JPH0627517B2 - Engine controller - Google Patents

Description

TECHNICAL FIELD The present invention relates to an engine control device using a technique called learning control.

(Prior Art) Conventionally, in order to perform control for maintaining an appropriate operating state of an engine with good responsiveness even when the operating state changes, a control amount corresponding to the operating state of the engine is used. It is known to store and update the stored value (learning value) in accordance with the above, and perform learning control for reflecting the stored value (learning value) in the control. For example, Japanese Patent Laid-Open No. 56-23566 discloses an ignition timing. A device adapted for learning control is shown. That is, this device checks the knocking occurrence state as the engine operating state, and classifies the retard amount (retard angle amount) of the ignition timing obtained accordingly into the operating state of the engine (for example, engine speed and intake negative pressure). The stored value is stored in a corresponding storage area of the storage means having a plurality of storage areas, and the stored value of the retard amount is updated to be a learning value, and the ignition timing is set based on the learning value stored in the storage means. I'm trying to control.

By the way, in the conventional device of this type, in order to increase learning opportunities for each storage area, the storage area of the storage means is set corresponding to a relatively large division of the operating state. In some cases, a considerably large error may occur between the learned value and the proper value. That is, even if the division of the storage area is increased to some extent, the learning value can be sufficiently adapted to various operating states in the area as long as it is obtained in the average operating state in the area. If the learning value is obtained in an operating state that is biased to one side (for example, low rotation and low load side), the other side (for example, high rotation and high load side)
The error between the learning value and the proper value becomes large in the operating state biased to.

In order to reduce such an error, it is conceivable to previously provide the storage means with a large number of storage areas into which the operating state is divided into small sections. However, if the storage areas are simply divided in advance, learning opportunities for each storage area (opportunities to which the driving state belongs) are reduced. Therefore, it is necessary to sufficiently store and correct the learning value in each storage area. Becomes difficult.

(Object of the Invention) In view of such circumstances, the present invention focuses on the fact that the difference between the learning values of the adjacent storage areas becomes larger than the predicted value when the learning value is not appropriate. (EN) An engine control device capable of obtaining an appropriate learning value and performing highly accurate learning control over the entire operating range.

(Structure of the Invention) The structure of the present invention will be described with reference to the functional block diagram shown in FIG. 1. An operating state for detecting an operating state of an engine (presence or absence of knocking influenced by external factors or a state such as an air-fuel ratio). A detection unit 1; a storage unit 2 for storing the control amount of the engine for each storage region obtained by dividing the map on which the operating state of the engine (engine rotational speed and load state determined mainly by the driving operation) is divided into a plurality of regions; According to the operating state of the engine detected by the operating state detecting means 1, the control amount stored in the storage area of the storing means 2 corresponding to the operating state at that time is corrected and this control amount is used as the learning value. The learning control means 3 and the learning value stored in the storage means 2 are used to control the control target (for example, control based on the learning value or feedback reflecting the learning value). In the control device of the engine for performing the control), the following comparing means 4 and storage area dividing means 5 are provided. That is, the comparison means 4 compares the new learning value X obtained by the learning control means 2 and the learning value Xnear of the control amount of the storage area adjacent to the storage area in which the new learning value is to be stored. The difference between the two (|
X-Xnear |), the storage area dividing means 5 subdivides the storage area in which the new learning value X is to be stored and increases the storage area when the difference is a predetermined value or more. .

That is, normally, the storage areas are set in correspondence with a somewhat large division of the operating state so as to increase the learning opportunities for each storage area. However, when the learning values of the adjacent storage areas are significantly different, the learning value is appropriate. Therefore, the learning value is optimized by subdividing the corresponding storage area.

