JPH04203337A - Air-fuel ratio controller - Google Patents

Abstract

PURPOSE:To make control parameters such as basic injection time proper by providing the map of air-fuel ratio compensating coefficient based on air flow sensor output voltage to adjust control coefficients independently separately. CONSTITUTION:A control circuit 60 acquires air-fuel ratio compensation coefficient by the feedback control of air-fuel ratio. In the control circuit 60, there is provided the map of air-fuel ratio compensation coefficient based on the output voltage of an air flow sensor 24. It is thus possible to adjust injector coefficient which is a control coefficient, ineffective injection time of an injector 12 and air flow characteristics independently respectively, making control parameters such as basic infection time proper.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention relates to an electronic fuel injection system for internal combustion engines such as gasoline engines, and in particular to an air injection system that is optimal for initial calibration of control constants and for calibrating changes over time. This invention relates to a fuel ratio learning control device.

[Conventional technology]

Conventionally, learning control of a gasoline engine is described in Japanese Patent Laid-Open No. 111034/1983.

[Problem to be solved by the invention]

The above conventional technology writes the air-fuel ratio correction coefficient generated in closed-loop control into a steady-state learning map for engine speed and load, and corrects changes in control constants due to engine differences and aging. Since no correction is performed and correction is made using the air-fuel ratio correction coefficient, an error occurs in the value of the basic injection time, which is the load equivalent value, and the ignition timing control also searches the map based on the basic injection time, which includes the error. ,
There was a problem that proper ignition timing control could not be performed.

An object of the present invention is to independently adjust the injector coefficient, which is a control coefficient, the invalid injection time of the injector, and the air flow rate characteristic, to optimize control variables such as the basic injection time.

[Means to solve the problem]

In order to achieve the above objective, a correction characteristic indicating the appropriateness or inappropriateness of each control coefficient is calculated from the air-fuel ratio correction coefficient formed and learned through closed-loop control, and optimization of each control coefficient is realized. By doing so, control variables such as the basic injection time can be optimized, and appropriate fuel injection and ignition timing can be achieved.

[Operation] The air-fuel ratio correction coefficient reflects the error of the control constant from the appropriate value. For example, when the injector coefficient is larger than the appropriate value, a large amount of fuel is injected, and the air-fuel ratio correction coefficient becomes a small value due to feedback control.

The influence of the injector coefficient contributes uniformly to the air-fuel ratio correction coefficient in each operating region. Furthermore, an error in the air-fuel flow rate characteristics has an effect on the equal air flow rate line where the air flow rate is equal. Therefore, considering the effects of errors in the injector coefficient, injector invalid injection time, and air flow rate characteristics, indicators indicating each error are calculated from the air-fuel ratio correction coefficient, each control coefficient is optimized, and each control variable is adjusted appropriately. become

〔Example〕

Hereinafter, the present invention will be explained with reference to embodiments shown in the drawings.

In order to comprehensively control the operating conditions of automobile gasoline engines, improve exhaust gas conditions, and improve fuel efficiency, a control device using a microcomputer captures signals from various sensors that indicate the engine's operating conditions. 2. Description of the Related Art Electronic engine control devices (hereinafter referred to as EEC) are used, which perform various controls such as fuel supply amount and ignition timing to obtain optimal engine operating conditions.

An example of a system in which EEC is applied to a fuel injection type internal combustion engine is disclosed in Japanese Patent Application Laid-Open No. 55-134721.
This example is shown in Figures 1 and 2.
This will be explained with a diagram.

FIG. 1 is a partial sectional view schematically showing the entire engine control system. In the figure, intake air is supplied to the air cleaner 2. Throttle chamber 4. Pass through intake pipe 6.

It is fed into the cylinder 8. Exhaust pipe 1 from cylinder 8
o and is emitted into the atmosphere.

The throttle chamber 4 is provided with an injector 12 for injecting fuel.
The fuel injected from the throttle chamber 4 is atomized in the air passage of the throttle chamber 4 and mixed with the intake air to form a mixture, and this mixture passes through the intake pipe 6 and enters the intake valve 20 by opening the valve.
It is supplied to the combustion chamber of cylinder 8.

A throttle valve 14 is provided near the outlet of the injector 12. The throttle valve] 4 is configured to be mechanically interlocked with the accelerator pedal, and is driven by the driver.

An air passage 22 is provided upstream of the throttle valve 1°4 of the throttle chamber 4, and a hot-wire air flow meter made of an electric heating element, that is, a flow rate sensor 24, is disposed in this air passage 22, and a flow rate sensor 24 is disposed in the air passage 22. A changing electrical signal AF is taken out. Since the flow rate sensor 24 consisting of this heating element (hot wire) is provided in the bypass air passage 22,
It is protected from high-temperature gas generated during backfire from the cylinder 8, and is also protected from contamination by dust and the like in the intake air. The outlet of the bypass air passage 22 is opened near the narrowest part of the venturi, and the inlet thereof is opened on the upstream side of the venturi.

The injector 12 is constantly supplied with pressurized fuel from the fuel tank 30 via the fuel pump 32, and when an injection signal from the control circuit 60 is given to the injector 12, the fuel enters the intake pipe 6 from the injector 12. Fuel is injected.

