JPH076482B2 - Knotting control method for internal combustion engine - Google Patents

Description

The present invention relates to a knocking control method for an internal combustion engine, and particularly to a comparison between a corrected retard amount for performing a retard angle at a relatively high speed depending on the presence or absence of knocking and the presence or absence of knocking. The present invention relates to an improvement in learning control of a method of correcting a basic ignition advance and controlling knocking by performing a retard advance at a relatively slow speed and changing a learning retard amount by a learning control.

The conventional learning control method is based on the ratio Q / N of the intake air amount Q to the engine speed N or the load determined by the negative pressure of the intake pipe to determine the basic ignition advance θ _BASE to the read-only of the microcomputer. It is stored in memory (ROM) in the form of a map, the ignition advance angle θig for actually controlling the igniter is calculated based on the following equation (1), and the knocking control is performed using this ignition advance angle. is there.

θig = θ _BASE − (θ _KG + θ _K ) (1) However, θ _KG is determined by the engine speed and load and is changed by learning control in order to bring the level of knocking to a predetermined level. The learning retard amount, θ _K, is a correction retard amount that retards the ignition timing when knocking occurs and advances the ignition timing when knocking does not occur.

Here, the correction delay angle amount θ _K is obtained as follows.
First, the vibration of the engine is detected using a knocking sensor composed of a microphone, etc., and a predetermined multiple K · b (where K is a proportional constant) of the average value (back ground) b of the engine vibration and the peak value of the engine vibration. a is obtained, and this peak value α and the value of K · b are compared. The peak value a is K ・ b
When it exceeds the value of No., it is judged that knocking has occurred, and as shown in the following formula (2), the correction delay amount θ _K is changed so that the ignition timing is delayed by a predetermined crank angle (eg 0.4 ° CA) per occurrence of knocking. To do.

θ _K ← θ _K + 0.4 ° CA (2) When the peak value a is less than or equal to the value of K · b, it is determined that no knocking has occurred, and the first timer is used to set a predetermined time (for example, It is determined whether 48 msec) has elapsed, and when a predetermined time has elapsed, the correction retard amount θ _K is changed so that the ignition timing advances by a predetermined crank angle (for example, 0.08 ° CA) as shown in the following expression (3). .

θ _K ← θ _K −0.08 ° CA (3) Further, the learning delay amount θ _KG according to the engine condition is calculated as follows. First, as shown in FIG. 1, the random access memory (RAM) of the microcomputer is provided with addresses 0 to 23 for storing the learning retard amount in association with the engine speed N and the load Q / N, and the learning map is prepared. Is created. The engine speed N and the intake air amount Q are taken in and a point (N, Q /
Find the addresses of 4 RAMs surrounding N). Now, as shown in FIG. 2, the RAM addresses surrounding the point indicating the current engine condition are n (n = 0.1, ... 16), n + 1, n + 6, n + 7, and the learning retard amount θ _KG n at the address n. , The learning delay amount θ _{KG (} n _{+1 )} is stored in the address n + 1, the learning delay amount θ _{KG (} n _{+6 )} is stored in the address n + 6, and the learning delay amount θ _{KG (} n ₊₇₎ is stored in the address n + 7. Be present. Then, the difference in engine speed between the addresses is X, the difference in load between the addresses is Y, the difference in engine speed between the address n and the point indicating the current engine condition is x, and the difference between the address n and the current engine is x. Assuming that the load difference between the point indicating the condition is y, the learning retard amount θ _{KG at the} point indicating the current engine condition is obtained by the two-dimensional interpolation method shown in the following equations (4) to (6). To be

The learning control of the learning map is performed as follows. First, a second timer that determines the learning control time according to the current engine condition and a third timer that determines the learning control time regardless of the engine condition are prepared. When it is detected by the second timer that a predetermined time (for example, 48 msec) has elapsed, it is judged whether or not the correction delay angle amount θ _K is changed to exceed the predetermined crank angle (for example, 4 ° CA), and the correction angle is corrected. When the amount θ _K exceeds a predetermined crank angle, four predetermined learning values (for example, 0.04 ° CA) on the learning map surrounding the point indicating the current engine condition described above.
Learning control to add is performed. On the other hand, when it is detected by the third timer that a predetermined time (for example, 16 seconds) has elapsed, a predetermined learning value (for example, 0.01 CA) is obtained from the learning delay amounts of all the addresses on the learning map regardless of the occurrence of knocking.
The learning control is performed so that the ignition timing is advanced by subtracting.

