JPH0742886B2 - Control device for internal combustion engine - Google Patents

Description

TECHNICAL FIELD The present invention relates to an improvement in a control device for an internal combustion engine.

(Prior Art) An engine control device is provided with a sensor that detects knocking, and when knocking occurs, the ignition timing is controlled to a retard side or the air-fuel ratio is changed to a side where knocking does not occur (rich side). There is something to do (for ignition timing control,
(See JP-A-60-26172). Here, a level that produces a light knocking state is set as a predetermined level, and the ignition timing and the air-fuel ratio are controlled with this predetermined level as a target value (such control is hereinafter referred to as "knock control").
Is performed. This is because strong knocking, which adversely affects the engine, must be avoided, but light knocking does not itself have a negative effect on the engine, and the fuel efficiency is rather improved by improving the combustion efficiency.

On the other hand, it is desirable to control the ignition timing so that the maximum torque can be obtained with the same fuel amount from the viewpoint of improving the output. Therefore, there is one in which control is performed so that the ignition timing becomes MBT (minimum advance value giving maximum torque) (this control is hereinafter referred to as "MBT control") (see Japanese Patent Laid-Open No. 61-16268). . For example, the crank angle at which the in-cylinder pressure of the engine is the maximum
Called. ) The ignition timing is corrected and controlled by the feedback amount based on the deviation between the detected value and the target value so that the detected value of Θpmax or the indicated mean effective pressure Pi matches the predetermined target value.

Also, knocking control and MBT control are often combined. Here, in the relationship between the two, knocking control is prioritized. Below, the control of the combination of both is referred to as “knock-MBT control”.

Further, for the purpose of eliminating errors due to individual differences and changes with time, there has been proposed a device provided with a learning function for ignition timing control and air-fuel ratio control (for air-fuel ratio control, Japanese Patent Laid-Open No. 55-
(See Japanese Patent No. 96339).

(Problems to be solved by the invention) By the way, when the engine is under high load, the engine is likely to knock and a large output is required. Therefore, it is necessary to increase the output by thickening the air-fuel mixture while performing knock control. However, even in the high load range, when it is in a steady state, the output is not required as much as in the transient state, so if the air-fuel ratio can be made as lean as possible within the range where running performance is not wasted, fuel consumption is wasted. Can be suppressed.

If you try to control like this, knock-MB
It is necessary to control both the T control and the air-fuel ratio control, that is, both the ignition timing and the air-fuel ratio as required. However, in the conventional example, since the content of the knock control is control of only one of the ignition timing and the air-fuel ratio, both of them are not controlled. This is because the main focus is only on avoiding knocking, and it is not taken into consideration that the lean air-fuel ratio is made lean and the fuel efficiency is improved at the time of high load steady state.

On the other hand, when the high load is steady, it is necessary to take measures against the heat load in addition to knocking, that is, control so that the exhaust gas temperature does not exceed a predetermined value. This is because, for example, if the temperature of the exhaust valve wake exceeds a predetermined value at the time of high rotation or the temperature of the catalyst wake exceeds the predetermined value at the time of low rotation, the engine body and the catalyst are damaged.

Furthermore, considering that the air-fuel ratio and the ignition timing vary due to the difference in the compression ratio, the detection accuracy of the air amount sensor, or the cooling efficiency of the combustion chamber, and this deviates from the maximum torque point and the minimum fuel consumption point, It is also necessary to add a learning function.

The present invention aims to solve such conventional problems.

(Means for Solving Problems) The present invention, as shown in FIG. 1, includes a sensor 1 for detecting knocking, and a means 2 for converting a detection signal of the knocking sensor 1 into a signal according to the knocking intensity. , A means 3 for detecting the maximum in-cylinder pressure crank angle Θpmax or the indicated mean effective pressure Pi (these are referred to as "MBT correlation amount"), and a signal corresponding to the knocking intensity and a target value thereof while keeping the air-fuel ratio constant. And the knocking intensity exceeds a predetermined value, based on the deviation between the MBT correlation amount and the ignition timing feedback amount (for example, β) obtained from each of the detected value and the target value of the MBT correlation amount and the learning value (LADV) related to the ignition timing. Knock to bring the ignition timing closer to MBT within the range where there is no knock-MBT control means 4
And means 5 for updating the learning value related to the ignition timing on the basis of the feedback amount (β) of the ignition timing that minimizes the deviation between the detected value of the MBT correlation amount and its target value, and keeps the ignition timing constant. However, the air-fuel ratio feedback amount (for example, δ × α) obtained from the deviation between the signal corresponding to the knocking intensity and its target value and the learning value (LF
BA) for performing knock control, and a learning value for the air-fuel ratio based on an air-fuel ratio feedback amount (δ × α) that minimizes the deviation between the detected value of the MBT correlation amount and its target value. And means 10 for determining whether or not the exhaust temperature (for example, the temperature of the exhaust wake or the catalyst wake) is a control condition that is equal to or less than a predetermined value at the time of high load steady state,
And a means 11 for switching the knock-MBT control by the ignition timing and the knock control by the air-fuel ratio at a predetermined timing (for example, every predetermined time or every predetermined rotation) when it is determined that this control condition is satisfied. . In addition, 8 is an operating condition detecting means, and 9 is an exhaust temperature detecting means.