The air-fuel mixture sucked in through the intake valve 20 is compressed by the piston 50. A spark from a spark plug (not shown) causes combustion, and this combustion is converted into kinetic energy. The cylinder 8 is cooled by cooling water 54. The temperature of this cooling water is measured by a water temperature sensor 56, and this measured value TV is used as the engine temperature.

An 02 sensor 142 is provided at the collecting part of the exhaust pipe 10, and measures the presence or absence of 02 in the exhaust gas and outputs a measured value λ.

In addition, the crankshaft (not shown) is attached at each reference crank angle and at a certain angle (for example, 0.5
A crank angle sensor is provided that outputs a reference angle signal and a position signal for each angle (degrees).

The output of the crank angle sensor, the output signal TW of the water temperature sensor 56, the output signal of the A/F sensor 142, and the heating element 24
The electrical signal AF from the ignition coil AF enters a control circuit 60 consisting of a microcomputer, memory, etc., and becomes an input for controlling the injector 12 and ignition coil 62.

Further, the throttle chamber 4 is provided with a bypass 26 connected to the intake pipe 6 across the throttle valve 14, and this bypass 26 is provided with a bypass valve 61 whose opening and closing are controlled.

This bypass valve 61 faces the bypass 26 provided by bypassing the throttle valve 14, and is controlled to open and close by a pulse current, and the membrane area of the bypass 26 is changed depending on the lift amount, and this lift amount is determined by the control circuit. The drive unit is driven and controlled by the output of 6. That is, the control circuit 6
0 generates an opening/closing cycle signal for controlling the drive unit, and the drive unit operates the bypass valve 6 based on this opening/closing cycle signal.
Adjust the lift amount in step 1.

The EGR control well 90 controls the passage between the exhaust pipe 10 and the suction pipe 6, and controls the amount of EGR flowing from the exhaust pipe 10 to the suction pipe 6.

Therefore, in the second time, the injector 12 is controlled to control the air-fuel ratio (A/F) and control fuel increase and decrease.
Bypass valve 61 and injector 12 can perform engine speed control (ISC) during idling,
Furthermore, the amount of EGR can also be controlled.

FIG. 2 is an overall configuration diagram of the control circuit 60 using a microcomputer, which includes a sensor 1 to a central processing unit (hereinafter referred to as CPU) and a read-only memory (hereinafter referred to as R and OM). , a random access memory (hereinafter referred to as RAM), and an input/output circuit 108 .

The CPU 102 calculates input data from the input/output circuit 108 using various programs stored in the ROM 104, and returns the calculation results to the input/output circuit 108 again. RAM 106 is used for intermediate storage necessary for these operations. CPU102°ROM104. RA, M2O
3. Various types of data are exchanged between the input/output circuits 108 via a bus line 110 consisting of a data bus, a control bus, and an address bus.

The input/output circuit 108 includes a first analog-to-digital converter 122 (hereinafter referred to as ADCI), a second analog-to-digital converter 124 (hereinafter referred to as ADC2), an angle signal processing circuit 126, and a 1-bit It has an input means with a discrete input/output circuit 128 (hereinafter referred to as DI○) for inputting and outputting information.

A battery voltage detection sensor 132 (hereinafter referred to as V
It is written as B2. ) and air-fuel ratio sensor 142 (hereinafter referred to as A/FS
It is written as ) and the output from the multiplexer 162 (hereinafter referred to as
It is written as MPX. ), MPX162 selects one of them and converts it to analog/digital conversion circuit 1.
64 (hereinafter referred to as ADC). ADC16
The digital value that is the output of 4 is stored in register 166 (hereinafter referred to as R
It is written as GE. ) to be heset.

The angle sensor 146 (hereinafter referred to as ANGLS) outputs a signal (hereinafter referred to as REF) indicating a reference crank angle, for example, 180' crank angle (in the case of a 4-cylinder engine), and a signal indicating a minute angle, for example, 1 degree crank angle. (Hereinafter referred to as POS.)
is output and applied to the angle signal processing circuit 126, where the waveform is shaped.

DI0128 has an idle switch 148 (hereinafter referred to as IDL) that operates when the throttle valve 14 has returned to the fully open position.
It is written as E-3W. ) and top gear switch 150 (
Hereinafter, it will be referred to as TOP-8W. ) and starter switch (
Hereinafter, it will be referred to as 5TART-8W. ) is input.

Next, a pulse output circuit and a controlled object based on the calculation results of the CPU will be explained. Injector control circuit 1134
(hereinafter referred to as INJC) is a circuit that converts the digital value of the calculation result into a pulse output. Therefore, the pulse INJ having a pulse width corresponding to the fuel injection amount is INJC11.
34 and applied to the injector 12 via an AND gate 1136.

The ignition pulse generation circuit 1138 (hereinafter referred to as IGNC) includes a register (hereinafter referred to as ADV) for setting the ignition timing and a register (hereinafter referred to as DWL) for setting the primary current energization start time of the ignition coil. and CP
These data are set by U. AND gate 1 to amplifier 62 for generating pulse IGN based on set data and supplying primary current to the ignition coil
Apply this pulse IGN via 140.

A control circuit (hereinafter referred to as 15C) controls the opening rate of the bypass valve 61.
It is written as C. ) 1142 through an AND gate 1144. l5C
C1, 142 is a register l5CD that sets the pulse width
and a register l5CP for setting the pulse period.

