JPH086669B2 - Ignition timing control device for internal combustion engine - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to an ignition timing control device for an internal combustion engine.

(Prior Art) From the viewpoint of improving fuel efficiency, the ignition advance value is controlled by performing advance or retard correction so that the minimum ignition advance value (so-called MBT) required to generate the maximum shaft torque is obtained ( Various types of devices have been proposed in which the control of the ignition advance value to the MBT is referred to as "MBT control" hereinafter). However, MBT
Since knocking may occur during control, it is possible to detect a value related to knocking (for example, cylinder pressure) to determine whether knocking has occurred, and when knocking occurs, delay a large amount by a predetermined amount (R _KNOCK ). After that, the MBT control is performed again (see Japanese Patent Application Laid-Open No. 58-82074). This is directly related to the MBT and the ignition advance value that starts knocking when the ignition advance value is advanced (this advance value is hereinafter referred to as "knock limit advance value"). This is because the relative positions of these differ depending on the operating conditions.

For example, since knocking is unlikely to occur in the low load operation range, the knock limit advance value is on the advance side of the MBT as shown in FIG. 5, so that MBT control can be performed without causing knock. On the other hand, in the high load operation range, knocking is more likely to occur, so the knock limit advance value will be on the retard side of MBT as shown in Fig. 6, so this time MBT
It becomes impossible to control. Therefore, in this case, control for advancing and retarding the knock limit advance value as the target position (this control is hereinafter referred to as "knock limit control") is performed. That is, when a knock occurs, a predetermined amount (R _KNOCK )
However, the control is performed in such a manner that the retard is corrected in a stepwise manner, and then gradually (every constant value) is advanced until the knock occurs.

(Problems to be Solved by the Invention) By the way, such a device has knock limit control and MBT.
In combining the control, it can be said that the MBT control is adopted instead of the advance angle correction that is performed when the knock is not generated in the knock limit control, and the advance angle correction in the MBT control is performed after the delay angle correction when the knock is generated. Controlled to the target position. Here, a small value is used for the advance speed (advance correction amount per unit cycle) A _MBT1 in the MBT control, so if the ignition advance value greatly _deviates from the target position due to knocking, the advance speed will be small. However, it is supposed to cause another inconvenience.

Explaining this with reference to FIG. 5, it is unlikely that knock will occur as long as the ignition advance value is in the vicinity of the target position in MBT control, but considering that knocking is stochastic development, The knock occurs with a certain probability even during such an operation, and when the knock occurs, the knock is greatly shifted to the retard side. Here, even if the small advancing speed near the target position contributes to the stability of control, the first is to quickly return to the target position when it greatly deviates from the target position. This prolongs the recovery time and causes unnecessary torque loss during that time.

Further, in the high load operation range as shown in FIG. 6, the retard angle correction and the MB in the knock limit control are performed near the knock limit advance value.
Although the advance angle correction in the T control is repeated, the target position (knock limit advance angle value) is restored at a small advance angle speed here as well, so that the ignition advance value becomes a retard angle as a whole. Here, the more the vehicle is retarded from the knock limit advance value, the farther it is from the target position, so that the generated torque is unnecessarily reduced in this case as well. Furthermore, even if this is compared with the case where only the knock limit control is adopted alone, the advance speed adopted in this case is usually faster than the advance speed in the MBT control, and as a result, only the knock limit control is executed independently. The control is less efficient than when it is used.

Therefore, in order to speed up the return to the target position, it seems that the control gain adopted in the MBT control should be increased in advance, but it is not so simple. For example,
In one mode of the MBT control, the crank angle position (Θpmax) where the combustion pressure has the maximum value is a predetermined crank angle (10 to 10) after the compression top dead center.
It is known that the torque generated by the engine becomes maximum when the ignition advance value is set so as to come to the position of 20 °), so feedback control is performed with this position as the target position (Θpmax ^* ). There is. Taking this as an example, Θpm
Even if the ignition advance value is fixed, ax is scattered with a certain distribution, and the distribution also changes greatly depending on the intake air-fuel ratio. Therefore, the feedback control amount (advance speed) is
Θpmax detected over a predetermined period is averaged, and a value based on the deviation between this average value pmax and Θpmax ^* is adopted. For this reason, if the control gain is uniformly increased because the return is delayed when the position is far from the target position, the stability of the control is deteriorated when the position is near the target position.

