 15, the basic fuel pulse width Tp is corrected by the feedback correction coefficient Lalpha and the operating state correction coefficient Coef calculated as described above, and the corrected result is used as the final fuel injection pulse width Ti (= Tp × Coef). × Lalpha)
Set to.

On the other hand, when it is determined in P10 that the condition for the lean air-fuel ratio control is satisfied, the injection is controlled with the lean air-fuel ratio as a target, and the EGR is stopped. Thus, the learning and storage of the feedback correction coefficient Lalpha in P9 is not performed. Here, since the feedback correction coefficient Lalpha learned and stored in P9 is a value learned when the EGR is executed with the target air-fuel ratio being near the stoichiometric air-fuel ratio, when the EGR is stopped, , The effect of the recirculation of unburned gas and oxygen by EGR, the correction coefficient does not match the required level even at the same engine load and rotational speed. If the learning result is used as it is, a control error (target lean air-fuel ratio) Is controlled to a leaner side).

However, since the difference in the correction request level is due to the presence or absence of EGR, the EGR rate is obtained from the learning result at the time of EGR execution and the EGR rate data at the time of learning.
The required correction level at the time when is stopped can be estimated. Therefore, if it is determined in P10 that the lean air-fuel ratio control condition is satisfied, the process proceeds to P11, and the current operating condition is determined based on the EGR rate map (see FIG. 8) referred to in the control of the EGR control valve 15. The EGR rate at the time of learning the feedback correction coefficient Lalpha corresponding to is obtained.

In the next P12, a correction rate REGR for correcting the feedback correction coefficient Lalpha learned at the time of EGR execution to a value corresponding to the EGR stop state is set based on the EGR rate obtained in P11. (Fig. 6
reference). At P13, the correction rate REGR set at P12 is
By multiplying the feedback correction coefficient Lalpha stored in correspondence with the corresponding operating condition on the learning map, the feedback correction coefficient Lalpha learned in conformity with the execution state of EGR is corrected in conformity with the EGR stop state. Convert to coefficient LLalpha.

At P14, the basic fuel pulse width Tp is corrected using the feedback correction coefficient LLalpha and the operating state correction coefficient Coef corresponding to the EGR stop state obtained at P13, and the final fuel pulse width Tp for lean air-fuel ratio control is obtained. The fuel injection pulse width Ti (= Tp × Coef × LLalpha) is obtained. According to such a configuration, learning of the feedback correction coefficient Lalpha is performed only when the target air-fuel ratio is set near the stoichiometric air-fuel ratio and EGR is performed. , The value is corrected to a value corresponding to the lean air-fuel ratio control in which the EGR is stopped, in other words, a value corresponding to the difference in the correction request level depending on the presence or absence of the EGR, and is used during the lean air-fuel ratio control.

Therefore, using the learning result at the time of EGR execution, appropriate learning correction can be performed not only at the time of EGR execution but also at the time of EGR stop, and immediately after the EGR is stopped (to the lean air-fuel ratio control). Immediately after the transition), appropriate learning correction can be expected. Therefore, there is no need to make a rich shift in the target lean air-fuel ratio in advance in anticipation of the deviation of the learning correction control, and it is possible to improve the fuel efficiency by promoting the leaning of the target air-fuel ratio.

In the above embodiment, the engine load (basic fuel pulse width Tp) and the engine speed N are stored in the EGR rate table referred to when obtaining the EGR rate and the learning table for learning and storing the feedback correction coefficient Lalpha.
If the grid points determined by e are common, there is no need for interpolation calculation for comparison and reference as data corresponding to the same engine load and rotation speed, and an increase in calculation speed and improvement in control accuracy can be expected.

In the above embodiment, the air-fuel ratio feedback learning control for obtaining the target air-fuel ratio using the wide-range air-fuel ratio sensor 10 as the combustion state detecting means has been described.
For example, a configuration may be adopted in which an in-cylinder pressure sensor that detects the pressure in the cylinder is provided as the combustion state detecting means, and the combustion pressure is feedback-controlled. Next, a second embodiment will be described.