(Embodiment) FIGS. 2 and 3 show an embodiment of the device of the present invention. In this embodiment, knocking is detected as an operating state of the engine, and the retard amount of ignition timing is controlled as the control amount of the engine. Is shown. FIG. 2 shows the overall structure of the apparatus of this embodiment. In this figure, 11 is an engine cylinder, 12 is a combustion chamber in the cylinder 11, 13 is an intake passage, and 14 is an exhaust passage. The throttle valve 15 and the fuel injection valve 16 are provided. Further, an ignition plug 17 is provided in the combustion chamber 12, and an ignition coil 18, a distributor 19 and an igniter 2 are provided for the ignition plug 17.
0 is arranged to constitute a known ignition device, and the ignition coil 18 is connected to a battery (BAT).

Further, 21 is a control unit (EC
U), which is a crank angle sensor 22 that detects a predetermined crank angle, and a knock sensor 23 that is mounted on the wall surface of the cylinder 11 and that detects vibration due to engine knocking.
And a negative pressure sensor 24 for detecting an intake negative pressure downstream of the throttle valve 15 corresponding to the engine load, and outputs a control signal to the igniter 20.

The control unit 21 includes a CPU as shown in FIG.
25, a ROM 26 for storing a program and the like, a RAM 27 for storing a learned value of an ignition timing retard amount described later and other various data, an input circuit 28 for a signal from the crank angle sensor 22, and a knock sensor 23. An input circuit 29 for processing a signal from the negative pressure sensor 24, a multiplexer 30, and an A / D converter 3
1, a timer 32 and a drive circuit 33 for outputting a control signal to the igniter 20, and a reference clock counter 34. Reference numeral 35 is a bus in the control unit 21.

The RAM 27 has a basic learning value map M ₁ as schematically shown in FIG. 4 (a) as a storage means for storing the learning value of the ignition timing retard amount in a plurality of storage areas, and the storage area division For this purpose, a divided learning value map M ₂ as shown in FIG. The basic learning value map M ₁ is ignited in a storage area for each of a plurality of learning zones Z ₁ , Z ₂ ... Zn ... Which is a relatively large partition of the map showing the operating state according to the intake negative pressure and the engine speed. The learning value θl of the time retard amount can be stored. Further, the divided learning value map M ₂ has a storage area for each zone obtained by subdividing the learning zone into a plurality of pieces (for example, four divisions), and for example, when a division process described later is performed on a certain learning zone Zn. 4 so that the respective may store learned value θl the divided storage areas of each learning zone Zn ₁ ~Zn ₄ was. RAM2
Although not shown in FIG. 7, a map storing a learning count CNT described later for each learning zone Z ₁ , Z ₂ ... Zn ... Corresponding to the basic learning value map M ₁ , a divided learning zone (Z
_{1 to} Z ₄ ), a map that stores the number of learning times CNT, a division flag map that stores the division flag data Fdiv described below that indicates the presence or absence of division for each learning zone (Z ₁ , Z ₂ ... Zn ...), and the like. Has been.

Then, the CPU 25 controls the learning control means according to the present invention by performing control according to a flowchart described later.
The comparing means and the dividing means are configured, and the ignition timing is controlled according to the learned value of the ignition timing retard amount according to the operating state and the knocking occurrence state.

The control by the CPU 25 will be described with reference to the flowcharts of FIGS. 5 and 6 (a) to (d).

FIG. 5 shows a Mayroutine. First, in step R1, the initial values of each data of θk, θsum, θkn, Nkn, θl area, CNT area, and Fdiv area, which will be described later, are set to 0, and then the engine rotation is performed in step R2. Then, the basic ignition timing θb is calculated based on the intake load, and the process of step R2 is repeated.

FIGS. 6A to 6D show a series of interrupt routines,
This interruption is performed at every constant crank angle (for example, BTDC 60 °). When the interruption is started in FIG. 6 (a), first in steps S1 and S2, the knock data (KN) and the intake negative pressure data (BST) based on the signals from the sensors 23 and 24 are input, and the engine according to the interrupt cycle. The rotation speed (rpm) is calculated. And according to the operation state checked in with the intake negative pressure data and the engine rotational speed, is stored it examines the current learning zone by division of the basic learned value map M ₁ in step S3 in Znew register further step S4~ in S6, if the split flag data Fdiv corresponding to the current learning zone Znew is 1, it has been amended stored contents of Znew register examines the current learning zone by segment division learning value map M _2.