(Operation) Knock by ignition timing while maintaining a constant air-fuel ratio
In the combined control of the MBT control and the knock control by the air-fuel ratio while keeping the ignition timing constant, the point where the deviation between the detected value of the MBT correlation amount and its target value is the minimum ensures the maximum torque. Meanwhile, it is sought as a point (fuel efficiency control point) where the fuel consumption is the lowest. This prevents wasteful consumption of fuel while obtaining maximum torque during steady high load where output is not required as much as during transition.

Moreover, each control of the two merged controls does not change the ignition timing and the air-fuel ratio at the same time, but only one of them is changed. Therefore, compared with the case where the ignition timing and the air-fuel ratio are changed at the same time. Control is simplified, and the response of the control is enhanced.

Furthermore, since each control is executed at a predetermined timing, for example, at every predetermined time, it means that an optimum control point is being searched for at a constant cycle, and thus even if there is a slight driving change, it is possible to sufficiently cope with it. it can.

Further, by ensuring that the detected exhaust temperature falls within the exhaust temperature limit, the exhaust function and the catalyst are protected from the heat load. Needless to say, the control is maintained with high accuracy by adding the learning function.

(Embodiment) An embodiment in which the present invention is applied to a V-type 6 cylinder engine is shown in FIG. 2. 23B of 23A is a washer cylinder pressure sensor to be mounted on a washer portion of an ignition plug, and outputs of sensors 23A and 23B. The charge amount to be converted is converted into a voltage value by the charge amplifiers 24A and 24B, and the output of this charge amplifier is input to the multiplexer 40 as a cylinder pressure signal. Using this in-cylinder pressure signal, the detected value of the in-cylinder pressure maximum crank angle (Θpmax) is calculated.

On the other hand, the output of the charge amplifier is the knocking signal detection circuit 25.
Inputs to A and 25B to separate only high frequency signals specific to knocking, and this knocking signal is also multiplexed.
Entered in 40.

A crank angle sensor 29 generates a signal for each unit angle (for example, 1 °) of the clan angle and a signal (reference signal) for each predetermined angle (120 °), and is input to the input / output port (I / O) 39. . The engine speed N is obtained from the unit angle signal, and the cylinder is discriminated from the reference position signal.

31 is a sensor that detects the intake air amount Qa (for example, a flap type or hot wire type air flow meter) 32 is a sensor that can detect not only the stoichiometric air-fuel ratio but also other air-fuel ratios, and 33 and 34 are exhaust air, respectively. The sensors detect the exhaust temperatures of the valve downstream and the catalyst downstream, and the signals from each sensor are input to the multiplexer 40.

Reference numeral 35 is a control unit to which various signals from the sensors are input, and includes a CPU 36, ROM 37, RAM 38 and I / O 39, a multiplexer 40, an A / D converter 41 and a digital port 42. In the control unit 35, the fuel injection pulse width (output injection pulse width) Ti to be output and the ignition advance value (output advance value) to be output based on the operating condition signal and each feedback signal of the air-fuel ratio and the ignition timing. ADV is calculated respectively and converted into an injection pulse and an ignition pulse, which are output to a fuel injection valve and an ignition device (not shown).

Now, FIG. 3 shows the effects of the air-fuel ratio and the ignition timing on the engine under steady conditions (constant speed and constant load) near the throttle valve fully open. The procedure is as follows. The cylinder pressure maximum crank angle Θpam as the MBT correlation amount
It will be explained in the case of adopting ax. The characteristics of the knock limit, the exhaust temperature limit and the MBT line shown in FIG. 3 are not limited to the ones that are determined at the time of one high load steady state, and the relative position network of each characteristic changes due to a slight difference in operation change.