However, although the above learning control has an advantage that the program becomes short, it is not possible to perform the appropriate learning control according to the engine condition, and thus, the large knocking occurs in which the nosking control becomes inaccurate. There was a problem that the engine output decreased.

In order to solve the above problems and perform appropriate learning control under all engine conditions, the distances between the point indicating the current engine condition and the four points surrounding this point are calculated, and weighted according to this distance. As shown in the following equation, learning values are distributed and learning control is performed.

Here, α is a learning value, and γn, γn ₊₁ , γn ₊₆ , and γn ₊₇ are distances on the learning map between a point indicating the current engine condition and four addresses surrounding this point.

However, in such learning control, it is necessary to find the square root of the sum of the square of the engine speed and the square of the load in order to find the distances γn, γn _{+ 1,} etc. Therefore, the program of the microcomputer is long and the control of the computer is long. There is a problem that it is difficult to take in. Further, because of this problem, the calculation time is long, and it is necessary to reduce the frequency of knocking control at high engine speed. For this reason, there are problems that knocking occurs frequently at high engine speeds, and the output is reduced due to excessively retarded ignition timing.

To deal with such a problem, Japanese Patent Laid-Open No. 56-23566 proposes a control method in which the operating states of the engine are divided and learning is performed for each division.

However, according to the control method of the publication, when the knocking is occurring, the retard amount of the ignition timing is increased, and when the knocking is not occurring for a predetermined time, the retard amount is decreased,
Since this learning is to store the retard amount as it is for each operating state, the retard amount corrected for knocking due to a transient state or noise is also learned. That is,
Since knocking due to noise or the like is also learned, there is a problem in that the next time the driving state is the same, the engine may be retarded too much, and the output may be reduced.

The present invention has been made to solve the above problems, and shortens the program, increases the learning value of the address near the current engine condition and reduces the learning value of the address far from the current engine condition, It is an object of the present invention to provide a knocking control method for an internal combustion engine, which is capable of performing appropriate learning control under the above engine conditions and capable of performing stable ignition timing control even against knocking due to noise or the like.

In order to achieve the above object, the present invention provides a correction retard amount for advancing / retarding an ignition timing according to knocking from a basic ignition advance determined according to an engine speed and a load, and a specific ignition retard amount. It is stored for each operating state, and in order to bring the knocking to a predetermined level, it is learned that when the correction delay amount exceeds a preset predetermined range, it is retarded by a predetermined value and when it is less than the predetermined range, it is advanced by a predetermined value. From the learning retard amount changed by the control, the ignition timing obtained by subtracting the sum of the learning retard amount calculated corresponding to the current operating state is ignited to control the knocking of the internal combustion engine. In the knocking control method, the learning delay amount is stored in association with the specific driving state surrounding and adjoining the current driving state, and the stored learning retard amount is among the specific driving states. , Current engine A first step for distributing the predetermined value at a ratio of a difference between a current engine speed and an engine speed in the specific operating state with respect to a difference in engine speed between operating states sandwiching a rotational speed; The predetermined value distributed in the first step as a ratio of the difference between the current load and the load in the specific operation state with respect to the difference in load between the operation states sandwiching the current load. The second step to be distributed, and the learning amount of the driving state close to the current driving state among the specific driving states becomes large, and the learning amount of the driving state far from the current driving state becomes small, 1st and 2nd
The third step of adding or subtracting the predetermined value distributed in the step of, and is learned. As a result, the program becomes shorter, so it can be incorporated into a limited program of the micro computer, and because the calculation time becomes shorter, learning control can be performed in the same way at high engine speed and low engine speed. The unique effect of improved control is obtained.

Further, in the above-mentioned configuration of the present invention, the predetermined value in the outer region including the upper and lower limit values of the engine speed and the upper and lower limit values of the load corresponding to the stored learning retardation amount in the knocking control region is controlled by the knocking control. It is preferable to make it smaller than a predetermined value inside the region. By doing this, the accuracy of learning control in the boundaries of the learning map and in the notking control area outside the learning map is improved, and the opportunities for learning control increase, so it is necessary to change the learning map according to changes in engine conditions. You can respond promptly.