Therefore, when the ignition advance value is in the vicinity of the target position, the stability of control is achieved even if the return speed is reduced, and when the ignition advance value is far from the target position, the early return to the target position is prioritized. Therefore, when performing advance correction after retarding when knocking occurs, the control gain used to calculate the advance correction amount in MBT control during a predetermined time or a predetermined number of ignitions is the control gain when knocking is not occurring. It is possible to make it larger than the gain.

However, in the case where the retard correction amount at the time of knock occurrence is changed according to the magnitude of the knock level and the retard angle correction amount is controlled to increase as the knock level increases, the progress is made after the retard angle correction at the knock occurrence. There may be a case where the control gain or the time for increasing the control gain or the number of times of ignition at the time of correcting the angle is increased, but the value cannot be maintained if it remains a constant value. For example, when the maximum level of knock occurs, the retard correction amount also becomes the maximum, so the control gain is set to a large constant value and the control of the large value is made to correspond to the maximum retard correction amount (or the maximum knock level). When the time (constant value) for making the control gain a large value so as not to cause the over-advanced angle even in the gain is set in advance, the knock level at the time of knock occurrence is small, and therefore the retard correction amount is also very small. At times, the above-mentioned predetermined time is too long, resulting in an over-advanced angle, which greatly deviates from the target position for MBT control.

Further, when a knock of a small level occurs due to this over-advance, it may be over-advanced again, and it may not be possible to substantially shift to the MBT control.

In order to prevent such a situation, if the control gain is set to a small fixed value, or the time when the control gain is used and the number of ignitions are set to a small value, it is possible to make a large delay correction by the maximum level knock. The return of the ignition advance value to the vicinity of the target position of the MBT control is greatly delayed, and the original purpose of the present application, ie, the early return to the target position, cannot be achieved.

The present invention has been made by paying attention to such a conventional problem, and in the case of controlling so that the retard correction amount at the time of knocking is changed according to the magnitude of the knock level, the target of the ignition advance value is set. It is an object of the present invention to provide a device that accelerates the return to the position but does not cause an excessive advance.

(Means for Solving Problems) In the present invention, as shown in FIG. 1, a parameter relating to either or both of the generated torque and the combustion state of the engine (this parameter is referred to as "output parameter"). For detecting a value, a means 2 for detecting a value related to knocking, a means 3 for determining whether or not knocking occurs from this related value, and a maximum shaft torque is generated when knocking does not occur. The ignition advance value is corrected by the advance correction amount or the retard correction amount calculated by using a predetermined control gain based on the parameter so that the ignition advance value becomes the minimum ignition advance value required for the ignition advance value. First means for controlling 4
And an ignition timing control device for an internal combustion engine, comprising: a second means 5 for retarding the ignition advance value by a predetermined amount according to the magnitude of the knock level when the knock occurs. After the correction, the control gain used to calculate the advance correction amount when performing advance correction is made larger than the control gain when no knock is generated for a predetermined time or a predetermined number of ignitions, and the retard correction amount when knock is generated. Alternatively, the control gain used for calculating the advance angle correction amount when the knock angle is delayed or the advance angle correction is performed is set to a value according to the magnitude of the knock level. A means 8 is provided for making the time for making the control gain larger and the number of ignitions to a value according to the retard correction amount or the knock level when a knock occurs. In addition, 6 is an output means and 7 is an ignition device.

(Operation) When the ignition advance value is in the vicinity of the target position, the return speed is reduced to improve the control stability, and when the ignition advance value is largely separated from the target position, the early return to the target position is prioritized.
When the advance angle is corrected after the knock angle is corrected, the control gain used for calculating the advance angle correction amount for a predetermined time or a predetermined number of ignitions is set to be larger than the control gain when the knock is not generated. , Is used to calculate the retard correction amount in the case of changing the retard correction amount at the time of knock occurrence according to the magnitude of the knock level and controlling the retard correction amount to increase as the knock level increases. There may be a case where the control gain or the time for increasing the control gain or the number of ignitions is increased and the value cannot be maintained if the value remains constant. For example, when the retard correction amount due to knock is extremely small, when performing the advance correction after the retard correction, the control gain adjusted when the retard correction amount due to knock is maximum is used as it is for a fixed time or constant ignition. The ignition timing is over-advanced by correcting the advance angle during the number of times, or by increasing the control gain during the time adjusted when the amount of retardation correction by knocking is maximum or during the number of ignitions Leave the target position.

