JPS61106943A - Control device of engine - Google Patents

Abstract

PURPOSE:To improve accuracy of learning control largely by providing a memory control value renewal means reflecting a final control value obtained by feedback control according to the difference of running condition to memory range. CONSTITUTION:A memory means stores control value corresponding to memory range sectioned in the vicinity of each running condition. A basic control value setting means obtains basic control value corresponding to a desired running condition in each running zone. A feedback control means performs feedback control based on a basic control value. A memory control value renewal means reflects a final control value obtained by feedback control to a memory range contributing to calculation of the basic control value among memory ranges according to the difference of running condition. Thus, accuracy of learning control can largely be improved without increasing memory capacity for map.

Description

DETAILED DESCRIPTION OF THE INVENTION (Industrial Application Field) The present invention relates to an engine control device, and mainly relates to an engine control device suitable for controlling the fuel injection amount and ultimately the air-fuel ratio with high precision. .

(Prior Art) Conventionally, fuel-injected gasoline engines have been used to correctly control the air-fuel ratio of the mixed gas supplied to the engine. In addition to detecting the operating state from moment to moment using methods such as A so-called map control method has been proposed that supplies a fuel injection amount corrected by an amount or a correction amount.

As a result of long-term operation, variations in the functions of various parts of the engine occur due to aging changes such as wear of intake valves and cams, resulting in variations in the air-fuel ratio even if the detected operating state is the same. , the oxygen concentration in the exhaust gas is detected by an O2 sensor, and based on this 0□sen signal, it is determined from time to time whether the air-fuel ratio is correctly controlled to the stoichiometric air-fuel ratio. The so-called fuel injection amount is learned in advance by sampling the fuel injection amount or the correction amount of the fuel injection amount at an appropriate timing, updates the map information, and controls the fuel injection amount using the map information updated in this way. It has developed into learning control (Japanese Patent Application Laid-Open Nos. 55-96339 and 1983-
(See Publication No. 54750).

However, in conventional map-based learning control, the engine operating range is divided into a relatively small number of operating zones.
Since one control amount (fuel injection amount and its correction amount) was set in each driving zone, the same control amount was set in each driving zone even if the driving conditions were different. Although the control amount will change discontinuously (stepwise) in the adjacent operation zone, the air-fuel ratio will be
A: There was a problem that it was not possible to increase the accuracy beyond a certain level.

Moreover, even when updating the control variables stored in the map through learning, the sampling is not always performed at a representative operating state within each operating zone, and the operating state at the time of the previous update sampling is not necessarily sampled. Since the operating conditions at the time of this update sampling are not necessarily the same, considerable variation in the control amount itself stored in the map cannot be avoided even after repeated learning, and it is difficult to increase the air-fuel ratio beyond a certain level of accuracy. The problem was that it couldn't be done.

On the other hand, it is not impossible to solve the above problems by dividing the engine operating region into smaller areas and making each operating zone itself smaller, but in that case, the storage capacity for storing the map However, the problem remains that it lacks practicality, such as increased information processing time and longer information processing time.

(Object of the Invention) The present invention has been made in view of the above-mentioned problems, and provides a single engine control device that can significantly improve the accuracy of learning control without increasing the storage capacity for maps. The purpose is to provide

(Structure of the Invention) As shown in FIG. 1, an engine control device according to the present invention divides an engine operating region into a plurality of operating zones each divided into at least three operating states, and each of the operating zones is partitioned into a plurality of operating zones. A storage means for storing control values corresponding to a memory area partitioned near an operating state, and a storage means for storing a control value corresponding to an arbitrary operating state within each operating zone in a memory area of an operating state partitioning this operating zone. A basic control value setting means that calculates the basic control value by interpolation from the control value, a feedback control means that performs feedback control based on the basic control value, and a final control value obtained by the feedback control in the memory area for calculating the basic control value. The memory control value updating means contributes to the control value and reflects the difference in the operating state in the memory area.