By the way, the basic ignition advance angle θ _BASE, that is, MBT (Minimum S
As shown in Fig. 3, the park Advance for Best Torque changes according to the engine speed as shown by the curve C ₁ , and minute knocking occurs when it is difficult to cause knocking when the air is wet. The ignition timing is as shown by the curve C ₂ , and the minute knocking occurrence ignition timing is as shown by the curve C ₃ when the knocking is likely to occur such as when the air is dry. Therefore, in order to always perform the same knocking control under the condition that the knocking is likely to occur and the condition where the knocking is unlikely to occur, in the above-described configuration of the present invention, the corrected ignition advance angle is in the predetermined range (for example, 2 ° C. A ≦ θ _K ≦ It is preferable to control the learning delay amount so that it becomes 4 ° CA).

Next, an example of an engine to which the present invention is applied is shown in FIG. As shown in the figure, this engine includes an air flow meter 2 as an intake air amount sensor provided on the downstream side of an air cleaner (not shown). The air flow meter 2 is composed of a compensation plate 2A that is rotatably provided in the damping chamber, and a potentiometer 2B that detects the opening degree of the compensation plate 2A. Therefore, the intake air amount Q is detected as the voltage output from the potentiometer 2B. Also,
An intake air temperature sensor 4 that detects the temperature of intake air is provided near the air flow meter 2.

A throttle valve 6 is arranged downstream of the air flow meter 2, and a surge tank 8 is provided downstream of the throttle valve 6.
Is provided. An intake manifold 10 is connected to the surge tank 8, and a fuel injection valve 12 is arranged so as to project into the intake manifold 10. The intake manifold 10 is connected to a combustion chamber 14A of the engine body 14, and the combustion chamber 14A of the engine is connected via an exhaust manifold 16 to a catalytic converter (not shown) filled with a three-way catalyst. Further, the engine body 14 is provided with a knocking sensor 18 configured by a microphone or the like for detecting vibration of the engine. Reference numeral 20 is a spark plug, 22 is an O ₂ sensor for controlling the air-fuel mixture in the vicinity of the stoichiometric air-fuel ratio, and 24 is a cooling water temperature sensor for detecting the engine cooling water temperature.

The spark plug 20 of the engine body 14 is a distributor.
26, the distributor 26 is connected to an igniter 28. The distributor 26 is provided with a cylinder discriminating sensor 30 and an engine rotation angle sensor 32, which are composed of a pickup and a signal rotor fixed to the displacer shaft. The cylinder discrimination sensor 30 is, for example, an electronic control circuit configured by a microcomputer or the like for a cylinder discrimination signal for each crank angle of 720 degrees.
The engine rotation angle sensor 32 outputs a crank angle reference position signal to the electronic control circuit 34 every 30 degrees of the crank angle.

As shown in FIG. 2, the electronic control circuit 34 includes a random access memory (RAM) 36, a read only memory (ROM) 38, a central processing unit (CPU) 40, and a clock (CLOCK) 41. And a first input / output port 42, a second input / output port 44, a first output port 46, and a second output port 48. The RM36, ROM38, CPU40, CLOCK4
1, the first input / output port 42, the second input / output port 44, the first output port 46 and the second output port 48 are connected to the bus 50.
Connected by. The air flow meter 2, the cooling water temperature sensor 24, the intake air temperature sensor 4, and the like are connected to the first input / output port 42 via a buffer (not shown), a multiplexer 54, and an analog-digital (A / D) converter 56. It is connected. This multiplexer 54 and A / D converter 56
Are controlled by signals output from the first input / output port 42. The O ₂ sensor 22 is connected to the second input / output port 44 via a buffer (not shown) and a comparator 62, and the cylinder discrimination sensor 30 and the engine rotation angle sensor 32 are connected via a waveform shaping circuit 64. ing. Also, the second
A knotting sensor 18 is connected to the input / output port 44 via a bandpass filter 60, a peak hold circuit 61, a channel switching circuit 66 and an A / D converter 68. This bandpass filter is connected to a channel switching circuit 66 via an integrating circuit 63. This channel switching circuit
The control signal output from the second input / output port 44 for inputting one of the output of the peak hold circuit 61 and the output of the integration circuit 63 to the A / D converter 68 is input to 66. The peak hold circuit 61 has a second input / output port.
Reset signal is input from 44. Further, the first output port 46 is connected to the igniter 28 via the drive circuit 70, and the second output port 48 is connected to the fuel injection valve 12 via the drive circuit 72.