At this time, in the present invention, when the retard correction amount becomes the maximum due to the occurrence of the maximum level knock, the control gain for the advance correction immediately after that becomes the maximum corresponding to the maximum retard correction amount. Or, the time to increase the control gain or the number of ignitions becomes maximum according to the maximum retardation correction amount, while the retardation correction amount becomes extremely small due to knocking of a small level, according to this, Immediately after that, the control gain at the time of correcting the advance angle becomes smaller, or the time for increasing the control gain or the number of ignitions becomes smaller, so there is no over-advance in any case regardless of the magnitude of the knock level. . As a result, even when controlling the retard correction amount when a knock occurs according to the magnitude of the knock level, the ignition timing is set to the target position (MBT or knock limit advance value) while preventing over-advance. And can be quickly returned.

 An example will be described below.

(Embodiment) FIG. 2 shows a block diagram of an embodiment when the present invention is applied to a 6-cylinder engine, but there is no difference from the conventional example. That is, in this example, the in-cylinder pressure is adopted as the knock-related value, and the crank angle position where the in-cylinder pressure is maximized is adopted as the output parameter. In the former case, the pressure sensor 11 converts the change in the cylinder pressure into a voltage and outputs this signal as a cylinder pressure signal S ₆ . For the latter, the knock detection circuit 12 has a frequency near the knock (6 to 8 kHz).
It is composed of a band pass filter whose pass band is z) and produces and outputs a knock signal S ₇ from the in-cylinder pressure signal S ₆ .

Since the ignition timing control needs to be performed in synchronization with the crank angle, a sensor (crank angle sensor) that detects a reference position of the crank angle (for example, a predetermined position before compression top dead center) and a unit angle (for example, 1 °). ) 13 is provided, and in a multi-cylinder engine, it is necessary to perform cylinder discrimination, and a sensor (cylinder discrimination sensor) 15 for this purpose is provided.

In addition to this, the air amount sensor 14 and throttle valve fully closed switch
Although 16 is also installed, it is not directly related to this invention.

In the control unit 20 to which these signals S ₁ , S ₂ , S ₄ , S ₆ , S ₇ are input, the ignition advance value (a numerical value representing the crank angle before compression top dead center.) Is calculated based on these signals. The ignition advance value is calculated in the CPU 21 and stored in the ignition register of the input / output port 24 so that the spark ignition is performed when the calculated ignition advance value is reached. Then, the crank angle reference position signal S ₁
When the unit angle signal S ₂ is counted from the input, the signal (ignition signal) S ₈ for shutting off the primary current of the ignition device 25 is output when the ignition advance value is matched.

Next, the calculation of the ignition advance value executed in the control unit 20 will be described. The knock signal S ₇ is used to determine whether knock has occurred.
T control is performed, and when the knock occurs, the ignition advance value is retarded stepwise by a predetermined amount. Such a function is sufficient by adding a control routine to be executed by the CPU 21 when the control unit 20 is composed of a microcomputer. FIG. 3 shows this control routine. The drawing is executed every time the reference position signal S ₁ is input.

In the figure, first, the MBT executed when no knock occurs
Explaining from the control, this function is performed in steps 35, 36 and 38. That is, the actual crank angle position Θpmax at which the in-cylinder pressure is maximum is measured from the in-cylinder pressure signal S ₆ and this value Θpmax is measured.
Feedback control is performed so that pmax matches the predetermined target position Θpmax ^* . This is because the target position (crank angle position where the generated torque is maximum) Θpmax ^* is a predetermined crank angle (10 to 20) after the compression top dead center regardless of operating conditions.
It is noted that it is known to be located almost at ()). For example, if only the ignition advance value is changed under the same other conditions, the position of Θpmax comes earlier as the advance angle increases. Therefore, if Θpmax <Θpmax ^* , it means that the advance angle is too _{far and} the _RMBT1 retard correction is at a speed, because it is too retarded to compensate advance at a rate of a _MBT1 in the case of the reverse (step 31,35,36,31,35,38).

Here, R _MBT1 and A _MBT1 are the retardation speed and the advance angle when the knock is not generated, and are calculated from the control gain when the knock is not generated. For example, when showing a routine for calculating a feedback control amount in the proportional integral control in Figure 4, the deviation from the target position ε (= Θpmax-Θpmax ^*) and the control gain (proportional gain K _P1, integral gain K _I1) Toka proportional Minute P (= K _P1
× ε) and the integral I (= K _I1 × Σε, where Σε is the total sum of ε), and from these sums (P + I), the retardation correction amount (delayed velocity) R _MBT1 per unit cycle and the unit The amount of advance correction (advance speed) A _MBT1 per cycle is calculated (51 to ₅₆ , ₅₁ , ₅₂ , ₅₈ to 61). Note that R _MBT1 and A _MBT1 are not biased to either side by returning at the same speed regardless of whether the target position _deviates to the retard side or the advance side, Often equal values. In this example as well, the values of R _MBT1 and A _MBT1 are made equal by making the control gains equal (step
53, 58).