(Effects of the Invention) In the present invention, instead of storing one control value in correspondence with each of a plurality of operating zones obtained by dividing the engine operating region, the control value is stored in the vicinity of each operating state that partitions the operating zone. One control value is stored in correspondence with each of the memory areas, and the basic control value corresponding to any operating state within each operating zone is interpolated from the control value of the memory area of the operating state that partitions this operating zone. Since it is determined by GE, highly accurate basic control values that correctly correspond to each operating state can be determined and controlled.

Furthermore, the final control value obtained by the feedback control is reflected in the memory area that contributed to the calculation of the basic control value by the memory control update means, depending on the difference in the operating state, so that each memory By repeatedly updating the control value corresponding to the area with highly accurate data, the stored control value (learning value) can be brought closer to a correct value without variation.

As described above, accurate control can be performed over a wide operating range without increasing memory capacity.

(Example) An example in which the present invention is applied to fuel control of a fuel injection type gasoline engine will be described based on the drawings.

As shown in FIG. 2, this fuel control system includes a computer 4 that commands the fuel injection amount to the injector 3 that injects fuel into the intake passage 2 of the air cleaner 1, and sensors that provide various data necessary to the computer 4. Basically configured.

The throttle opening sensor 5 detects the throttle opening of the throttle valve 6 in the intake passage 2, and the engine load is detected from the throttle opening detected by the throttle opening sensor 5.

The rotational speed sensor 7 is installed on the crankshaft of the engine E and detects the engine rotational speed N from the rotational speed of the crankshaft.

The intake air amount sensor 8 is disposed upstream of the intake passage 2 and detects the intake air amount Q8 taken into the engine E.

The o2 sensor 9 is provided upstream of the catalyst 11 in the exhaust passage 10 of the engine E, and detects whether the air-fuel mixture sucked into the combustion chamber 12 of the engine E is lean or rich from the oxygen concentration in the exhaust gas.

The intake air temperature sensor 13 detects the temperature of intake air.

Coolant temperature sensor 14 detects the coolant temperature in engine coolant passage 15 .

As shown in FIG. 3, the computer 4 includes an I10 interface 16 for inputting the outputs of the sensors 5, 7, 8, 9, 13, and 14 or outputting commands to the injector 3, and a central processing unit. (CPU) 17,
It has a basic structure including a read-only memory (ROM) 18 and a random access memory (RAM) 19, which are connected by a data bus 20 and an address bus 21.

The above ROM 18 contains engine speed N and intake air amount Q.
A program for calculating the basic fuel injection amount Tm based on 1 and various programs for various calculations described later are stored.

The following learning value map is preset and stored in the RAM 19.

That is, as shown in FIG. 6, the engine operating range is divided into a plurality of operating seas 7Z, zz depending on the engine speed N=N+, N2, N3, N4 and the throttle opening L=Ll, L2, L3. , Zl, ... Zl3,
As shown in FIG. 7, each grid point PL (N,
Each memory area MZ is set in the vicinity of L) as an area where the engine speed difference is within a predetermined value ΔN and the throttle opening difference is within a predetermined value ΔL, and each memory area MZ is set at each address of the RAM 19. Each memory area M
A learning map is created by storing the learning value ΔTG corresponding to each memory area M Z + at the address corresponding to Z1, and a part of the RAM 19 is used as learning map storage means.

Incidentally, as shown in FIG. 8, the memory area MZ is located at each grid point P.
It may be set as a nearby area within radius r from □.

In addition, each memory area MZ is stored at each address of the RAM 19.
, and each memory area MZ.

A learning flag map is created in which a learned flag or an unlearned flag is stored at an address corresponding to the learned flag, and a part of the RAM 19 is used as learning flag map storage means.

The CPU 7 of the computer 4 executes the main routine shown in the flowcharts of FIGS. 4 and 5 to determine the width of the fuel injection pulse, and outputs the fuel injection pulse to the injector 3 by an interrupt signal during the execution of the main routine.

Next, the fuel control routine shown in the above flowchart starts at the I10 interface 16 by means of a start signal.
and after initializing the necessary data, the following step 81
-312 and repeat this.