(1) At high load, the feedback control of the air-fuel ratio is stopped, the air-fuel ratio is set to the rich side, and the ignition timing is set to the retard side. In this case, the control is started when it is determined that it is the steady time. Now, suppose that the control start is point L in the figure.

(2) First, a knock -MBT control by the ignition timing as late Θpmax of detecting the air-fuel ratio as constant becomes its target retarded (described later theta ₂₎ at that time. When this control is performed for a predetermined time, the absolute value (| ΔΘpmax |) of the deviation between the detected value of Θpmax and its target value is settled to a point where it becomes a minimum, and the point settled in the figure is point M.

(3) When the M point is searched, the learning value (LADV) regarding the ignition timing is updated based on the feedback amount (β) of the ignition timing at the M point.

(4) This time, the ignition timing at that time is made constant, and knock control by the air-fuel ratio is performed. However, | ΔΘpmax | is stored in association with the feedback amount (δ × α) of the air-fuel ratio that changes during knock control. When this knock control is performed for a predetermined time, the knock limit N point is reached.

(5) When N points are searched, the stored group of | ΔΘ
The learning value (LFBA) regarding the air-fuel ratio is updated based on the feedback amount (δ × α) of the air-fuel ratio at the point where the value is the minimum value from pmax |.

(6) Within the exhaust temperature limit as long as the high load is steady, the above (2)
The processes (5) to (5) are repeated at a predetermined timing.

To summarize these controls, knock-MBT control based on ignition timing and knock control based on air-fuel ratio are alternately performed to make the air-fuel ratio as lean as possible while seeking for the maximum torque point. Is to be performed within the exhaust temperature limit. However, FIG. 3 is only an example, and the control including this example is performed in FIG. That is, FIG. 4 shows the flow of the entire control, and has a so-called OS meaning.

In FIG. 4, the knock-MBT control (step) based on the ignition timing that has been conventionally performed is used in its entirety, and the air-fuel ratio control (step) is slightly modified to perform the combined control of the present invention. Therefore, these prerequisite controls will be described first.

(I) Air-fuel ratio control In the air-fuel ratio control, the output injection pulse width Ti is calculated by Ti = Tp × TFBA × LFBA × α + Ts as shown in FIGS. 5 (A) and 5 (B) ( Step 64). However, Tp (= K × Qa
/ N, where K is a constant. ) Is the basic pulse width, TFBA is the basic value that gives the target air-fuel ratio, LFBA is the learned value of the air-fuel ratio, α
Is a feedback amount calculated based on the signal from the air-fuel ratio sensor 32, and Ts is an invalid pulse width. In addition, with TFBA
LFBA is read from the table with N and Qa as parameters (steps 51 and 52). Ti is stored in the I / O register in angle synchronization (step 65).

However, unlike the conventional example, in this example, a lean coefficient δ of the air-fuel ratio is newly introduced, and Ti is calculated by Ti = Tp × TFBA × LFBA × δ × α + Ts (steps 62 and 64). The reason why δ is introduced here is that it is necessary when knock control by the air-fuel ratio is performed in the step of FIG. Therefore, in the step, δ only has a value of 1.0.

(II) Knock-MBT control The control is divided into knock control and MBT control depending on whether knocking occurs. In both cases, the feedback amount of the ignition timing is understood as the same symbol β. As shown in FIG. 6 (C), whether or not knocking has occurred is determined by the integrated value (S) of the knocking signal obtained during one combustion of one cylinder and its moving average value ().
(Steps 81-83,86) and S
If ≧ SL + (where SL is a predetermined value), it is determined that knocking has occurred and a knock flag (FKN) is set (steps 83, 8).
Four). Here, the flag being set means that the value of the flag becomes 1. The same meaning will be used below.

(I) Knock control As shown in FIG. 6 (B), when the value of FKN is determined, if there is knock, a value delayed from the previous feedback amount (β ′) by a predetermined value d ₁ (> 0). (Β′−d ₁ ) is set as the current feedback amount β (steps 73 and 74). In addition,
"'" Means the previous value. This symbol will be used interchangeably below.

(Ii) MBT control If there is no knock, the current feedback amount β
Is calculated by β = β ′ + k ₁ × ΔΘ pmax (steps 73 and 75). However, Derutashitapmax In the equation the deviation of the detected value of Θpmax and its target value theta _2, k ₁ is a constant.