In the ROM 38 of the electronic control circuit 34, the map of the basic ignition advance angle θ _BASE represented by the engine speed and the intake air amount, the basic fuel injection amount, and the like are stored in advance, and the signal from the air flow meter 2 and the engine speed are stored. The basic ignition advance and the basic fuel injection amount are continuously output by the signal from the angle sensor 32, and the basic ignition advance and the basic fuel injection are performed by various signals including the signals from the cooling water temperature sensor 24 and the intake air temperature sensor 4. The corrected ignition advance and the corrected fuel injection amount are added to the amount, and the igniter 28 and the fuel injection valve 12 are controlled. The air-fuel ratio signal output from the O ₂ sensor 22 is used for air-fuel ratio control that controls the air-fuel ratio of the air-fuel mixture to near the stoichiometric air-fuel ratio. Further, the learning map shown in FIG. 1 is previously stored in the RAM 36 of the electronic control circuit 34.

Next, an implementation order when the present invention is applied to the engine as described above will be described in detail. In the description of the embodiment of the present invention, the fuel injection control, the air-fuel ratio control, the main routine of the ignition timing control, etc. are the same as the conventional ones, and therefore the description thereof will be omitted, and the knocking control related to the present invention will be omitted. Only the routine will be described.

FIG. 6 shows an interrupt routine every 30 ° CA when the present invention is carried out by using a microcomputer. First, at step 81, the engine speed N is obtained from the rotation time based on the signal from the engine rotation angle sensor 32, and the step is performed.
At 82, the number of interrupts after the cylinder discrimination signal is input from the cylinder discrimination sensor 30 is counted and a flag indicating the current crank angle is set. Next, in step 83, it is determined whether or not the flag set in step 82 is a top dead center (TDC) flag. Step if not currently at top dead center
If it is at the top dead center at 88, it is determined at step 84 whether the knock gate is closed or not. If the knock gate is open, close it at step 85; if the knock gate is closed, go to step 85.
At 86, the channel switching circuit 66 is switched so that the engine vibration signal output from the knocking sensor 18 is transferred to the band bus filter 60, the integrating circuit 63, and the channel switching circuit 66.
To the A / D converter 68 to start the A / D conversion of the average value of the engine vibration, that is, the back ground level. Next, at step 87, the closing time of the knock gate is closed.
At t ₁ , that is, the next time when the notch gate is closed is calculated and the time coincidence interrupt A is set.

Next, in step 88, it is judged whether or not the crank angle has reached 90 ° CA BTDC (before top dead center) based on the flag set in step 82. If the crank angle is not 90 ° CA BTDC, proceed to step 91. If the crank angle is 90 ° CA BTDC, proceed to step 89.
The odor correction retardation amount θ _K is updated and the ignition timing is calculated (the details will be described below).
In step 90, the time coincidence interrupt B is set to find the time when the igniter 28 is turned on based on the ignition timing calculated in step 89 and the current time, and the igniter on flag is set. Then, in step 91, it is judged whether or not the crank angle reaches 60 ° CA BTDC, and 60 ° CA
If it is not BTDC, return to the main rating and 60 °
If it is CA BTDC, the igniter off time is calculated in step 92, the time coincidence interrupt B is set, and the igniter on flag set in step 90 is cleared.

Next, the time coincidence interrupt A shown in FIG. 7 will be described.
This interrupt routine is for obtaining the peak value of the engine vibration, and an interrupt is made at the time set in step 87 of FIG. 6, and the peak value held in the peak hold circuit 61 in step 93 is changed to the channel switching circuit. The value is input to the A / D converter 68 via 66 to start the A / D conversion of the peak hold value and return to the main routine.

FIG. 8 shows the routine of the time coincidence interrupt B. When the time set in step 90 and step 92 in FIG. At step 94, it is judged whether the igniter on flag is set, that is, whether this flag is 1 or not. When the flag is set, the igniter is turned on at step 96, and when the flag is set, the igniter is set at step 95. Off,
Return to the main routine.

FIG. 9 shows an A / D conversion completion interrupt routine, which is executed when the background level A / D conversion and the peak hold value A / D conversion are completed. First, in step 97, it is determined whether the knock gate is currently open. When the knock gate is closed, the A / D conversion value converted in step 86 in FIG. 6 is stored in the memory of the RAM 36 to be the back ground level b in step 98, and in step 99 the not gate is opened to return to the main routine. . On the other hand, when the knock gate is open, the A / D converted value converted in step 93 of FIG. 7 is stored in the memory of the RAM 36 to be the peak value a, and the knock gate is closed in step 101 to return to the main routine.