The EGR amount control pulse generation circuit 1178 (hereinafter referred to as FEGRC) that controls the EGR control well 90 has a register EGRD for setting a value representing a pulse duty and a register EGRP for setting a value representing a pulse period. The output pulse EGR of the EGRC is applied to transistor 90 via AND gate 1156.

Further, the 1-bit input/output signal is controlled by the circuit DI○128. The input signal is IDLE-SW double signal 5
There are TART-8W signal and TOP-8W signal. Also,
The output signal includes a pulse output signal for driving the fuel pump. This DIO is provided with a register DDR192 for determining whether a terminal is used as an input terminal, and a register DOU71.94 for latching output data -11,-.

The mode register 1160 is a register (hereinafter referred to as MO
Write it as D. ), for example, this mode register 116
AND gate 1136 by setting the instruction cell 1 to 0.
, 1140, 11.44. .

All 1156 are activated or deactivated. By setting the command in the MOD register 1160 in this way, rNJc, IGNC, l5C
It is possible to control the stop and start of C output.

A signal DIo1 for controlling the fuel pump 32 is output to DI0128.

Therefore, if such EEC is applied, almost all controls related to the internal combustion engine, such as air-fuel ratio control, can be appropriately performed, and the strict exhaust gas regulations for automobiles can be fully satisfied.

In the EEC shown in FIGS. 1 and 2, the injector 1
The fuel injection according to No. 2 is periodically performed intermittently in synchronization with the rotation of the engine, and the fuel injection amount is controlled by controlling the valve opening time of the injector 12 in one injection operation, that is, the injection time TI. It is.

Therefore, in one embodiment of the present invention, this injection time Ti is basically determined as shown below.

TL= α”Tp・(Kt+Kt Ks)(1+ΣK
+) +Ts...(1) Here, Kconst: Injector coefficient TP
Two basic injection times α: Air-fuel ratio correction coefficient Ts: Injector invalid injection time KL: Steady-state learning coefficient: Transient learning coefficient: Various correction coefficients Ks: Shift coefficient Qa: Incoming air flow rate N: Engine rotation speed, i.e. The basic fuel injection time T is determined by equation (2) from the engine intake air flow rate Qa and the rotational speed N, and the 02 sensor 142 After correcting the air-fuel ratio by feedback by changing the air-fuel ratio correction coefficient α using the signal λ, a more accurate stoichiometric air-fuel ratio can be obtained.
Furthermore, the steady-state learning coefficient Kg is used to correct variations in the characteristics of various actuators and sensors related to air-fuel ratio control and changes over time, and the transient learning coefficient is used to correct acceleration and deceleration. , subtracting the shift coefficient during sudden deceleration from this to obtain the fuel injection time T1
This is what determines the

Here, the learning coefficient Km will be explained. 0□Sensor 1
42 outputs a binary signal (high, low level voltage) depending on the presence or absence of oxygen in the exhaust gas. It is well known that the air-fuel ratio correction coefficient α is increased or decreased in steps based on this binary signal, and then the air-fuel ratio is controlled by gradually increasing or decreasing it. FIG. 3 shows the state of the air-fuel ratio correction coefficient α which detects whether the air-fuel ratio is rich or lean based on the output signal λ of the 02 sensor.

Here, the air-fuel ratio correction coefficient α when the signal of the 02 sensor is inverted, the extreme value from lean to rich is αm&X, the extreme value from rich to lean is α1n, and the average value αav
e is calculated using the following formula.

The idea of this average value is well known, but in this example, the average value αave is the upper limit value (T, U, L) and the lower limit value (T, L
, L), the average value αave and 1.0
For the deviation of , let be the steady learning correction amount. This calculation of the steady learning correction amount Kg is performed in all areas where feedback correction is being performed by the 02 sensor.

FIG. 4 shows a table in which the steady learning correction amount Kt is written. In this table, Kt is written at division points determined by the basic fuel injection time Tp and the engine speed N. This learning timing is when the 1 division point does not change and the number of extreme values reaches n times - 15=. The table shown in FIG. 4 is defined as a steady learning map. This steady learning map is divided into dividing points (
Here, 64 points) All of the points can be filled in through learning.
Practically impossible. Therefore, it is necessary to create unlearned dividing points by referring to learned dividing points.

Therefore, next, this method of creation will be explained.

FIG. 5 shows an example of a buffer map and a comparison map, which are used to create a steady learning map and wait for the same number of points as the division points of the steady learning map.

FIG. 6 shows a block diagram of a routine for creating a steady learning map. In (1), the steady learning map and comparison map are all cleared, and the steady learning correction amount is written to the buffer map. However, at this point, double writing to the buffer map is not performed. In (2), when the number of writes in the buffer map becomes 0, the contents of the buffer map are transferred to the comparison map, and in (3), the buffer map is Create everything and transfer the contents to the steady learning map. In (4), the contents of the comparison map are retransferred to the buffer map. From this point on, a value of 1 is used to calculate the fuel injection time. Up to this point, in equation (1), is 1.
It is 0. In (5), the steady learning correction amount is written in both the steady learning map and the buffer map, the air-fuel ratio correction coefficient α is set to 1.0, and the contents of the buffer map and the comparison map are compared.