Here, in the waveform diagram of the in-cylinder pressure shown in FIG. 7, predetermined crank angles Θ ₁ , Θ ₂ and Θ ₃ are determined after the compression top dead center (TDC), and the in-cylinder pressure P corresponding to the crank angles of these 3 points is determined. _{If 1} , P ₂ and P ₃ are detected, the detected value of Θpmax is obtained from this detected value by using a quadratic curve approximation, and ΔΘpmax is calculated from the deviation of the obtained detected value of Θpmax from a predetermined target value Θ _2. Is required.

(Iii) Then, as a whole, the output advance value ADV is calculated by ADV = TADV + LADV + β as shown in FIG. 6 (B) (step 76). However, TADV is the basic advance value, LADV is the learning value of the ignition advance value, and TADV and LADV are N.
It is read from the table with Qa as a parameter (6th
(A). ADV is also stored in the I / O register in angle synchronization (step 77).

Now, returning to FIG. 4, the contents of FIG. 5 (A) to FIG. 7 described above are steps, but each step of FIG. 4 described below is also configured corresponding to these steps. This will be described in detail with reference to the routines shown in FIGS. Note that these routines are executed at regular intervals (for example, 4m
every sec) or every predetermined crank angle.

The control of the present invention is started when it is determined in step that the control condition is satisfied. FIG. 8 shows this control condition determination routine, in which the engine load equivalent amount (for example, intake air amount Qa and intake throttle opening) is equal to or greater than a predetermined value (Qa ₀ ) and the difference between the engine load equivalent amount of the previous time and this time. (ΔQa) and the difference between the previous and current rotational speeds (ΔN) are both predetermined values (ΔQa ₀ and Δ
If it is equal to or less than N ₀ ), it is determined that the high load is in a steady state and a flag (FCONT) is set (steps 91 to 94).

Then, when it is determined that the high load is in steady state (FCON
T = 1), proceed to the step, and judge whether the detected exhaust temperature is within the exhaust temperature limit. This is executed by the routine shown in FIG. In the figure, the exhaust temperature (about the temperature EXTEMP of the exhaust valve wake at high speed, about the temperature CATTEMP of the catalyst wake at low speed) and the predetermined value (Te for EXTEMP, Tc for CATTEMP) are shown. Compared (step 10
2,105), that is, EXTEMP and Te are compared, CATTEMP and Tc are compared, and there are two flags (FOK1 and FOK2) depending on the comparison result.
Stand up. Here, the numbers 1 and 2 shown in each flag mean high rotation speed and low rotation speed, respectively. Note that Te and Tc are read from a table having N and Qa as parameters.

When it is determined that the exhaust temperature is within the exhaust temperature limit based on the values of these two flags (FOK1 = 1 or FOK2 =
1) The process proceeds to step, and at predetermined time intervals (for example, every few seconds), step and step are alternately switched to be performed.

Among these steps, in the step, knock-MBT control by ignition timing is performed while keeping the air-fuel ratio at that time constant. Except that the air-fuel ratio is kept constant, knock-MBT control by ignition timing is similar to that executed in steps, and the contents of the steps are shown in FIGS. 6 (A) and 6 ( As described above in B). As a result, the point where the absolute value (| ΔΘpmax |) of the deviation between the detected value of Θpmax and its target value becomes the minimum is searched for within a predetermined time.

Next, in step, the learning value regarding the ignition timing is updated based on the feedback amount (β) of the ignition timing at the point where | ΔΘpmax | is minimized. For example, the learning value (LADV) regarding the ignition advance value is calculated by using LADV = LADV '+ β.

On the other hand, in step, knock control based on the air-fuel ratio is performed while keeping the ignition timing constant. In the knock control based on the air-fuel ratio, the air-fuel ratio is set to the rich side when it is determined that there is knock, and the air-fuel ratio is set to the lean side when there is no knock. This is because, as is clear from FIG. 3, the air-fuel ratio may be set to the rich side in order to move the operating point away from the knock limit.

The tenth routine performs such control, and when it is determined that there is knocking, the previous air-fuel ratio leaning coefficient (δ ′) is set to the predetermined value (d ₂ , but d ₂ > 0) for the fuel increase. ) Is calculated as the leaning coefficient (δ) of this time, and conversely, when there is no knock, a predetermined value (d ₃ ,
However, δ is the value obtained by subtracting d ₃ > 0 (step
111,112, 111,113). Even if the exhaust temperature exceeds the exhaust temperature limit in the step, it is sufficient to set the air-fuel ratio to the rich side from FIG. 3 in order to return it to the exhaust temperature limit. Therefore, the air-fuel ratio enrichment control is performed in the step. ing. FIG. 11 shows the contents of the steps, and the predetermined value d ₄ (> 0) corresponds to d ₂ in FIG.