FIG. 9 is a control function block diagram of the second embodiment. The control block diagram of the first embodiment shown in FIG. 3 is different from the control block diagram of FIG. Only the differences. The lean combustion feedback correction coefficient storage map L (stop-time correction value storage means) stores the learning result at the time of execution of EGR in the lean operation correction coefficient calculation means J based on the EGR rate at the time of the learning. Coefficient LLalph corrected and corrected for EGR stop (for lean combustion)
a is a map that stores a for each operating state as a learning value LMLalpha for lean combustion.

That is, in the second embodiment, the learning result at the time of EGR execution is corrected to a value suitable for EGR stop based on the EGR rate, and the corrected value is stored as the EGR stop learning value. By doing so, learning for EGR stop (lean combustion) can be performed based on the learning result at the time of EGR execution. Therefore, relatively accurate learning correction is performed from the beginning of the transition to lean combustion, and furthermore,
By performing the learning on the basis of the learning correction level (correction coefficient obtained from the learning result at the time of performing the EGR), it is possible to perform the learning capable of responding to the correction request for each operation state with high accuracy at the time of lean combustion. ing.

Here, according to the flowchart of FIG.
The state of the learning control in the second embodiment will be described in detail.
In the flow chart of FIG. 10, since the processes from P21 to P26 are exactly the same as those from P1 to P6 in the flow chart of FIG. 4, the description of each step of P21 to P26 is omitted, and the processing contents from P27 will be described.

In P27, it is determined whether or not the operating range is such that the lean air-fuel ratio should be set to the target air-fuel ratio. Specifically, as shown in FIG. 7, the engine load (basic fuel pulse width Tp) and the engine speed Ne correspond to a lean air-fuel ratio control region set as parameters, and the cooling water temperature T
that w is equal to or higher than a predetermined temperature,
This is the condition for lean air-fuel ratio control.

If it is determined in P27 that the lean air-fuel ratio control condition is not satisfied, and the air-fuel ratio near the stoichiometric air-fuel ratio is set as the target air-fuel ratio, the process proceeds to P28. At P28, the actual air-fuel ratio detected based on the output value of the wide area air-fuel ratio sensor 10 is determined based on the feedback correction coefficient Lalpha for correcting the basic fuel pulse width Tp based on the learning value stored in the map. The calculation is performed so as to approach the target air-fuel ratio.

In P29, the feedback correction coefficient L
It is determined whether or not the driving condition allows learning of alpha. If it is determined in P29 that the learning condition is satisfied, the program proceeds to P30
To the calculated feedback correction coefficient Lalph
a is used as correction request data corresponding to the current operating state (engine load, rotation speed), and the engine load (basic fuel pulse width T
Table in which grid points are determined using p) and rotation speed Ne as parameters (feedback correction coefficient storage map H)
To learn by updating.

At P31, the calculated feedback correction coefficient Lalpha and operating state correction coefficient Coef are calculated.
To correct the basic fuel pulse width Tp to set the final fuel injection pulse width Ti corresponding to the normal air-fuel ratio control state. On the other hand, when it is determined in P27 that the lean control condition is satisfied, the program proceeds to P32, in which it is determined whether or not the target lean air-fuel ratio and the actual air-fuel ratio detected by the wide-range air-fuel ratio sensor 10 substantially match. .

When it is determined in P32 that there is a difference between the target lean air-fuel ratio and the actual air-fuel ratio, the program proceeds to P33, in which the EGR rate applied when the EGR is executed under the current operating conditions is mapped. (See FIG. 8), and in the next P34, an EGR correction rate REGR is obtained from the EGR rate (see FIG. 6). In P35, the feedback correction coefficient Lalpha learned and stored at the time of EGR execution (normal air-fuel ratio control) is multiplied by the EGR correction rate REGR, thereby obtaining a correction coefficient LLalph corresponding to the EGR stop state.
Convert to a.

Subsequently, at P36, the feedback correction coefficient LLalpha calculated at P35 is used as a lean combustion learning value LMLalpha as a map for lean combustion (FIG. 11).
When the process proceeds from P36 to P38, the final fuel injection pulse width Ti for the lean air-fuel ratio control is determined based on the feedback correction coefficient LLalpha obtained by the correction using the EGR correction rate REGR. Is calculated.