In step S1, the engine rotation speed N and throttle opening are determined from the rotation sensor 7, throttle opening sensor 5, intake air amount sensor 8.0□ sensor 9, cooling water temperature sensor 14, and intake air temperature sensor 13, respectively, and the intake air amount Q3 .0□Sensing signal p, cooling water temperature Qw, and intake air temperature Q8 are read through the I10 interface 16.

In step S2, the engine rotation speed N is a set value N preset in the ROM 18. (For example, No = 10
It is determined whether the engine rotation speed N is equal to or higher than the set value N, such as when starting the engine E. When it is less than the set value N, the process shifts to Sl, and the engine speed N becomes the set value N. In the above case, the process moves to S3.

In S3, the basic fuel injection amount T is calculated based on the engine speed N and the intake air amount Qll by a calculation program that has been input in advance to the ROM 18.

In S4, a temperature correction coefficient KT is calculated based on the cooling water temperature Q8 and the intake air temperature Q3 using a calculation program that has been input into the ROM 18 in advance. That is, as a result, the amount of fuel is corrected to increase when the engine is cold or when it is cold.

)□・I In S5, engine speed N
A learned value ΔT6, which is a fuel correction amount for increasing/decreasing the basic fuel injection NTB with respect to the current operating state determined by The following calculation is performed by interpolation according to the calculation program.

That is, as shown in FIG. 9, it is assumed that the current operating state is point P in operating zone Z, and the grid points that partition this operating zone Z8 are P, , P, , Po, and these grid points The memory area corresponding to is MZ.

, MZJ, MZIIXMZ, the learning values set in these memory areas are ΔTG1, ΔT04, ΔTG, respectively.
ffi, ΔTGI, point P is a, c and d
Assume that the position is determined as shown in the figure.

Point P is in the above memory area MZ, , MZJ, MZ and MZl
If it does not fall within any of the lattice points p, ', p, , P, , , P7, the learned value of point P, that is, the learned value corresponding to the current operating state. For example, the following linear interpolation formula is used to reflect the difference in operating conditions up to the point. )Δ
TGLp, - [(ΔT, i, b+ΔTGJ, a) d+(
ΔT, ff1. b+ΔTG, 1. a) c)/
((a + b)
When entering either Z or l, the learned value of the P point, that is, the learned value ΔTG corresponding to the current operating state (+11 is the learned value of the memory area where the P point is located. In other words, each memory Since the area is set in the vicinity of each grid point, the same learning value Δ is used for each memory area.
Using TG1. Even if x, the resulting decrease in accuracy is negligible, and the purpose is to omit useless interpolation calculations and perform more realistic control.

Note that, of course, other simpler interpolation formulas may be used instead of the linear interpolation formula described above.

In S6, the F/B correction value ΔT is calculated based on the 0
F is calculated.

This F/B correction value ΔT is based on the previous air-fuel ratio being lean,
When the current air-fuel ratio is also lean, the F/B correction value ΔT
F is corrected to increase, and on the other hand, when the previous air-fuel ratio was rich and the current air-fuel ratio is also rich, the F/B correction value Δ
TFfJ< was corrected to decrease, and when switching from lean to rich, the previous increase correction was too large, so this time F
It is revised downward based on /B correction value ΔTy, and on the other hand, when switching from rich to lean, the previous decrease correction was too large, so it is increased based on the current F/B correction value ΔT. Fixed.

In S7, the routine of the flowchart in FIG. 5 (step S70) is executed only when a predetermined learning value update condition is satisfied.
~579), the learning value ΔT of the learning value map is updated.

To explain the update of the learning value ΔTG, in S70, the necessary number of F/B correction values ΔTF are stored in the memory of the RAM 19, and in 371, whether the update condition of T6 is satisfied with the learning value inputted in advance in the ROM 18 or not. It is determined whether or not. When the update conditions are met, the process moves to S72, and when the update conditions are not met, the learning value ΔT is
, and moves to S8 without updating.