Then, in step 114 of FIG. 10, feedback control of the air-fuel ratio is performed using the leaning coefficient δ. What is executed by this control is the contents of FIG. 5 (A) and FIG. 5 (B), and the air-fuel ratio control is executed with δ × α as the feedback amount. As a result, the operating point reaches the knock limit after a lapse of a predetermined time.

However, although not shown, .THETA.pmax is detected also in step, and | .DELTA..THETA.pmax | is stored corresponding to the feedback amount (.delta..alpha.) Of the air-fuel ratio for each predetermined crank angle.

Then, in step, | ΔΘ stored for a predetermined time
δ × α that minimizes | ΔΘpmax | from the data group of pmax |
And the selected air-fuel ratio feedback amount (δ × α)
The learning value for the air-fuel ratio is updated based on For example, the learning value LFBA regarding the air-fuel ratio is calculated by using δ × α as LFBA = LFBA ′ × (δ × α).

In this way, not only the knock-MBT control by the ignition timing while keeping the air-fuel ratio constant, but also the knock control by the air-fuel ratio while keeping the ignition timing constant is performed by controlling both the air-fuel ratio and the ignition advance value. By doing so, the maximum torque is obtained, and the air-fuel ratio and the ignition advance value that give the operating point where the fuel consumption is the lowest are obtained.

Moreover, each control of step and is alternately executed every predetermined time (step). This means that the optimum control point cannot be immovable because each of the characteristics shown in FIG. 3 relatively changes due to a difference in load condition even though it is in a high load steady state. Therefore, the optimum control point, which fluctuates due to the difference in load conditions, is searched by performing the control every predetermined time.

The predetermined time that determines the switching cycle is the time that is expected to settle to the point where | ΔΘpmax | becomes minimum by the knock-MBT control by the ignition timing, the knock-MBT by the air-fuel ratio.
It is necessary to set the time taken to reach the knock limit by control and the responsiveness.

Next, the operation of this example will be described with reference to the example of FIG. 3. From the point L of the control start point, knock-MBT control by ignition timing is performed while keeping the air-fuel ratio at that time constant, and the MBT point is Seeking (step ~). In the figure, M point is MB
If T is set, the maximum torque can be obtained with the same air-fuel ratio by using the ignition timing for the same point.

Then, from the point M, while maintaining the ignition timing at that time constant, knock control by the air-fuel ratio is performed and the point N of the knock limit point is sought, and in the meantime, | Θpmax |
The point at which is minimal is selected (step ,,).
In the figure, if the MBT line is horizontal, the fuel consumption is also lowest at the N point, which is also the knock limit point. Therefore, N
The point becomes the optimum control point, and the maximum torque can be obtained within the knocking limit and the fuel consumption can be minimized by performing the control using the air-fuel ratio and the ignition timing at the N point. In other words, even in the high load range where knock-MBT control is performed, fuel efficiency can be improved by making the air-fuel ratio lean as much as possible in a steady state.

Note that the N point does not always become the practically optimum control point. The MBT line may not be horizontal as shown in FIG. 3 but may have a slight inclination. In this case, the point M is selected as the optimum control point, but in reality, the intersection point P between the MBT line and the knock limit is set. This is because it is more suitable as the optimum control point. However, due to differences in operating conditions
There is no doubt that the N point will be selected when the MBT line becomes nearly horizontal, and it can be said that such control is practically acceptable as a whole control.

Moreover, in terms of control, rather than changing the air-fuel ratio and the ignition timing at the same time so as to follow the MBT line from the M point to the P point, changing the air-fuel ratio alone with the ignition timing kept constant simplifies the control. The responsiveness is enhanced.

Further, since the learning function is provided, the control accuracy can be kept high without being affected by the compression ratio, the detection accuracy of the air amount sensor, the variation in the cooling efficiency of the combustion chamber, and the like.

Further, since the relative relationship of each characteristic shown in FIG. 3 is established in a certain high load steady state, the MBT line may be shifted upward, for example, due to a slight shift of the engine load after finding the optimum control point N. . Even in a high load condition where Qa ≧ Qa ₀ is satisfied in Steps 91 and 92 of FIG. 8 showing a control condition determination routine and a steady condition where ΔQa ≦ Qa ₀ is satisfied,
This is because Qa changes within that range.