When the difference in the compared contents reaches a certain number, (6)
In this step, the routines (2) to (4) are repeated.

According to this embodiment, since the steady-state learning correction Kt stores the deviation from 1.0, the air-fuel ratio correction coefficient α can be controlled around 1.0 with - times of correction, and the harmful exhaust gas Components can be reduced.

In addition, in the steady learning map shown in Fig. 4, the map values in the rightmost column and bottom row are used when the basic fuel injection time is Tp7 or more and the engine speed is N7 or more, so even in the power range, the optimum values are always maintained. It is possible to perform corrections that increase the power.

Next, the negative side of the learning routine for the steady learning coefficient Km will be explained with reference to the flowcharts of FIGS. 7 and 8.

The process according to this flowchart is repeated at a predetermined period after the engine starts, and first, step 30"C
It is determined whether or not the -02 feedback control is being entered, and if the result is Yes, the process proceeds to step 302. If the result is No, proceed to step 332. In step 302, the signal of the 02 sensor is λ=1 (stoichiometric air-fuel ratio A/F=1
4.7) is crossed. If the result is NO, proceed to step 332 and perform a well-known integration process (not shown).
will be carried out. If the result is Yes, step 304
Then, the average value αave shown in equation (3) is calculated. In step 306, the average value αave is determined as shown in FIG.
Determine whether it is within the lower limit value and the result is Yes.
If so, normal feedback control is being performed, so the counter is cleared in step 326, and step 33 is performed.
Head to 2. on the other hand.

If the average value αave is outside the upper and lower limits, step 3
In step 08, the difference between the average value αaVe and 1 is set as the steady learning correction amount. Next, in step 310, as shown in FIG.
The current dividing point determined from the basic fuel injection time Tp and the engine speed N is calculated, and in step 312, it is compared with the dividing point one time before this routine to determine whether the dividing point has changed. . If the dividing point has changed (Yes
), the dividing point at which the steady learning correction amount Kg is placed has not been determined, so the process proceeds to step 326. If the dividing point has not changed, increment the counter in step 314,
In step 31G, it is determined whether the counter has reached n. If the counter value is not n (No), step 33
Head to 2. When the counter value reaches n (Yes), the counter is cleared in step 318 and the process proceeds to step 320. In step 320, it is determined whether or not the first generation of the steady learning map, which is the operations (2) to (4) explained in FIG. 6, has been performed. If the map has not yet been created, the process proceeds to step 332 and subsequent steps, and the operation (1) described in FIG. 6 is performed. In step 322, it is determined whether the division point has already been written.

=19- Since it has already been written (Yes), proceed to step 332 without doing anything. If the result is NO, in step 324, the steady learning correction amount calculated in step 308 is written to the dividing point. If the first steady learning map has been created in step 320 (Yes), the process proceeds to step 328 and subsequent steps, and operations (5) and (6) explained in FIG. 6 are performed.
In step 328, the steady learning correction amount is added to the division points of the steady learning map and the buffer map. Then, in step 330, the air-fuel ratio correction coefficient is set to 1.0.

Therefore, by repeating the processing according to these steps 300 to 332, the process described in FIG.
1) (5) The operations of (6) are obtained.

Next, using the flowchart shown in Fig. 8, (
2) The operations of (3) and (4) will be explained.

At step 350, it is determined whether the first steady learning map has been created. If you haven't created it yet (No), step 3
The process proceeds to step 54, where the number of buffer map writes is checked. When the number reaches m, the process proceeds to step 356, but if the number has not yet reached m, the process proceeds to step 370.

If the first steady learning map has been created in step 350 (Yes), then in step 352 the difference in data between the buffer map and the comparison map is checked. If there are Q differences in the contents of the buffer map and comparison map.

Proceeding to step 356, a steady learning map is created.

If there are no Q differences in the contents, proceed to step 370.

At step 356, a map creation flag is set;
Prohibit writing learning results. In step 358, the contents of the buffer map are transferred to the comparison map, and step 3
At 60, a steady learning map is created using the buffer map. In step 362, the contents of the created buffer map are transferred to the steady learning map, and in step 364, the contents of the comparison map are transferred to the buffer map. At step 366, a flag indicating that a steady learning map has been created is set. This flag is used for determination at step 350 and step 320 in FIG.

In step 368, the map creation flag set in step 356 is reset.

The above-described portion is an example of a method for creating a steady learning coefficient by engine control by a microcomputer, zero feedback using the output of the 02 sensor, and steady learning of the air-fuel ratio correction coefficient. Next, the characteristic index obtained from the air-fuel ratio correction coefficient for two different operating states, which is the central part of the invention of this embodiment, and the control constants (injector air retention, hot water injection rate, We will explain the action of leaving Koshinkari 1) naturalized).

First, a state in which the control constants do not match the physical characteristics of the engine, sensor, and actuator will be referred to as an unmatched state. Using the unmatched injector coefficient Kconst, the invalid injection time Ts, and the air flow rate characteristic Q&, the fuel injection time is determined by the learning coefficient of equation (1): Kg=1.0, Kt=O.

Assuming that K s ” O is in a steady unlearned state, the following formula % Formula % (4) The value Kconst* that matches each of this state
When written using *Qa*, Ts*, and COE F*, the following formula is obtained.

T+ = Kcon5t* + Qa*/ N-COE
F umbrella + Ts*...(5) From equations (4) and (5), the following equation holds true.