FIG. 10 shows an interrupt routine that is performed every predetermined time (for example, 4 msec) for counting the time when the knocking does not occur and the learning control time. First, in step 102, the count value of the counter TIME1 for determining the time when no knocking occurs is incremented by 1, and in step 103, the count value of the counter TIME2 for determining the time for learning control is incremented by 1. At the next step 104, the count value of the counter TIME1 is
Judge whether it is 12 (48msec) or less. When the count value exceeds 12, the count value of the counter TIME1 is set to 12 in step 105, and the count value becomes 12
In the following cases, it is determined in step 106 whether the count value of the counter TIME2 is 12 or less. Here, when the count value exceeds 12, step 107
At, the count value of the counter TIME2 is set to 12, and the process returns to the main routine. When the count value is 12 or less, the process returns to the main routine.

Next, the detailed routine of step 89 in FIG. 6 will be described with reference to FIG. The crank angle is 90 ° at step 88 in FIG.
When it is judged that the CA BTDC has been reached, it is judged in step 122 whether or not it is in the knocking control region, and in the next step 108, the peak value a stored in step 100 of FIG.
The back ground level b stored in the ninth step 98 is compared with a value K · b obtained by multiplying the back ground level b by a constant K. When the peak value a exceeds the value K · b, it is determined that knocking has occurred, the correction retard angle θ _K is increased by a predetermined angle (for example, 0.4 ° CA) in step 110, and no knocking occurs in step 112. Counter that counts time
Clear the count value of TIME1. On the other hand, when the peak value a is equal to or less than the value K · b, it is determined that the knocking does not occur, and it is determined in step 109 whether the count value of the counter TIME1 is equal to or more than the predetermined value (12). When the value is above the specified value, the step is started because no knocking has occurred for the specified time.
In 111, the correction delay angle θ _K is set to a predetermined angle (for example, 0.08 ° C
A) After decrementing, in step 112 clear the counter TIME1. If the count value is less than the predetermined value in step 109, the process proceeds to step 113. Step 1
In 13, the basic ignition amount θ _K obtained as described above and the learning retard amount θ _KG obtained by the two-dimensional interpolation method from the learning map are used to obtain the basic ignition as shown in the above equation (1). The ignition advance angle θ ig that actually controls the igniter is calculated by correcting the advance angle θ _BASE .

Next, a routine for obtaining the learning retard amount θ _KG corresponding to the current engine condition from the learning map and performing learning control will be described. FIG. 12 shows this routine from the middle of the main routine.

First, in step 124, the knocking control area (Q / N ≧
0.6) or the non-knocking control area (Q / N <0.6). If it is the non-knocking non-control area, proceed to the main routine, and if it is the notking control area, step 114.
In step 4, the RAM addresses of four points surrounding the point indicating the current engine condition determined by the engine speed N and the load Q / N are obtained on the learning map. Next, in step 115, the obtained value is 4
By the two-dimensional interpolation method (the routine of the two-dimensional interpolation method will be described later) based on the data stored in the RAM address of the point, that is, the learning delay amount stored in the RAM of the four points and the address The learning retard amount θ _KG of the point indicating the current engine condition is calculated, and the calculated value is stored in a predetermined location of RAM. In step 116, it is determined whether or not the count value of the counter TIME2 for obtaining the learning control time counted in step 103 of FIG. 10 is a predetermined value (for example, 12) or more. If the count value is less than the predetermined value, the process returns to the main routine, and if the count value is more than the predetermined value, the count value of the counter TIME2 is cleared in step 117, and then step 1
In step 18, the value of the corrected retard amount θ _K updated in steps 110 and 111 of FIG. 11 is determined, and learning control is performed based on the value of the corrected retard amount (the learning control routine will be described later. To).