Further, in steps 36 and 38 of FIG. 3, the ignition advance value (Rold) calculated last time is added to the retard correction amount R _MBT1 or the advance correction amount A _MBT1 is subtracted from the ignition advance value (Rne) of this time.
w) means that it is required.

Next, when it is determined that knock has occurred, the retard angle is stepwise corrected by a predetermined amount R _KNOCK (steps 31, 4).
2). In this case, the predetermined amount R _KNOCK is changed according to the magnitude of the knock level, and the value becomes maximum when the maximum level knock occurs, and becomes extremely small when the small level knock occurs. Here, the purpose of adding R _KNOCK is a fail-safe measure.
A relatively large value is adopted for R _KNOCK , and therefore the correction amount R _MBT1 , which is adopted in MBT control when no knock occurs,
Compared with A _MBT1 , R _KNOCK ＞ R _MBT1 and R _KNOCK ＞ A _MBT1
Is.

Next, the characteristic part of the present invention is that the control gain (K _P1 ,
The advance angle correction in MBT control is performed with a control gain (K _P2 , K _I2 ) that is a value that is larger than K _I1 ) and that corresponds to the amount of retard correction (or the magnitude of the knock level) when a knock occurs. This function is achieved by adding new steps 40, 41, 32 to 34 and 37, 39 in the control routine shown in FIG.

First, in order to perform after the delay angle correction when the knock occurs, the control flag (F) is set in the control cycle in which the stepwise delay angle correction is performed due to the knock occurrence (F = 1), and from the next control cycle, this It is preferable to determine whether or not the flag F is set, and to carry out when the flag F is set (steps 31, 40, 31, 37).

Then, when the flag F is set, the advancing speed after the stepwise retard correction is performed by the newly introduced large control gain.
A _MBT2 (A _MBT2 > A _MBT1 ) is calculated (steps 37, 39). For example, in the case of proportional-integral control, when the flag F is set as shown in FIG. 4, by referring to a predetermined table from the retard correction amount R _KNOCK obtained in step 42 of FIG. A control gain (proportional gain K _P2 , integral gain K _I2 ) other than the control gain (K _P1 , K _I1 ) is read out, and A _MBT2 (= K _P2
• Calculate ε + K _I2 · Σε) (steps 57, 62 to 65).

Here, the predetermined table has a characteristic that the control gain increases as the retard correction amount R _KNOCK increases.
Even if the control gain is large, the value of K _P2 and K _I2 is K
It cannot be said that it is larger than the values of _P1 and K _I1 , and the optimum value is determined by matching. As a result, when the knock occurs, the ignition advance value stepwise deviated from the target position by R _KNOCK is pulled back to the target position by the value of A _MBT2 per unit control cycle.

However, the return speed to the target position can be high only when it is far from the target position, and if the return speed is high until it is close to the target position, it will prolong the settling time to the target position. Therefore, it is necessary to limit the period for increasing the control gain to the predetermined time τ. This τ is roughly determined by the value of the retard correction amount R _KNOCK and the advance velocity A _MBT2 when the knock occurs. For example, assuming that knock does not occur during the return and that the advance speed A _MBT2 is a constant value, the ignition advance value retarded by R _KNOCK in steps is returned at the A _MBT2 speed. The minimum time of (R _KNOCK / A _MBT2 ) × T ₀ (where T ₀ is the control cycle) is sufficient to return to the target position, and this time is set as τ. In addition,
To be exact, this τ is also determined by matching.

More specifically, as shown in FIG. 7, if τ is set in accordance with the maximum retardation correction amount and the advance angle speed A _MBT2 at that time, even if the retardation correction amount is small, the same τ Thus, the ignition advance value is restored.