Kconst ・Qa/N-Co EF ・a +
Ts= Kconst'l"Qa*/N' Co E
+Ts* for F' (6) Here, using Tp*=Kconst*·Qa*/N, the following equation is obtained.

α(N, T P”) = E1・E2・E3・E
4...(7)E1=(Ts*Ts)
/ (Engineering plcOEF*(N, Tp*)) +1...
(8) E2=Kconst*/Kconst
-(9)E3=4ra6h
...(10) E4=COEF*(N, Tp*)
/C0EF(N,Tp)...(11)(7)~(11
) according to

Ts; El (mainly a function of TP*, (see Figure 13) Kcolst; E 2 (constant) Qa;
It can be seen that E 3 (function of Qa) COEF; E 4 (function of N, Tp*), etc. are each reflected in α as a product.

Next, α(N r T p*) is set to equal Q, as shown in Figure 9.
Consider the case where N and T p * are divided so that the lines are arranged diagonally. Here, for simplicity, a 4×4 map is considered, and the learning value of α is assumed to be the intersection point of the grid. Next, (7
) FIG. 10 shows the error E for each factor when unmatching occurs according to the equation. The value of α map at this time is
Shown in Figure 1. Here, it is assumed that C0EF in equation (11) is almost matched, and E4=1.

Here, for the vertical axis Tp, the value of El is al, C2,...
・and on the diagonal line is the unmatching E3 of Q.
The values cl, c2°... are applied, and the unmatching term E2 of the injector coefficient is bl for all map values.
is on.

The coefficients of the α map at this time are regarded as a matrix, and each element thereof is designated as M r i as shown in FIG.

Each element of the matrix reflects each unmatching factor in the form shown in FIG. For example, as shown in FIG. 12, a1 to a3 normalized by C4 are obtained by dividing the elements of the matrix. Therefore,
By being able to capture the characteristics of Tp with respect to Tp, for example, Ts can be corrected and matched based on the tendency shown in FIG. be able to.

Next, FIG. 14 shows that Qa is corrected in the same manner as FIG. 12. At this time, it is standardized to C4.

Taking into account the above characteristics, the matching coefficient in the unmatched state is determined as follows.

Kacdl; Kconst correction (scalar; 1 variable) Kgcd2 (N, Tp); Ktr-correction (N, Tp
Map; number of N divisions, number of Tp divisions) K*ca3; For Ts correction (scalar; 1 variable).

a4 (Qyu); Q& correction (vector; Q& division number) Considering (4) and (5), these coefficients may be set as follows.

Ktcdl = Kconst*/ Kconst
...(12) KtCd2 (N, T
p)=COE F*(N, Tp*)/COE F*(N
, Tp*)...(13) Ktca3=TsNeTs
=・(14)Ktca4 (Qa)”Qa*/Qa
(15) Also, the fuel injection is performed using the following formula.

...(16) Tp'=Kconst-Kmcdl・1Kaca4
Qa Shea/N Inuue Presentation' = Kmc+1 ・Kzci4 (Qa)Kconst-
Qa/ N=Ktcal ・Kmca4 (Qa)・T
p - (17) According to equations (16) and (17), from the change in the coefficient appearing in α due to 0□ feedback, Kconst and T
Allocate coefficients to be corrected for each s and Qa. especially,(
As shown in formula 17), the basic injection time is determined by the correction Kzc+1 of the injector coefficient (Kconst) and the correction Ktc of Qa.
It is corrected by the product of d4(Qa). Furthermore, as shown in the following equation, the fuel injection time TI is C0EF with respect to Tp'.
Calculated by multiplying the product of ' and α, battery correction voltage TS'
are added and calculated.

TI=Tp' ・C0EF- exposed cd2 N Tp ・
α+(Ts+Ktci3)COEF'
Ts'...(18) Based on the above results, we separated the fuel injection time, which was conventionally corrected by α, according to the necessity of occurrence, and in particular, changed the basic injection time T, to (
17) It can be corrected as shown in the equation. In other words,
Separate learning for each occurrence factor can be achieved.

Below, we will discuss the matching procedure based on the above analysis.

First, set Kconstl T5 and Qa, perform 02 feedback to realize various operating states, and calculate α(N).

TF) Perform constant learning of the map. At this time, (1
) Steady learning is performed under operating conditions such that the various correction terms in the equation are O except for Ktra, which is the air-fuel ratio correction coefficient of feedback control during steady state, that is, under the following conditions.

After warm-up, start operation and check steady operation (1ΔN1 × ΔNS
, lΔTPI<ΔTps), α(N.

'rp).

Next, separate each factor from α. Here, first, the air flow rate is corrected. From the characteristics of the elements of the matrix in Figure 11, the value normalized by C4, which is the value of E4 when Q a 4 as shown in Figure 14, is calculated by dividing the elements of the matrix as shown in the table. I understand that. From the table, it can be seen that there are several calculation methods depending on the element. When all the α maps have been learned, it is preferable to perform averaging processing if it is effective based on the degree of dispersion of values. Furthermore, when the number of learning α maps is small, Qa correction may be performed by acquiring the minimum necessary value.

In other words, for Kacd4 (Qa), first, the following equation is corrected to make the relative error constant at 1/c4.