FIG. 13 shows a detailed routine of the two-dimensional interpolation method in step 115 of FIG. In this two-dimensional interpolation routine, the map shown in FIG. 1 is used as the learning map, and the four RAM addresses indicating the current engine conditions are assigned to the second map.
As shown in the figure, n, n + 1, n + 6, and n + 7. First, in step 130, it is judged whether or not the current load Q / N is less than or equal to the upper limit value of the load on the learning map, that is, 1.2 [/ rev.]. If the load exceeds 1.2 [/ rev.], 1.2 is stored in register n in step 131 and the load is
If it is 1.2 [/ rev.] Or less, the current load speed Q / N is stored in the register n in step 134. In step 135, the current engine speed N is the upper limit value of the engine speed on the learning map. That is, it is determined whether the speed is 6000 [rpm] or less. When the engine speed exceeds 6000 [rpm], 6000 is stored in the register m in step 136, and when the engine speed is less than 6000 [rpm], the current engine speed N of step 137 is stored. Store the value in register m. In step 138, it is judged whether or not the value of the register n is equal to or more than the lower limit value of the load on the learning map, that is, 0.6 [/ rev.], And when the value of the register n is less than 0.6, the register n of the register n is read in step 139. Value 0.6
When the value of the register n is 0.6 or more, the process proceeds to step 140. Then, in step 140, the value of the register m is the lower limit of the engine speed on the learning map, that is, 1000.
If the value of the register m is less than 1000, the value of the register m is set to 1000 in step 141, and if the value of the register m is 1000 or more, the process proceeds to the next step 142. . As a result, when the current engine speed N and the load Q / N are values on the learning map, the values are stored in the registers m and n, respectively, and the current engine speed N and the load Q / N are learned by the learning map. When the upper and lower limits are exceeded, the upper and lower limits are stored in registers m and n, respectively.

Steps 142 to 149 are a routine for selecting four points on the learning map. First, in step 142, the load value at address 0 from the value of register n is 0.6 [/ re
The value obtained by subtracting v.] is stored in the register n. Next, in step 143, the value of the register n is divided by the load scale interval of 0.2 [/ rev.], The integer part of the quotient is stored in the register n, and the remainder of the quotient is stored in the register y. The value of this register y is equal to the value y of the load from address n in FIG. 2 to the point indicating the current engine condition.
In addition, the integer part of the quotient stored in the register n is a column of addresses that are closest to the point indicating the current engine condition and are equal to or less than the point indicating the current engine condition (horizontal arrangement of addresses, for example, addresses 0 to 5). Is the first column). Then, in step 144, the value of the register y is further divided by 0.2 [/ rev.]. Therefore, finally, the value corresponding to y / Y in the equation (6) described above is stored in the register y.

In step 145, a value obtained by subtracting the value 1000 [rpm] of the engine speed at address 0 from the value in register m is stored in register m, as described above. Next, in step 146, the value of the register m is divided by the scale interval of the engine speed, 1000 [rpm], the integer part of the quotient is stored in the register m, and the remainder of the quotient is stored in the register x. The value of the register x is equal to the value x of the engine speed from the address n in FIG. 2 to the point indicating the current engine condition. In addition, the integer part of the quotient stored in the register m is
The row of the address that is closest to the point indicating the current engine condition and is equal to or less than the point indicating the current engine condition (the vertical arrangement of the addresses, for example, the arrangement of addresses 0, 6, 12, 18 is the first row)
Indicates the line number. Then, in step 147, the value of the register x is further divided by 1000 [rpm]. Therefore, finally, the above-mentioned (4) and (5) are stored in the register x.
The value corresponding to x / X of the expression is stored.

Next, in step 148, the value of the register n is multiplied by 6 and stored in the register n, and in the next step 149, the value of the register n and the value of the register m are added and stored in the register n. As a result, the four lower left corner addresses surrounding the current engine condition, that is, the address number of the nth address in FIG. 2 are obtained and stored in the register n.

In step 150, the learning delay amount θ _KG n stored at address n on the learning map is read out and stored in register A, and the learning delay amount θ _{KG (} n stored at address n + 1 is read.
₊₁₎ is read and stored in the register B, the learning delay amount θ _{KG (} n ₊₆₎ stored in the address n + 6 is read and stored in the register C, and the learning delay amount stored in the address n + 7 is read. The quantity θ _{KG (} n ₊₇₎ is read and stored in the register D.
Then, in step 151, the value of register B is subtracted from the value of register A and the value of register x is multiplied, and the value of register A is added to the value and stored in register E. In step 152, the value of register D is subtracted from the value of register C, the value of register x is multiplied, and the value of register C is added to the value and stored in register F. Finally, in step 153, the value of the register F is subtracted from the value of the register E, the value of the register y is multiplied, and the value of the register E is added to the value to learn the point indicating the current engine condition. The amount of retardation is θ _KG .