Next, in step S7, it is checked whether or not the driving state is in the same learning zone (Znew = Zold) between the present time and the previous time, and if the determination result is YES, the normal knock control in step S66 and later described later is performed. If the determination result is NO, the process proceeds to step S8.

Steps S8 to S14 are processes for correcting the learning value, and when the operation state shifts to a learning zone where
The learning value θl of the learning zone Zold before the transition is corrected. That is, based on the knock number Nkn that has occurred in the same learning zone, which has been obtained in steps S69 and S70 described below up to the previous time, and the cumulative value θsum of the ignition timing retard at the time of knock occurrence, if the knock number Nkn is not 0, the average value is obtained. The ignition timing retard amount (θave = θsum / Nkn) is calculated (steps S8 and S9), and a new learning value θl is set to [θl = θl · (1-K) based on this and the old learning value θl and the learning efficiency K. +
K · θave] (step S10). Then, the new learning value θl is saved in the storage area corresponding to the target zone Zold of the map M ₁ or M ₂ (step S11). Further, data CN indicating the number of learning value updates
If T1 has not reached the predetermined value, it is incremented (steps S12 and S13), the data CNT1 is saved in the area corresponding to the target zone Zold of the map of the learning number storage value (step S14), and then the 6
The process moves to step S15 in FIG. Note that step S8
When the result of the determination in is NO, the process directly proceeds to step S15.

Steps S15 to S31 in FIG. 6B are processes for dividing the storage area when the division condition is satisfied for the target zone Zold. In this process, when the number of times of learning is small or the learning value is appropriate, there is no division in order to increase the chances of learning, and even if learning is sufficiently performed, there is a significant difference in learning value between the adjacent zones. When the size is large, the process is divided into pieces.

That is, as the processing for checking the division condition, first, the division flag data Fdiv of the target zone Zold is 0 (undivided) or 1
(Divided), and if not divided, it is checked whether the learning count data CNT1 has reached a predetermined value (step S15).
~ S17), if the division is completed or the number of times of learning is less than the predetermined value, the division processing is not performed and the process proceeds to step S32 described later. If the number of learnings has reached a predetermined value without division,
After searching the adjacent zone (step S18) and setting the adjacent zone number Nnear as the initial value of the counter value C ₀ (step S19), the learning count CNT of the adjacent zone number Znear is read from the map in order from the first adjacent zone. Then, it is checked whether or not the data Cnear has reached a predetermined value (steps S21 and S22), and if the determination result is YES, the learning value θl of the target zone and the learning value θnear of the number of adjacent zones Znear read from the map are determined. It is checked whether or not the absolute value of the difference is greater than or equal to a predetermined value (steps S23 and S24). Step until these determination results is NO number adjacent zones undetermined _(value of _C 0 -1 was replaced with _{C 0)} becomes 0 S25 to
The determinations of steps S21 to S24 are repeated through S27,
When the conditions determined in steps S21 to S24 are not satisfied for all the adjacent zones, the determination result in step S26 is YES, and in this case, the division process is not performed and the process proceeds to step S23 described later.

Target zone Zold and at least one adjacent zone Znear
If the conditions determined by the determination in steps S21 to S24 are satisfied between and, division processing is performed. This process is the division flag data Fdiv corresponding to the target zone Zold as 1 from (step S28), at step S10 in the all storage areas of the four zones fraction divided learning value map M ₂ corresponding to the target zone Zold The learned value θl thus obtained is saved (step S29), and the target zone Zold is set to the third value, for example.
In the case of the zone Zn in the figure, θl is saved in each storage area of the zones Zn ₁ to Zn ₄ . Further target zone Z
The division flag data Fdiv is saved in the storage area of the division flag map corresponding to old, and the target zone Zo
All the stored values CNT of the learning numbers of the four zones corresponding to ld are cleared (steps S30 and S31). Even when this division processing is performed, step S in FIG. 6 (c).
32, but in this case, the learning count has been cleared and the integration condition described later does not hold, so that the process actually moves to FIG. 6 (d).