K shit cd4 #(Qa) = c i / c 4
...(19) Qal'(Qa)=KtCd4#(
Qa)・'Qal・(20)Kmed4- # (
Qa) = c 4 ・Ktca4 # (Qa) −
(21) At the same time as performing the Qa child table correction as shown above, the diagonal elements of the α map are calibrated as shown in the following equation. This is a correction made because the influence factor on the α map is eliminated by calibrating using the Qa child table equation (20).

MIJ=MIJ ・(c 4/c k)
−(22) However, k=j−i+4 ・−(23)
1+ j; ]-+ 2+ 3+ 4 That is, j
−i=uniform correction for constant diagonal elements. As a result, the term related to Qa in the α map becomes ci=c4.

The above can be summarized as follows.

First, the Qa error characteristic is flattened using the learned α(N,'rp) map, and a Qa correction table is created. Here, the α map is also corrected in accordance with the Qa correction.

Regarding the correction of Kconst and Ts, the following two methods are proposed. One is to correct Ts by dividing the coefficient of the α map matrix, and the other is T p
This is a method of correcting Ts by plotting ' and T.

As shown in Fig. 13, there is unmatching in the invalid injection time Ts, and when Ts*Ts is not 1'0'', it exhibits a hyperbolic characteristic with respect to Tp, and becomes 1 when T is large. Therefore, for example, Tpl or T
p2 low load region al/a4. to the correction term of Ts so that the value of a2/a4 approaches 1. Find the optimal value of Ts by increasing or decreasing d3. Here, for example, when Ts is small, a 1 in the low load range as shown in FIG.
/a4.

Since a 2 / a 4 is greater than 1, K consideration ca
Perform the operation to increase 3. Even when Ts is large, Kzcd3 is made small using the same method. At this time, al
If the value of /a4 shows a stable increase/decrease tendency, the magnitude of increase/decrease may be set as shown in the following equation in order to increase the convergence density of Ts.

Kmc, +3 = Kmc++ 3 + (const
ant) la 1 / a 4 (24) Calculation of coefficients such as a 1 / a 4 is as shown in FIG.

Assuming that when the optimal value of T5 falls within a predetermined range in the above procedure, the value of α map becomes C8, which is almost constant, then the value of each element is set to be near "1" by a common factor. From (5) to (10), E 1 = E in equation (7)
1 and E 3 =Qa4*/Qa4, a linear equation holds true.

Kconst*/ Kconst Hc 4 = a
s − (25) Kmca 1 = Kcons
t*/ Kconst ・= (26) Also, from equation (21), Ktcd4(Qa)=ci=c4(ci/c4)=c4
-Kmcd4#(Qa)...From equations (27), (26), and (27), the following equation holds true.

Ktc< 1 ・Kmci4 (Qa)= Kcons
t*/ Kconst・C4・Ktcd4#(Qa)
-(28) Here, using equation (29), the following equation is obtained.

Ktcdl ・Kmca4 (Qa)= as ・
Kmca4- # (Qa) (29) The above can be summarized as follows.

From the revised α map, a characteristic value that depends on Ts unmatching and is a function of Tp is calculated, and the Ts correction value is optimized using this characteristic value as a reference value.

Next, the product of the uniform error of Q& and the error rate of Kconst is determined using the value of the common coefficient of the α map that has been approximately flattened by the above-described operation. By the above operations, Ts and 1K
You can calibrate const.

Next, K CON using T p ′ T s plot method
5 and the correction of Ts.

When Qa child table correction is executed, Q.

is Qa' = Qa*/c 4. At this time, the fuel injection is as follows.

Tt=Kconst-Kmca4 #Qa'Qa/
N-C0E F*TP' (Kconst*/ Kconst-Qa4'/Q
a4-3+Ts*...(30) =32- Here, when (Tp', T+) is obtained and plotted, the plotted trajectory becomes a straight line as shown in Figure 15, and Tp'= The intercept O becomes the volume Ts, and the slope of the straight line takes a value as shown in equation (29). Here, it is considered that C0EF*=1 in a well-matched state, so if the slope of the straight line is ks, the following equation holds true.

Kconst*/ Kconst' Qa4 */ Q
a4 = Ktca 1 ・c 4 = k, s...
-(31) With the above, matching of coefficients is possible in the same way. The above can be summarized as follows.

Tp and T calculated using the corrected Qa child table
+ is obtained in each steady state of operation, and ('rp+Tj)
From the slope of the straight line formed by the uniform error of Q, l and K CO
Find the product of the errors of n S t. Also, the correction value of Ts is determined from the intercept of Tp=O.

By repeating the following operations, each control constant can be optimized. By optimizing each control constant, the calculation of the basic injection time can be optimised, and the ignition timing can also be set appropriately, making it possible to achieve comprehensively appropriate engine control.

Above, the specific operation of this embodiment will be explained using the flowchart. FIG. 16 shows the configuration of the control constant correction device. First, the air-fuel ratio feedback means 400 generates the air-fuel ratio correction coefficient α using the 02 feedback. The steady-state learning means 500 performs the steady-state learning shown in FIGS. 7 and 8, and learns the air-fuel ratio correction coefficient in a steady state. Next, the characteristic index calculation means 600 uses the air-fuel ratio correction coefficient that has been steadily learned.
Accordingly, the characteristic index regarding each control constant is calculated. With reference to the characteristic index, the control constant correction means 700
Correct the control constants and optimize them.