Next, the learning control in step 118 of FIG. 12 will be described in detail with reference to FIGS. 14 and 15. In step 160, it is determined whether or not the corrected retard amount θ _K is equal to or larger than the first predetermined crank angle (for example, 2 ° CA). When it is determined in step 160 that the corrected retard amount θ _K is at the first predetermined crank angle end, in step 161, the learning value α is set to the first predetermined value (0 or a positive or negative value, for example, -0.12 ° CA). ) And proceed to step 164. On the other hand, if it is determined in step 160 that the corrected retard amount is greater than or equal to the first predetermined crank angle, in step 162, the corrected retard amount θ _K is equal to or larger than the first predetermined crank angle. It is determined whether the crank angle is not less than a predetermined crank angle (for example, 4 ° CA). Corrected retard angle θ _K is 4
If it is equal to or more than ° CA, the following routine is executed. If the corrected retard angle θK is less than 4 ° CK, the learning delay α is set to a second predetermined value (for example, 0.12 ° CA) in step 163, and the processing proceeds to step 164. move on.

Steps 164 to 167 show a learning control routine, and the steps 144 and 1 shown in FIG.
At 47, the learning values α are temporarily stored in the register and the learning value α is distributed as shown in the following equations (11) to (14) to perform learning control.

θ _KG n ← θ _KG n + α (1-x) (1-y) …… (11) θ _{KG (} n ₊₁₎ ← θ _{KG (} n ₊₁₎ + α x (1-y) …… (12) θ _{KG (} n ₊₆₎ ← θ _{KG (} n _{+6) +} α (1-x) y …… (13) θ _{KG (} n ₊₇₎ ← θ _{KG (} n ₊₇₎ + α xy …… (14) Learning value A detailed routine for distributing α will be described with reference to FIG. In step 170, the learning value α is multiplied by x to obtain α ″, and in step 171, α ″ obtained in step 170 is subtracted from the learning value α to obtain the value α ″.
And In step 172, α obtained in step 171
Is multiplied by y and temporarily stored in the register α '. The value of the register α'is expressed by the learning value α, α (1-
x) y, which is the learning amount at address n + 6. At the next step 173, the value of the register α'is subtracted from α obtained at step 171, and the result is stored in the register α. The value of the register α is α (1-x) (1-y), which is the learning amount at the address n. Also, in step 174, step 1
The value α ″ obtained in 70 is multiplied by y and stored in the register α. The value of this register α becomes αxy, which is the learning amount at the address n + 7. Then, in step 175, from α ″ obtained in step 170 to the register The value of α is subtracted and stored in the register α ″. The value of this register α ″ is αx
(1-y), which is the learning amount at address n + 1.

As a result, the learning amount at the address near the point indicating the current engine condition becomes large, and the learning amount at the address far from the point indicating the engine condition becomes small, and learning control is performed, as shown in the following table. Learning control is performed so that the corrected retard amount becomes a value within a predetermined range according to the magnitude of the corrected retard amount.

Further, FIG. 16 shows changes in the corrected retard amount θ _K , the learned retard amount θ _KG , and the ignition timing θig over time. As can be seen, the retard amount theta _KG learning when the correction retard amount theta _K value of the predetermined range is constant, the retard amount learning when the delay correction amount theta _K exceeds the predetermined range theta _KG increases, and decreases when the corrected retard angle θ _K is less than the predetermined range.

Furthermore, FIG. 17 shows the variation of the ignition timing corresponding to the engine speed. In FIG. 16, the curves C _{1 to} C ₃ are the same as those in FIG. 3, and the correction delay angle θ _K is always constant regardless of whether knocking is likely to occur or not. To be understood.

Next, another embodiment of the learning control of the present invention will be described in detail with reference to FIG. In FIG. 18, the same parts as those in FIG. 14 are designated by the same reference numerals and the description thereof will be omitted. In the present embodiment, the learning control is performed by making the learning amount different between the knocking control region inside the learning map and the outer knocking control region including the boundary of the learning map.