In steps S32 to S45 in FIG. 6C, it is checked whether or not a predetermined integration condition is satisfied, and step S46.
In S59, the learning zones that have been divided when the integration condition is satisfied are integrated.

That is, as the process for checking the integration condition, first, it is checked whether the division flag data Fdiv is 1 (divided),
If the zone has been divided, the integration planned zone for the target zone is searched (step S33). For example, if the zone Zn ₁ according to the divided learning value map M ₂ is the target zone, zones Zn ₂ to Zn ₄ are searched. And the postscript Nuni, Ds
The initial value of each data of um and Zuni is set in each register (steps S34 to S36). Then, the number of learnings CN from the map for the planned integration zone zuni
T is read and it is determined whether or not the data Cuni has reached a predetermined value (steps S37, S38). If the determination result is YES, the learning value θl of the target zone and the learning value θuni of the integrated planned zone Zuni are determined. It is determined whether or not the absolute value of the difference is less than or equal to a predetermined value (steps S39 and S40), and if this determination result is YES, the cumulative value Dsum of the absolute values of the differences is obtained (step S41), and Number of zones (Nuni-
Step S until the value obtained by replacing 1 with Nuni) becomes 0
Processing for repeating the processing from 37 (step S42-
Perform S44). Further, if the cumulative value Dsum is obtained for all the integrated planned zones, it is checked whether or not this is within a predetermined value (step S45).

Through these steps S32 to S45, when the target zone has been divided and all the zones to be integrated have been sufficiently learned, and the absolute value and the cumulative value of the differences between the learning values are relatively small, the integration processing is performed. When the integration condition is not satisfied, the integration process is not performed and the process proceeds to step S60 of FIG. 6 (d). As the integration processing, the cumulative value Esum is obtained by adding the learning values read from the divided learning value map M _{2 in} order from the first integrated scheduled zone to the learning value θl of the target zone (steps S46 to S46).
53), and the average value (Esum / 4) obtained from this cumulative value is set as the integrated learning value θl (step S54). Then, the corresponding learning zone before division is searched, and this is set as the target zone Zold, and the learning value θl is saved in the storage area of the basic learning value map M ₁ corresponding to this target zone Zold (steps S55 and S56). This target zone Z
For old, the learning count storage value CNT is cleared and the division flag data Fdiv is set to 0 and saved in the storage area of the division flag map (steps S57 to S5).
9). Then, the process proceeds to step S60 in FIG. 6 (d).

Steps S60 to S74 of FIG. 6 (d) are processes for controlling the ignition timing, of which steps S60 to S65.
Is a preliminary process performed when the driving state shifts to another learning zone. In this process, the data of the current learning zone Znew is set in the Zold register (step S60) and the current learning zone Znew is set.
The learning value θl is read from the storage area of the map M ₁ or M ₂ corresponding to the above and set in the θk register (step S
61, S62) The learning count CN read from the map
T is set in the CNT1 register and the cumulative value θsu described later is set.
m and knock counter Nkn are set to 0 (step S6)
3 to S65).

Next, in steps S66 to S74, the ignition timing is controlled according to the knocking occurrence state. That is, the presence or absence of knock is checked (step S66), and when the knock occurs, the retard increase value θkn calculated as the magnitude function f (KN) is calculated.
Only by increasing the ignition timing retard amount θk (step S
67, 68), and further, the cumulative value θsum of the retard amount at the time of knock occurrence is obtained and the knock counter is incremented (steps S69, S70). Further, when the knock is not generated, the retard amount θk is reduced by a fixed minute value θi unless the retard amount θk is 0 (steps S71 and S72). After that, the final ignition timing θ is calculated based on the advance value θb of the basic ignition timing and the retard amount θk (step S73), and this is set in the timer (step S74) before returning to the main routine. The advance value θb of the basic ignition timing is stored in advance in the ROM.
It is stored in the map and can be obtained from this map according to the driving condition.