Here, the characteristic index regarding the control constant is a+/a4. shown in FIGS. 12 and 13. It shows values such as ct/c4, which are obtained by dividing the learned air-fuel ratio correction coefficient between elements. Since the air-fuel ratio correction coefficient is a value close to 1'', division can also be replaced by subtraction.

Next, specific processing steps will be explained using a flow diagram. 1st
FIG. 7 is a schematic flowchart, in which a characteristic correction routine 2000 is performed after the process of steady learning 500.

FIG. 18 is a schematic flowchart of the characteristic correction routine.

If the learning number is larger than the predetermined value NA, process 2020
In steps 2020 to 2050, the characteristic correction is not performed if the condition is not satisfied.

Detailed logic 2060 and simple logic 2070 are distributed. If the number of acquisitions of Q&characteristics is greater than a predetermined value NQA and the acquisition rate of T, characteristics is greater than NTS, detailed logic 2060 is executed; otherwise, simple logic 2070 is executed. FIG. 19 is a flow diagram of the simplified logic. First, matching status flag execution 2110 is executed. Here, matching (correction processing) is determined to be complete when the value of the learning map obtained through steady learning has a small amount of change as a whole with respect to the previous value. If the amount of change exceeds a predetermined value, it is determined as a matching error (correction processing error), and a flag is set. In judgments 2120 and 2130, judgments are made according to the flags. The process ends when matching is completed. Note that the completion flag is activated in a long cycle by a separate task not described here, and matching processing is performed periodically. Furthermore, when an error occurs, matching error processing 2150 is executed.

Here, the correction process is basically canceled and only the air-fuel ratio correction coefficient is controlled by steady learning. If the matching is not completed and there is no error, the correction process 2135 and subsequent processes are performed. First, in the switching 2135 of i, there are two systems of learning maps and other correction coefficients, and one of them is used as the current map, but since the other map is used as the calculation map, in this switching 2135, the current map is Switched to calculation map. Next, map condition search 2140 is executed. Here, in order to correct Kconst,
Search for values in the map of large areas of TP.

This is because, in the region where Tp is large, assuming that there is no variation in the Qa characteristics, the influence of Ts is small and only the influence of K CON S t appears. Next, a neutral average calculation 2160 of α is performed. Here, processing is performed to calculate the average value of the remaining α values excluding the maximum and minimum values among the α=36− values extracted in process 2140. If the number of α extracted here is two, the average value is calculated, and if it is one, the value is calculated as ACPPROC. In the next process, ALPROC is substituted for KLCDI, which is the correction value of Kconst. Subsequently, in map condition search 2180, α search is performed in an area where Tp is small, and processing 2140 is performed.

Similar to 2160, in process 2190, ALPROC
Calculate. The correction value KLCD3 of Ts is determined in step 2200.
It is calculated by multiplying by the gain KKKCD3 as shown in FIG.

Next, in map correction 1221.0, the learning map regarding Kconst and Ts is corrected. When the map correction is completed, the map system is switched and control is performed using the new coefficients. The above is the characteristic correction using simple logic.

Next, FIG. 20 shows a detailed logic flowchart.

Here, correction processing is performed on all of Qa and Kconst+Ts as unmatching.

First, initial processing 2410, 2420, 24.30°24
Up to 35.2450, matching completion and error determination processing are similar to the simple logic. Qa characteristic table calculation 2
At 440, characteristic indices are calculated, but if all characteristic indices are not available, interpolation calculations are performed to calculate all characteristics.

This interpolation calculation allows correction processing to be performed even if all learning has not been completed. Subsequently, in Ts characteristic table calculation 2460, a Ts characteristic index is calculated. Here, Ts
Since the characteristic index is a monotonous characteristic as shown in FIG. 13, if the obtained characteristic deviates from the characteristic shown in FIG. 13, the error flag FTSCMPER is turned ON. Next, at decision 2470, if there is an error, processing ends. If no error has occurred, KLCD3, which is a correction value of Ts, is calculated. Next, in map correction 2500, the learning map is corrected using Kconst+Ts, Qa. When the map correction 250o is completed, the map system is switched and control using new coefficients is performed.

The above is the correction based on detailed logic.

Based on the contents of the embodiments described above, the main parts of the present invention will be explained using FIG. 21.

FIG. 21 shows the Ts and Qa child table correction method.

First, regarding the method of correcting Ts, T p plate=1. , 5
Find m sec and Tph=3 msec, and let α at that time be α, , αh (2610), respectively. At this time, as shown in Fig. 11, α
If an error occurs between the values α1 and αh, it is an error caused by unmatching of Ts. Therefore, it is possible to correct the unmatching of Ts by making adverse use of this error.

The correction formula for Ts is shown in equation (32).

Ts*=Ts+KDTS ・(6m (!h)
''' (32) Here, KDT which is the Ts correction gain
S is equal to or greater than Tp (TP consideration - Tph), and this time Tph=1, 5 m5ec
and Tph” 3 m5ec, KDTS=3,
By substituting this into equation (32), Ts is corrected.