The above update conditions include that feedback control is being performed based on the 0□sense signal p, that the engine speed N is within a predetermined range, and that the engine E is warmed up and the cooling water temperature Q. At least the necessary conditions are that the temperature is 60° C. or higher, and that the output of the O2 sensor 9 is a predetermined level or higher.

In S72, the necessary number of F/B correction values ΔT stored above is
The average value of F is calculated, and this average value is used as the new learning value T
, is set as .

In S73, it is determined whether the current operating state is within the memory area MZ of any grid point P of the learned value map, and if it is within the memory area MZ, the memory area MZ is changed in 374. The learning value ΔTG is the current learning value Δ
The learning flag is updated in the TG, and a learned flag is set in the memory area MZ of the learning flag map stored in the RAM 19 in S75.

On the other hand, when it is determined in S73 that the current driving state does not fall within the memory area MZ of any grid point P1, the process moves to 376, and in 376, the driving zone Z to which the current driving state belongs, 4 surrounding areas dividing
There is an unlearned memory area MZ in the memory area MZ1 of the grid point P.
, and if the unlearned memory area MZj exists, the current learned value ΔT6 is tentatively input as the learned value ΔTG of the unlearned memory area MZj at 377, and 378 In this step, a learned flag is set in the corresponding memory area MZ of the learning flag map.

Furthermore, as a result of the determination in S76, the surrounding four grid points P,
If there is no unlearned memory area MZ in the memory area MZ, then the process moves to 379 and uses the current learned value ΔTc to extrapolate the memory area MZ of the four lattice points P8 around the current operating state. The learning value ΔT is calculated and updated, for example, as follows.

To explain again using FIG. 9, the current operating state is P.
The memory area MZ of each grid point P0, PJ, P,
, MZ, SMZllll, MZIl corresponding learning values ΔT'ct, ΔT G j %Δ'rt, ffi, Δ
T6. .

When updating, from point P to each grid point P, , p, , P,
The learning value ΔT is now recessed according to the difference in operating conditions up to ,P.
In order to reflect G, for example, the memory area M of the lattice point P□
For the learning value ΔT'ct of Z□, ×ΔTa, p++Δ
TGi(M-11). Note that ΔTaxN+ on the left side of the above equation is the learning value updated this time, and ΔTGi(N-11 on the right side)
is the previous learning value.

Each memory area [M
Learning values ΔT, jΔ corresponding to Z, , MZ, 11MZ, l
TG ΔTa11 is similarly extrapolated, determined and updated.

However, with the current learning value ΔTG(p) obtained from the current operating state (point P), the surrounding four grid points pi, p
, ,P, ,Pi, for example, the lattice point P! The learning value ΔT6 of the memory area MZ! Of course, it is also possible to update only the above information.

In this way, when updating the learning value of the learning value map with the current learning value ΔTc, T6 is set to the new learning value by extrapolation according to the difference in driving conditions, so the dispersion in the learning value is eliminated. As learning is repeated, the accuracy of the learned values improves and the learned value map is completed.

As mentioned above, at 87, the learning value ΔTG of the learning value map
is updated, and the process moves from S7 to 88.

In S8, an enrich correction coefficient KE for increasing the amount of fuel during high load is calculated based on the intake air amount Q or the throttle opening according to a program previously input to the ROM 18.

In S9, RO is calculated based on the time differential value (change rate) of the intake air amount Q or the time differential value (change rate) of the throttle opening.
The acceleration/deceleration correction coefficient is calculated by the program inputted in advance to M18.

This is to increase the amount of fuel during acceleration and to decrease the amount of fuel during deceleration.

In SIO, a lean control correction coefficient KL for performing lean control is calculated when a predetermined condition such as that the learning value ΔT of the learning value map has been updated a predetermined number of times is met.

In 311, the fuel injection amount T is T-(T, +ΔTG+ΔT,)Ka・KE'Ks・KL
It is calculated using the calculation formula.

Note that when temperature correction, enrichment correction, and acceleration/deceleration correction are not performed, A=1.0, KE-1,0, Ks=1, respectively.
．． 0, and when lean control is not performed, KL=1.
It is 0.