According to the engine control device as described above, the retard amount of the ignition timing is controlled based on the learning value, that is, in the control example shown in the above flowchart, when the operating state is kept in the same learning zone, the knocking occurrence state is generated. The normal feedback control (steps S66 to S74) according to is performed, but the learning value θl stored in the map M ₁ or M ₂ is used as the initial value of the retard amount immediately after the operation state shifts to a different learning zone. (Steps S60 to S6
5) As a result, the ignition timing is controlled with good responsiveness to changes in the operating state. Then, the learning value θl is sequentially corrected to be suitable.

Particularly, in the device of the present invention, basically, the learning value is stored in the basic learning value map M ₁ for each relatively large learning zone in order to increase the opportunity of learning for each driving zone. Even if the learning is repeated, if an appropriate learning value according to the average driving state of the target zone cannot be obtained,
This is discriminated by comparison with the adjacent zone, and the learning zone is subdivided (steps S16 to S32). Then, after the division, learning is performed for each of the subdivided learning zones and the learning value is stored in the divided learning value map M ₂ , so that the error of the learning value due to the variation of the driving state in the learning zone is small. To be done.

Incidentally, in the above-described embodiment, while the learning value of the ignition timing retard amount is obtained when the operating state shifts to a different learning zone,
Although the learning value read from the storage means is used as the initial value of the retard amount in the same learning zone, the learning value may be obtained and the learning value may be controlled even in the same learning zone.

Further, the control device of the present invention is not limited to the control of the ignition timing shown in the above-described embodiment, but when the control amount of another engine is learned and controlled, for example, the learning of the fuel injection amount that is stopped according to the output of the O ₂ sensor or the like is learned. It can also be adopted when controlling the air-fuel ratio based on the value.

(Effect of the invention) As described above, according to the present invention, the learning value of the control amount according to the operating state of the engine is stored for each storage area in which the map showing the operating state of the engine is divided into a plurality, and this learning is performed. When the control is performed based on the value, the new learning value obtained by the learning control means is compared with the learning value of the control amount of the storage area adjacent to the storage area where the new learning value is to be stored. Then, when the difference between the two becomes larger than a predetermined value, the storage area in which the new learning value should be stored is subdivided, so that the learning value is optimized by repeating learning for each storage area. It is possible to prevent the error caused by the variation of the driving state in the storage area while significantly improving the accuracy of the learning value.

[Brief description of drawings]

FIG. 1 is a functional block diagram showing the configuration of the present invention, FIG. 2 is a schematic diagram of an embodiment of the present invention, FIG. 3 is a block diagram of a control unit, and FIGS. 4 (a) and 4 (b) are FIG. 5 and FIGS. 6 (a) to 6 (d) are explanatory views schematically showing a map for storing learning values, and are control flowcharts. 1 ... Operating state detecting means, 2 ... Storage means, 3 ... Learning control means, 4 ... Comparison means, 5 ... Storage area dividing means.

Claims (1)

[Claims] 1. An operating state detecting means for detecting an operating state of an engine, a storing means for storing a control amount of the engine in each storage area obtained by dividing a map showing an operating state of the engine into a plurality of storage areas, and the operation. Learning control means for correcting the control amount stored in the storage area of the storage means corresponding to the operating state at that time in accordance with the operating state of the engine detected by the state detection means and using the control amount as a learning value; In a control device for an engine, which is configured to control a controlled object using the learning value, a new learning value obtained by the learning control means and a storage area in which the new learning value is to be stored are stored. Comparing means for comparing the learning values of the control amounts of adjacent storage areas with each other, and the new learning value is stored when the difference between the learning values compared by the comparing means is equal to or more than a predetermined value. Engine control apparatus is characterized by providing a storage area dividing means for increasing the number of storage areas of storage plurality subdivided come.