In step 180, it is determined whether the current engine speed N is larger than the lower limit value 1000 [rpm] of the learning map, and in step 181 the current engine speed N is reached.
Is smaller than 6000 [rpm] which is the upper limit of the learning map. Further, in step 182, it is judged whether the current load Q / N is larger than the lower limit value of 0.6 [/ rev.] Of the learning map, and in step 183 the current load Q / N is judged.
Judge whether / N is smaller than 1.2 [/ rev.] Which is the upper limit of learning map. If the engine speed is less than or equal to the lower limit value of the learning map or is greater than or equal to the upper limit value of the learning map, the learning value α is reduced to 1/2 at step 184, and step 164
From step 167, the learning value is distributed and learning control is performed. If the load is equal to or more than the lower limit value of the learning map or the upper limit value of the learning map, in step 185 the learning value α
Is halved, and the learning value is distributed in steps 164 to 167 to perform learning control. Further, when the current engine speed and load are within the range of the upper and lower limit values, the learning value α is left as it is and the step 164 is performed.
From step 167, the learning value is distributed and learning control is performed.

According to this embodiment, since the learning value is changed according to the region, it is possible to similarly perform learning control inside or outside the learning map, and the accuracy of learning control outside the learning map is improved. Further, since the opportunities for learning control increase, it is possible to obtain an effect that it is possible to promptly deal with the case where the learning map needs to change due to changes in various conditions.

As described above, according to the present invention, the learning delay amount is not changed until the correction delay amount exceeds the predetermined range, so that a sudden knock in a transient state or a delay angle to noise output is learned. This is avoided, stable learning control becomes possible, and optimal learning control can be realized in all engine operating states by weighting the learning.

[Brief description of drawings]

FIG. 1 is an explanatory diagram showing a learning map, FIG. 2 is an explanatory diagram showing a point indicating a current engine condition and four points surrounding this point, and FIG. 3 is a line showing a relationship between an engine speed and an ignition timing. Figure,
4 is a schematic diagram showing an engine to which the present invention is applied, FIG. 5 is a block diagram showing the electronic control circuit of FIG. 4, FIG. 6 is a flow chart of an interrupt routine every 30 ° CA, and FIG. 7 is Flow chart of time coincidence interrupt A, FIG. 8 is a flow chart of time coincidence interrupt B, FIG. 9 is a flow chart of A / D completion interruption routine, FIG. 10 is a flow chart showing interruption routine every 4 msec, and FIG. FIG. 12 is a flow chart of the learning control routine, FIG. 13 is a flow chart of the two-dimensional interpolation routine, and FIG. 14 is a flow chart of the learning control routine of the embodiment of the present invention shown in FIG. FIG. 15 is a flow chart of a routine for distributing learning values, FIG. 16 is a diagram showing changes in the correction retard amount, the learning retard amount, and the ignition timing over time, and FIG. 17 is the same engine as in FIG. A diagram showing the relationship between the rotational speed and the ignition timing, and the relationship between the correction retard amount and the learning retard amount. FIG. 18 is a flow diagram of the learning control routine of another embodiment of the present invention of Figure 12. 2 ... Air flow meter, 12 ... Fuel injection valve, 18 ... Notting sensor, 32 ... Engine rotation angle sensor, 34 ...
Electronic control circuit.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Toshio Suematsu 1 Toyota Town, Toyota City, Aichi Prefecture Toyota Automobile Co., Ltd. (56) Reference JP-A-56-23566 (JP, A) JP-A-56-106066 (JP, A) JP-A-57-105530 (JP, A)

Claims (1)

[Claims] 1. A correction retard amount for advancing and retarding an ignition timing according to knocking from a basic ignition advance angle determined according to an engine speed and a load, and stored for each specific operating state, In order to bring the knocking to a predetermined level, the learning delay is changed by the learning control so that when the correction delay amount exceeds a preset predetermined range, it advances by a predetermined value and when it falls below a predetermined range, it advances by a predetermined value. In the knocking control method of the internal combustion engine, which controls ignition by igniting at an ignition timing obtained by subtracting the sum of the learning retardation amount calculated corresponding to the current operating state from the angle amount, the learning The retard amount is stored in association with the specific operating condition that surrounds and is adjacent to the current operating condition, and the stored learning retard amount is the current engine speed among the specific operating conditions. Between operating states A first step of distributing the predetermined value at a ratio of a difference between a current engine speed and an engine speed in the specific operating state with respect to a difference in engine speed; A second step of further distributing the predetermined value distributed in the first step at a ratio of the difference between the current load and the load in the specific operation state with respect to the load difference between the operation states across the load; Of the specific driving states, the learning amount of the driving state close to the current driving state becomes large, and the learning amount of the driving state fast from the current driving state becomes small.
A third step of adding and subtracting the predetermined value distributed in the step of, and a knocking control method for an internal combustion engine, characterized by being learned by: