JPS63306248A - Air-fuel controller for internal combustion engine - Google Patents

Abstract

PURPOSE:To improve control accuracy by operating a transient learning value based on an air-fuel ratio feedback correction amount upon elapse of response delay time corresponding to engine temperature, and enabling corresponding learning for large shift of air-fuel ratio from a target level. CONSTITUTION:A control unit 35 operates a basic fuel injection quantity based on detection values of a throttle valve opening sensor 25 and a crank angle sensor 32, and carries out feedback correction of air-fuel ratio based on a detection value of a wide range air-fuel ratio sensor 34. Then fuel adhesion speed representing variation of engine load is obtained based on cooling water temperature, throttle valve opening and rotation, and upon judgement of transient state, the air-fuel ratio is corrected with a transient learning correction factor. The transient learning correction factor is updated corresponding to the air-fuel ratio feedback correction amount with a response time lag based on a detection value of a cooling water temperature sensor 33.

Description

DETAILED DESCRIPTION OF THE INVENTION (Industrial Application Field) The present invention relates to an air-fuel ratio control device for an internal combustion engine, and particularly to one equipped with learning [R].

(Prior Art) Electronically controlled fuel injection engines are actually widely used because of their high fuel metering accuracy, and fuel is intermittently supplied to the engine intake system from the fuel injection valve.

Here, since the valve opening pulse width of the injection valve is proportional to the amount of supplied fuel, the basic pulse width T based on the engine load (for example, intake air fiQ&) and the engine speed N is used for injection amount control.
The injection pulse width "I"i to be output is calculated by correcting p (=K-Qa/N, where K is a constant) according to other operating variables.

In addition, the stoichiometric air-fuel ratio at which the conversion efficiency of the three-way catalyst is optimal is determined by
Air-fuel ratio as control target (hereinafter referred to as "target air-fuel ratio")
On the other hand, an air-fuel ratio feedback correction amount LAMBDA is calculated based on the deviation from the actual air-fuel ratio (hereinafter simply referred to as "actual air-fuel ratio"), and this feedback correction amount is used to This has been corrected.

1" i "'l" p X COE F X L
A M B D A + T s (1) However, in the opening formula (1), COE F is the sum of various correction coefficients, and T s is the invalid pulse width.

However, in feedback control, once a deviation from the target air-fuel ratio occurs, the feedback control amount always exists, so by adding the feedback control amount in advance, the deviation from the target air-fuel ratio will be reduced during the next calculation. A learning function is provided to prevent this from occurring (for example, Japanese Patent Application Laid-open Nos. 58-104335 and 58-11083).
See No. 8. ).

(Problems to be Solved by the Invention) Incidentally, in the air-fuel ratio learning control described above, the average value of LAMBDA over the losing cycle of the feedback control cycle is employed as the learning value. If LAMBDA is the only parameter used to calculate the learning value, learning cannot be started until LAMBDA increases or decreases in a stable cycle while waiting for changes in operating variables to occur in LAMBDA. Therefore, it is not possible to deal with driving changes while waiting, and there is a limit to the improvement of learning accuracy. In addition, taking a specific method of feedback control, for example, differential-integral control, L
Since AMBDA does not have a characteristic that changes smoothly, a predetermined number of samples is required to obtain a fast average value, and a minimum period of time is required to secure this number of samples.

Therefore, the present applicant has proposed an air-fuel ratio control device that calculates a learned value using as parameters the main operating variables that affect the air-fuel ratio by M5 during steady state or transient operation.

This proposal is an example applied to a device (hereinafter referred to as rsPI device) in which an injector that serves all cylinders is installed near the intake throttle valve, and the effective injection pulse width 'I' e (= 1" i - T
s, where T i is the fuel injection pulse width to be output,
'I' s is the invalid pulse width. ) is QAINJI
: The basic pulse width 'rp to be applied and the correction amount related to fuel delay (hereinafter simply referred to as "fuel delay correction amount") K
ATHO8 is separated into two parts, and their sum Tc=(Tp
+KATHO3)XLAMBDA...(2) Calculated as follows. This is because the SPI system differs from a system in which the injection valve is provided in the intake boat of each cylinder by the amount of air flowing into the cylinder (hereinafter simply referred to as "cylinder air amount").

) Q c Y L and the air fi (QA
I N J ) do not necessarily match and there is a fuel supply delay for the fuel injected from the injector to reach the shilling, so we newly introduced KATHO3 for the latter fuel delay. This is what I did.

In addition, changes over time or differences in fuel properties will cause a decrease in accuracy, so each variable ('l'p and KA TH
A learning value (a learning value K B L RC for Tp, a learning value K B 1"L RC" for K ATHOS i) is introduced for each of the two OS's.

”re= (1' pX K B L RC1K A
l″HOS
DA...(3) Coconi, KBL
E<Cl! , the target air-fuel ratio ('FFBYA), the actual air-fuel ratio (AFBYA), and the feedback correction amount (LAMBDA) calculated based on the deviation of the rain air-fuel ratio as parameters. variables plus 1' p and K A T HO
Each is calculated using S as a parameter.

Furthermore, from equation (3), KBLRC is the learning value of 'rp (
Basic injection learning value), K B T L RC (i K
A They each mean the TlO2 learned value (fuel delay learned value), but the KBLRC calculation is executed when a steady state is determined, and the KBTLRC calculation is executed when a transient state is determined, so the following will be explained below. K B L RC is also referred to as the steady state learning value, and K B i''L ) (C is also referred to as the transient learning value.

Now, as shown in Fig. 10, the fuel delay during a transient period becomes larger as the cooling water temperature becomes lower. Therefore, when the temperature is low, the timing for calculating the learning value (hereinafter simply referred to as "learning timing"). If this is not taken into consideration, the learning function may not be fully utilized. For example, PlsB diagram and ff19 diagram show the air-fuel ratio sensor output when Tw is different under the same acceleration conditions (Figure 8 is at high water temperature and fjS9 is at low water temperature). When the water temperature is high, the speed change that occurs after acceleration occurs with good response, but when the water temperature is low, the response is delayed. Learning is effective in eliminating this lean phenomenon, and learning accuracy can only be improved by accurately grasping the degree of leanness.To do this, it is necessary to determine the response when the lean phenomenon occurs significantly, that is, during transient periods. The learning timing should be after the delay time has elapsed.

However, since the learning timing in the previous application is immediately after determining the transient period regardless of the difference in control, the learning timing deviates significantly especially at low water temperatures, reducing the learning accuracy. For example, as shown in Fig. 9. Even at low water temperatures as shown in
If the learning value is calculated at the same timing as the learning timing for high water temperatures shown in the figure, the learning timing for low water temperatures is too slow, so the degree of leanness will be underestimated, and the learning accuracy will decrease due to the underestimation. invite.

Note that the prior application also attempts to stop learning when the water temperature is low, but this is only in a steady state and does not focus on the learning timing.

This invention was made with attention to this point, and the response delay time is determined in advance according to the cooling water temperature, and after the response delay time has elapsed after it is determined that a transient state is occurring. It is an object of the present invention to provide an air-fuel ratio control device in which the learning timing is set as the learning timing.

(Means for solving the problem) In this invention, as shown in FIG. means 1 for detecting the actual air-fuel ratio; means 2 for detecting the actual air-fuel ratio; means 3 for calculating the feedback correction amount (LAMBDA) based on the deviation between the detected actual air-fuel ratio and the target air-fuel ratio; Hand Fi5 sets the response delay time (tR) according to ``w), Hand fi6 determines whether or not it is a transient IO (including not only acceleration but also deceleration), and Hand fi6 determines whether the IO is transient. means 7 for calculating a learning value (for example, KBTLRC'') at the time of transition based on the feedback correction 1ii LAMBDA after the response delay time tR has elapsed, and calculating the calculated learning value KBTLRC and the feedback correction amount 1°. A means 8 is provided for calculating a transient fuel injection amount by correcting the basic injection ft'l'p in AMBDA.Note that 4 is a means for detecting the engine temperature.

As the engine temperature, oil temperature, intake air temperature, etc. can also be used.

(Function) For example, when the cooling water temperature '- is high, the lean phenomenon appears with good response immediately after acceleration, so a short time (tR+) is set as the response time. The learning timing is set after the time elapsed.On the other hand, when the water temperature is low, there is a fuel delay l)
Since the appearance of the IJ-on phenomenon is delayed, a time (LR2) longer than the LRI is set in accordance with this, and the learning timing is set after the elapse of this LR2 time. Here, it becomes possible to learn the deviation from the target air-fuel ratio in response to the period when the deviation from the target air-fuel ratio becomes larger during a transient period, and thereby the accuracy of the air-fuel ratio by learning is improved particularly at low water temperatures.