On the other hand, when it is determined in P32 that the target lean air-fuel ratio substantially coincides with the actual air-fuel ratio, the feedback correction coefficient LMLalpha of the corresponding operating state is determined by referring to a map for lean combustion (see FIG. 11). And then P38
Then, the final fuel injection pulse width Ti is calculated based on the map reference value. As described above, in the second embodiment, the map for learning and storing the feedback correction coefficient Lalpha is used during the stoichiometric air-fuel ratio control (when the EGR is executed).
Map and for lean air-fuel ratio control (when EGR is stopped)
The map for the stoichiometric air-fuel ratio control is normally learned, while the map for the lean air-fuel ratio control has an air-fuel ratio deviation due to learning correction using map data. In such a case, when the result learned at the time of EGR execution is rewritten to a value obtained by correcting based on the EGR rate at that time, and when the target lean air-fuel ratio can be controlled by learning correction using map data, The correction is performed using the map data as it is.

According to this configuration, if the learning in the lean air-fuel ratio control state is not progressing and it is not possible to control the air-fuel ratio close to the target lean air-fuel ratio by the correction using the learning value,
Since the learning value obtained by correcting the result learned in the EGR execution state based on the EGR rate is applied, learning correction close to the required level can be performed from an early stage. By further progressing the learning for each operating state from the state, the correction control corresponding to the difference in the correction request for each operating state with high accuracy can be performed early in the lean air-fuel ratio control state.

[0048]

As described above, according to the present invention, E
Since the result learned in the GR execution state is corrected on the basis of the EGR rate at the time of performing the learning, the result is corrected to a correction value corresponding to the EGR stop state, and is used.
There is an effect that the effect of the learning control can be obtained with good response while responding to the difference in the correction request depending on the presence or absence of GR.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a basic configuration of the present invention.

FIG. 2 is a system schematic diagram showing one embodiment of the present invention.

FIG. 3 is a block diagram showing a control function of the first embodiment.

FIG. 4 is a flowchart showing control contents in the first embodiment.

FIG. 5 is a diagram showing a learning map of a feedback correction coefficient.

FIG. 6 is a diagram showing a correction rate REGR according to an EGR rate.

FIG. 7 is a diagram showing a lean combustion region.

FIG. 8 is a diagram showing a map of an EGR rate.

FIG. 9 is a block diagram showing a control function of the second embodiment.

FIG. 10 is a flowchart illustrating control contents according to the second embodiment.

FIG. 11 is a diagram showing a learning map for lean combustion.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Internal combustion engine 2 Intake manifold 3a-3d Fuel injection valve 4 Throttle valve 5 Control unit 6 Crank angle sensor 7 Air flow meter 8 Water temperature sensor 9 Throttle sensor 10 Exhaust manifold 11 Wide area air-fuel ratio sensor 12 Vehicle speed sensor 13 EGR device 14 EGR passage 15 EGR Control valve

Continuation of the front page (56) References JP-A-3-210038 (JP, A) JP-A-60-6045 (JP, A) JP-A-4-1439 (JP, A) (58) Fields investigated (Int .Cl. ⁷ , DB name) F02D 41/00-41/40 F02M 25/07 550

Claims (1)

(57) [Claims] A combustion state detection means for detecting a combustion state of the engine; an exhaust gas recirculation means for recirculating a part of the engine exhaust to an engine intake system at a recirculation rate according to an engine operation state; Exhaust gas recirculation execution control means for switching and controlling the execution / stop of recirculation according to the engine operating condition; and, when exhaust gas recirculation is being executed by the exhaust gas recirculation execution control means, an engine recirculation detected by the combustion state detection means. Correction value learning means for learning a feedback correction value for correcting the fuel supply amount to the engine based on the combustion state; and a correction value for storing the feedback correction value learned by the correction value learning means for each engine operating state. Storage means, and when the exhaust gas recirculation execution control means stops the exhaust gas recirculation, the feedback correction value stored in the correction value storage means is learned at the time of learning. A stop-time correction value setting means for correcting based on the gas recirculation rate and setting a feedback correction value corresponding to the exhaust gas recirculation stop state; and a fuel for correcting and controlling a fuel supply amount to the engine based on the feedback correction value A fuel supply control device for an internal combustion engine, comprising: supply correction means.