Therefore, in order to increase the control gain only during τ in the routine, a counter that counts the elapsed time from when the knock occurs at the same time as knock determination is started, and the counted counter values C and τ (Nτ ) Is compared (steps 31, 41, 32, 33) and the flag F is cleared after the counter value C matches Nτ (steps 32, 3).
Four). As a result, the MBT control when knock does not occur with the control gain reduced is restored (steps 37, 38, 57 to 6).
1). The counting by this counter may be executed in another time synchronization routine or rotation synchronization routine instead of being executed in the routine shown in FIG. In addition, if a stable area within a predetermined range is set around the target position and Θpmax falls within this range even if the predetermined time τ has not elapsed, advance angle correction with a large control gain is stopped. , MBT control when no knock occurs may be performed.

Next, when the target position of the MBT control is on the retard side with respect to the knock limit advance value (such as the low load operation range), the operation of this embodiment will be explained. Probabilistically even in such an operation range. Since knocking may occur, the ignition advance value at the target position is retarded by a predetermined amount R _KNOCK when knocking occurs (steps 31, 42).

According to this embodiment, since the flag F is set when the knock occurs, the control gain is increased after the retard correction (steps 31, 40, 57, 62), and the control gain (K _P2 ,
According to K _I2 ), the vehicle is returned to the target position at the advance speed of A _MBT2 (= K _P2 · ε + K _I2 · Σε) (steps 37, 39). Here, advance speed of the A _MBT2 is greater than advance speed A _MBT1 in MBT control when knock is generated, the time required for returning is shorter. That is, in this embodiment, the control gain that is larger than the control gain in the MBT control when the knock is not generated is performed only for the predetermined time τ, instead of immediately shifting to the MBT control when the knock is not generated after the stepwise retard correction. (K _P2 , K _I2 ) calculates the advance speed A _MBT2, and the return speed is increased by correcting the advance angle with this large advance speed. As a result, it is possible to prevent the generated torque from decreasing due to the delay in the return to the target position.

On the other hand, according to the conventional example, the MBT control is executed immediately after the retard correction with the small control gain (K _P1 , K _I1 ) when the knock is not generated, so that the advance position of A _MBT1 is used to reach the target position. The speed of this A _MBT1 is relatively small in order to stabilize the control in the vicinity of the target position, so it takes a long time to return to the target position, and the torque generated unnecessarily during this It lowers it. In addition, if the recovery time is long, knocking may occur during this time and it may not even be delayed by R _KNOCK . In this case, it will take a longer time to recover.

Needless to say, the flag F is lowered after the elapse of the time τ for increasing the control gain, and the MBT control is performed with a small control gain when no knock occurs (step 3).
2,34,35,37,38).

Also, when the target position of the MBT control is on the advance side of the knock limit advance value (such as in the high load operation range), according to this embodiment, the stepwise retard correction amount is set to R _KNOCK and the advance speed is set to Control to the knock limit advance value is performed as A _MBT2 (however, A _MBT2 <R _KNOCK ). This is similar to the conventional example in which only so-called knock limit control is independently adopted, and the knock limit advance value is tracked well. That is, in this embodiment, when the knocking is likely to occur,
At least the effect obtained when only the knock limit control is adopted is obtained. However, even during such an operation, in the conventional example, the ignition advance value tends to be retarded due to the delay in the return to the knock limit advance value, and the generated torque is reduced.

The present invention is further directed to a control for changing the retard correction amount when a knock occurs according to the magnitude of the knock level, and the value of the retard correction amount R _KNOCK is maximized by the occurrence of the knock at the maximum level. Control gain (K _P2 , K _I2 ) when the lead angle is corrected immediately after that, the delay angle correction amount R
If it becomes much smaller than the value of _KNOCK , the control gain (K _P2 , K _I2 ) for correcting the advance angle becomes small immediately after that,
It does not become over-advanced. As a result, even when controlling the retard correction amount when a knock occurs according to the magnitude of the knock level, the ignition timing is set to the target position (MBT or knock limit advance value) while preventing over-advance. And can be quickly returned.

Next, in FIG. 3, "1" is added to both the flag F and the counter value C to mean the first cylinder. This is to consider the case where the present invention is applied to a multi-cylinder engine in which a knock sensor is provided for each cylinder and the ignition advance value is controlled independently for each cylinder. One flag (F1 to F6) and six counters are activated. As a result, control is performed independently for each cylinder, and the torque loss prevention effect of the engine as a whole is improved.

In this example, only the control gain on the advance side is increased, but the control gains on both the retard side and the advance side may be increased at the same time. This is because the control gain on the retard side is not used substantially.