Now, when this error is within ±1%, it is determined that the correction of Ts has been completed. Next, moving on to Qa child table correction, the rule for starting Qa child table correction is that the output voltage of the air flow sensor is between 1.04V and 3.44V. Divide up to V into 16 regions every 160 mV and store α at that time in each memory (2710)L. When α is stored for 8 or more different regions, Qa child table correction is started. However, for 8 or more different areas, 16
Assume that the area is divided into three areas, large, medium, and small, and α is stored in two or more areas in each area. Further, the stored α within 3 days is used.

In the Qa child table correction method (2810) using α, as shown in FIG. i, = I H2
・−) to get. This corrected 8 or more Qai (i = 1
.

2...) Based on the air flow characteristic formula (equivalent to 4 hours)
The coefficients of are determined by the well-known method of least squares.

The above air flow rate characteristic equation is the voltage value Q output by the air flow sensor for the air flow rate Qa expressed by equation (17).
This approximates aV using a quartic function. In other words, Qav=a in formula (33), +atQa+a2Qa''+a2Qa3+a
4Qa'...(33) The coefficient at(j=o~4) is the correction term Q-+(i=1
．． 2...) is determined according to the least squares method (28
20) Do it.

After this, using the above air flow rate formula, recalculate the air flow value for the entire Qa child table area, that is, 64 points,
This calculated value is updated (2830).

Further, another embodiment is shown in FIG. 23.

FIG. 23 is a table in which α is written, and α is written at division points determined by the output voltage QaV of the air flow sensor and the engine speed N.

Therefore, if we take one of the engine speeds N, and the vertical axis is α and the horizontal axis is the output voltage of the air flow sensor, then T
If s contains an error, the result will be as shown in FIG. 24, and Ts can be corrected in the same way as in FIG. 13 described above.

Next, Qa table correction is obtained by averaging α for each output voltage of each air flow rate sensor. With the averaged α,
Correct the air flow characteristic value, use the corrected air flow characteristic value to find the air flow characteristic equation by the least squares method, and use this air flow characteristic equation to calculate the entire area of the Q&child table (64 points). Recalculate and update the air flow value.

〔Effect of the invention〕

According to the present invention, since the control constants can be corrected using simple logic, it is possible to significantly shorten the control constant matching adjustment of the control unit, shorten the development period, and ensure that the control constants are correct even when the actual vehicle is running. Since it is optimized, maintenance-free operation becomes easy. Furthermore, by monitoring the correction coefficient, it is possible to accurately grasp changes in the characteristics of sensors and actuators over time during periodic inspections and the like.

[Brief explanation of the drawing]

Fig. 1 is a block diagram of an embodiment of the present invention, Fig. 2 is a block diagram, Fig. 3 is an explanatory diagram of an air-fuel ratio correction coefficient, Fig. 4 is a block diagram of a learning map, and Fig. 5 is a block diagram. is a map configuration diagram, Figure 6 is a diagram showing the map change procedure, Figure 7 is a learning flowchart, Figure 8 is a map creation/change flowchart, and Figure 9 is the map configuration. Figure 10 is
Characteristic table. FIG. 11 is a diagram explaining the learned value, FIG. 12 is a diagram showing the characteristic index of Ts, FIG. 13 is a diagram showing the characteristic of Ts, FIG. 14 is a diagram showing the characteristic index of Qa, Figure 15 shows Tp
Figure 16 is a diagram showing the correction characteristics, Figure 17 is a schematic flow diagram of the characteristic correction routine, Figure 18 is a flowchart of the correction logic, and Figure 16 is a diagram showing the correction characteristics.
9 is a flowchart of simple logic, FIG. 20 is a flowchart of detailed logic, FIG. 21 is a flowchart of the main part of the present invention, FIG. 22 is a flowchart of correction of air flow characteristics, and FIG. This figure is a configuration diagram of the learning map, and FIG. 24 is a characteristic diagram of the air amount sensor. 24.02 sensor, 12...injector, 102.
-CPU, 1106-RA, 104・=ROM, 400
...Air-fuel ratio feedback means, 500... Steady learning means, 600... Characteristic index calculation means, 700...
Control constant correction means, 282o...air flow characteristics? 〇− ＾ ＾ ＾ ＾ to
＾、N
Tome Q 9 Kei Ryo Fake Mi 525 Gyakugi Electric Suspicion R', Hebei 4 Genus ξ and D TP' Grass, 4 mouths/7m

Claims (1)

[Claims] 1. An air-fuel ratio control device comprising a sensor that detects an air-fuel ratio based on engine exhaust gas components, and having an air-fuel ratio correction coefficient obtained by feedback control of the air-fuel ratio based on a detection signal of the sensor. An air-fuel ratio control device characterized in that a map of air-fuel ratio correction coefficients is provided based on the output voltage of the air flow sensor. 2. An air-fuel ratio control device that performs a process of correcting air flow characteristics when a predetermined number or more of map values of air-fuel ratio correction coefficients according to claim 1 have been learned. 3. An air-fuel ratio control device according to claim 2, characterized by having a function of determining the entire range of air flow characteristics by the least squares method based on the corrected data of air flow characteristics. 4. In claim 1, the basic injection time T
It is characterized by having a memory of air-fuel ratio correction coefficients for _P_h (large basic injection time) and T_P_l (small basic injection time), and performing the process set forth in claim 2 only when the difference between them is small. air-fuel ratio control device.