On the other hand, according to this example, regardless of the driving change,
The knock-MBT control by the ignition timing and the knock control by the air-fuel ratio are alternately performed in a constant cycle (step).
That is, since the control is configured to search for the optimum control point at a fixed cycle without immobilizing the optimum control point, it is possible to sufficiently cope with a difference in operation change that does not exceed ΔQa _0. Of.

Furthermore, since the actual exhaust temperature is detected and the control is performed only when the detected temperature falls within the exhaust temperature limit (step), compared with the case where the value is limited to a predetermined constant value, Therefore, the control accuracy in terms of measures against heat load, which is a problem when the load is high, becomes good.

Next, another embodiment uses the indicated mean effective pressure (Pi) instead of Θpmax. In this case, the change caused by adopting Pi is that the detected value of Pi is stored in association with the feedback amount of the air-fuel ratio in the step of FIG. 4 and | Θpmax
Instead of minimizing |, each feedback amount that maximizes the Pi detection value is selected.

Also in this example, the same operational effects as those in FIG. 1 are exhibited.

(Effect of the Invention) As described above, according to the present invention, when it is determined that the exhaust temperature is in the operation range of the predetermined value or less at the time of high load steady state, the ignition timing is changed while keeping the air-fuel ratio constant. Knock-MBT control and knock control by air-fuel ratio while keeping ignition timing constant are performed together, and based on each feedback amount that minimizes the deviation between the detected value of MBT correlation amount and its target value in each control. While updating the learning value related to the ignition timing or the air-fuel ratio, each control is switched at predetermined timing and alternately performed, so maximum torque can be obtained and fuel consumption can be minimized even if there is a slight change in operation. The optimum control point is searched periodically and with good response, which enables automatic operation to protect the exhaust pipe and catalyst and to obtain the maximum torque with less fuel at high load steady state. Be played.

[Brief description of drawings]

FIG. 1 is a block diagram of the present invention, FIG. 2 is a block diagram of a control system of an embodiment of the present invention, and FIG. 3 is a graph showing an air-fuel ratio and an ignition timing for showing a control method of this embodiment. FIG. 4 is a flow chart showing the overall control of this embodiment, FIG. 5, FIG. 6, FIG. 8 to FIG. 11 (however, FIG. 5 consists of (A) and (B), FIG. 6 is composed of (A) to (C).) Is a flow chart showing the concrete control of the steps shown in FIG. 4, and FIG. 7 is a cylinder pressure waveform chart for explaining the detection of .THETA.pmax. 1 ... knocking sensor, 2 ... signal conversion means, 3 ...
MBT correlation amount detection means, 4 ... Knock by ignition timing-MBT
Control means, 5 ... Ignition timing learning value updating means, 6 ... Air-fuel ratio knock control means, 7 ... Air-fuel ratio learning value updating means, 8 ... Operating condition determining means, 9 ... Exhaust temperature detecting means, 10 …… Control condition determination means, 11 …… Switching means, 21 ……
Engine body, 23A, 23B …… Cylinder pressure sensor, 24A, 24B …… Charge amplifier, 25A, 25B …… Knocking detection circuit, 29 ……
Crank angle sensor, 31 ... air amount sensor, 32 ... air-fuel ratio sensor, 33, 34 ... exhaust temperature sensor, 35 ... control unit.

Claims (1)

[Claims] 1. A sensor for detecting knocking, a means for converting a detection signal of the knocking sensor into a signal according to the knocking strength, a means for detecting an MBT correlation amount, and the knocking strength while maintaining a constant air-fuel ratio. Based on the learning value about the feedback amount and the ignition timing of the ignition timing obtained from each of the deviation between the signal and its target value and the deviation of the detected value of the MBT correlation amount and its target value, the knocking intensity MB of ignition timing within a range that does not exceed the specified value
Knock to approach T-means for performing MBT control, and means for updating the learning value for the ignition timing based on the feedback amount of the ignition timing that minimizes the deviation between the detected value of the MBT correlation amount and its target value. A means for performing knock control based on a feedback value of an air-fuel ratio obtained from a deviation between a signal corresponding to the knocking intensity and a target value thereof while keeping an ignition timing constant, and a learning value related to the air-fuel ratio,
Means for updating the learning value related to the air-fuel ratio based on the feedback amount of the air-fuel ratio that minimizes the deviation between the MBT correlation amount and its target value, and at a high load steady state and at an exhaust temperature under a predetermined control condition or less. A means for determining whether or not there is, and a means for switching the knock-MBT control by the ignition timing and the knock control by the air-fuel ratio at a predetermined timing when it is determined that this control condition is provided A control device for an internal combustion engine.