Further, in the embodiment described here, the in-cylinder pressure is adopted as the knock-related value and the crank angle position where the in-cylinder pressure is maximized is adopted as the output parameter, but the invention is not limited to these. For example, the ignition advance value is controlled so as to maximize the indicated mean effective pressure, or the ignition advance value is set so as to equalize the generated in-cylinder pressures among the cylinders for the purpose of reducing the roughness of combustion (irregular combustion). There is one that is controlled, and the present invention can be similarly applied to the case where these are adopted. In short, the output parameter has only to be related to the engine generated torque, the combustion state, or both, and the knock-related value is also the same. Also, the predetermined time τ
Alternatively, a predetermined number of times of ignition may be used instead of.

In the embodiment, the case where the control gain in the MBT control is increased according to the retard correction amount (or the magnitude of the knock level) at the time of knock occurrence is explained.
The number of ignitions for increasing the control gain in the T control can be set to a value corresponding to the amount of retard correction when knock occurs or the magnitude of knock level.

(Effects of the Invention) As described above, according to the present invention, it is determined whether or not knocking has occurred, and if no knocking has occurred, a predetermined control gain is set based on the output parameter so that the ignition advance value becomes MBT. The ignition advance value is controlled by correcting the advance angle or the retard angle with the advance angle correction amount or the retard angle correction amount calculated by using, while the ignition advance value is adjusted according to the magnitude of the knock level when a knock occurs. In the ignition timing control device for an internal combustion engine configured to correct the retard angle by a predetermined amount, the control gain used for calculating the advance angle correction amount when performing the advance angle correction after the retard angle correction when the knock occurs is set to a predetermined time. Alternatively, the value is made larger than the control gain when knock is not generated for a predetermined number of times of ignition, and is set to a value according to the retard correction amount or the knock level when the knock is generated, or when the knock is generated. Retardation correction of the control angle used to calculate the advance angle correction amount when performing the advance angle correction is made larger than the control gain when the knock is not generated, and the time to increase the control gain or the number of ignitions when the knock is generated Since the value is set according to the retard correction amount or the knock level magnitude, even when the retard correction amount at the time of knock occurrence is controlled to be changed according to the knock level magnitude, It is possible to quickly return the ignition advance value to the target position while preventing the advance angle.

[Brief description of drawings]

FIG. 1 is a conceptual configuration diagram of the present invention, FIG. 2 is a block configuration diagram of an embodiment of the present invention, and FIGS. 3 and 4 are explanations of operation contents executed in the control unit in FIG. It is a flow chart to do. FIG. 5 and FIG. 6 are generated torque characteristic diagrams when the relative positional relationship between the knock limit advance value and the MBT control target position is different. FIG. 7 is a waveform diagram for explaining the operation of the embodiment. 1 ... Output parameter detection means, 2 ... Knock related value detection means, 3 ... Knock determination means, 4 ... Ignition advance value control means, 5 ... Knock advance value correction means, 6 ... Output Means, 7 ... Ignition device, 8 ... Control gain / time / ignition number varying means, 11 ... Pressure sensor, 12 ... Knock detection circuit, 13 ... Crank angle sensor, 15 ... Cylinder discrimination sensor,
20 …… control unit, 21 …… CPU, 22 …… ROM,
23 …… RAM, 24 …… I / O port, 25 …… Ignition device.

Claims (1)

[Claims] 1. A means for detecting a parameter related to either or both of a generated torque and a combustion state of an engine, a means for detecting a value related to knocking, and whether or not knocking occurs from the related value. A means for determining and a lead angle correction amount calculated by using a predetermined control gain based on the above-mentioned parameter so that the minimum ignition lead angle value required to generate the maximum shaft torque when knocking does not occur or First means for controlling the ignition advance value by performing advance correction or retard correction with a retard correction amount, and retarding the ignition advance value by a predetermined amount according to the magnitude of the knock level when knock occurs In the ignition timing control device for an internal combustion engine, the control gain used to calculate the advance correction amount when performing the retard advance correction and the advance advance correction at the time of knocking is used for a predetermined time. Is greater than the control gain when knock is not generated for a predetermined number of times of ignition and is a value corresponding to the retard correction amount or the knock level magnitude when knock occurs, or the retard correction backward movement when the knock occurs When the angle correction is performed, the control gain used to calculate the advance correction amount is made larger than the control gain when the knock is not generated, and the time for increasing the control gain or the number of ignitions is the retard correction amount when the knock is generated. Alternatively, an ignition timing control device for an internal combustion engine is provided, which is provided with means for setting a value according to the magnitude of the knock level.