In S12, the fuel injection amount T calculated as described above is converted into a pulse having a width corresponding to the fuel injection amount T.

The pulse obtained in the main routine is applied to the injector 3 by an interrupt signal during execution of the main routine, and the fuel of the fuel injection amount T is injected.

According to the above embodiment, T6 is used for the learning value specified and updated in the memory area MZ near the lattice point P that partitions each driving zone Zi (driving state), and 7″J3E
(7) ilkfRMI: Calculated using 0.7, 4, 0.41, and a highly accurate learning value ΔTG that closely corresponds to each operating condition can be obtained, and the air-fuel ratio can be controlled correctly. Not only can the learning value ΔTG
When updating, the updated learning value formula T6 is calculated by extrapolation so that the difference in the operating state from the operating state at the time of sampling to each grid point P1 is reflected, and the updated learning value ΔTc is used for updating. The learning value map can be updated with a highly accurate learning value ΔTG, and each time the learning value ΔT is updated, the learning value map approaches a perfect one.

Furthermore, a memory area MZ is provided near each grid point P1.

is set, and the same learning value ΔTc is used within the memory area MZ, so the learning value ΔT is determined by interpolation.
When calculating G, interpolation calculations that are not very effective in improving the accuracy of the learned value ΔTG can be omitted.

The above embodiment relates to an example in which the air-fuel ratio of the engine is controlled by learning, but the present invention is not limited to learning control of the air-fuel ratio, but is also applicable to learning control of various control amounts in response to the operating state of the engine. Of course, it can be applied to things.

[Brief explanation of the drawing]

The drawings show an embodiment of the present invention, and Fig. 1 is a functional block diagram, Fig. 2 is an overall configuration diagram of the fuel control system, and Fig. 3 is a functional block diagram.
The figure is a schematic configuration diagram of the computer in Figure 2, Figure 4 is a flowchart showing the main fuel control routine, Figure 5 is a flowchart showing the learning value update routine, and Figure 6 is a flowchart showing the main routine of fuel control. An explanatory diagram of dividing the driving region into a plurality of driving zones, FIG. 7 is an explanatory diagram of a memory area set in the vicinity of grid points that partition each driving zone, and FIG. 8 is a diagram of a modified example. The equivalent diagram, Fig. 9, is an explanatory diagram when calculating the learning value corresponding to the operating state of point P by interpolation from the learning values in the memory area of the surrounding four grid points, and at the same time, it is an explanatory diagram when calculating the learning value corresponding to the operating state of point P by interpolation. FIG. 4 is an explanatory diagram when calculating learning values of the memory area of four surrounding lattice points. 3. Injector, 4. Computer, 5. Throttle opening sensor, 7. Rotation speed sensor, 8.
Intake air amount sensor, 9...0□ sensor. Patent applicant Mazda Motor Corporation agent
Toshio Okamura's drawing - (1. No change in γ in -) Procedural amendment (c) March 15, 1985 Patent Application No. 229956, filed in 1985 2, title of invention Engine sill control device 3, amendment Drifting with Hanyu Patent applicant address 3-1 Shinchi, Fuchu-cho, Aki-gun, Hiroshima Name
(313) Mazda Motor Corporation Representative Ken Yamamoto - 4, Agent Address 4-5-5 Nishitenma, Kita-ku, Osaka 530
No. Tokyu Marquis Koden 5, date of amendment order February 26, 1985, dispatch 6, subject of amendment Drawings

Claims (1)

[Claims] (1) Divide the engine operating region into a plurality of operating zones each defined by at least three operating states, and store control values corresponding to memory areas partitioned near each operating state that partition the operating zones. a storage means; a basic control value setting means for determining a basic control value corresponding to an arbitrary operating state within each operating zone by interpolation from control values in a memory area of operating states that partition this operating zone; and the basic control value. feedback control means that performs feedback control based on the feedback control; and memory control that reflects the final control value obtained by the feedback control in a memory area that contributed to the calculation of the basic control value among the memory areas according to the difference in operating conditions. An engine control device characterized by comprising: value updating means.