(Example) In the PtS2 diagram, one fuel injection valve 24 that covers all cylinders is installed in the intake passage 22 upstream of the intake throttle valve 21 (SPII arrangement).
, and shows a system diagram when the present invention is applied to an engine that uses the throttle valve opening (T O) instead of the air amount as the engine load signal in order to simplify the entire device.

Therefore, in this example, i' VO and 81 leap rotation speed N
The injection pulse width will be controlled as the basic operating variable.

In the figure, a throttle valve opening sensor 25, a sensor (crank angle sensor) 32 that outputs a crank angle basic position signal (Ref signal) and a unit angle signal (N signal), detect the engine cooling water temperature] A sensor (water temperature sensor) 33 that detects not only the stoichiometric air-fuel ratio but also a sensor (wide range air-fuel ratio sensor) 34 that detects other air-fuel ratios over a wide range are provided in each part of the engine. In addition, 26 and 27 are a plurality of electromagnetic valves (referred to as S■) provided in the passage 23 that bypasses the throttle valve 21,
This is for starting control. Further, 28 is a swirl control valve provided on the intake boat.

Signals from the sensors (25, 32 to 34) are input to the control unit 35.
Based on these signals, the optimum injection pulse width is calculated not only during steady state but also during transient times. Here, the calculation content of the injection pulse 1lij4T i to be output to the injection valve 24 is expressed as f:
Figures 53 to 7 (however, Figure 3 consists of Figures (A) and (13), and Figure 5 consists of Figures (A) to (13)).
C), the same shall apply hereinafter. ), but for the sake of convenience, the parts related to the present invention will be explained first, and then the entire control system will be outlined.

The control contents shown in these figures constitute one air-fuel ratio control system as a whole, and the details are as follows.
The figure shows the main routine for calculating the injection pulse width, and the tIS4 figure shows the variables used in the main routine (K A T l-
1OS), FIG. 5 is a routine showing learning contents, and FIG. 6 and vtjs7 are routines showing learning value rewriting processing. The numbers attached to the diagram represent processing numbers. Such control mainly controls the control unit 35 from the CPU, ROM, RAM, and earphones.
A microcomputer consisting of F-7x-X can be used.

In this case, T i and K A i'')l OS are executed in clock synchronization, and T i and K A i'' are executed in a background job.

Now, in this example, there are two learning correction coefficients (steady character correction coefficient K BL RC and transient learning correction coefficient KB TL
RC) is employed, and the calculations are performed separately for the steady state and the transient state, but in this invention, the learning timing for KB1' L RC is particularly an issue.

That is, the learning timing for K B TL RC is not immediately after the transient determination, but after the response delay time tR has elapsed since the fη constant.

Here, since tR corresponds to the fuel delay time shown in FIG. 10, it is assumed to be a value that increases as the cooling water temperature '1'- decreases. “To assign tR according to U w,
For example, it is possible to store a two-dimensional table (LR table) with the '1゛ field as a parameter in t(OM) and search the tR table from Tw at that time. Instead of , it can also be obtained from an arithmetic expression using ′rw as a variable.For example,
When the curve shown in FIG. 10 represents tR, if this curve is approximated by a straight line, then the straight line 1! tR=A BXTw(
However, L, A and B are positive constants. ), so
This formula also gives tR.

Then, to determine whether the time tR has elapsed since the determination of the transient state, a counter that counts at a constant cycle is started immediately after the determination of the transient state, and the second counter value CN'l'1' and the reference value L ( By comparing CN T i”≧L, tI<
It is determined that the time has elapsed. After the determination, a transient learning value is calculated.

Steps 101 and 102 in FIG. 5(B) calculate the response delay time tR and determine whether the LR time has elapsed.
q constant (step 82 in Figure 5 (A)) and ``transient learning (
This is a new part added between steps 103 to 109). In this example, it is determined that it is in a transient state by comparing the adhesion speed VMI' (described later), which indicates the amount of change in engine load, with the transient determination level (LKΔ'F) (that is, when VMF>LKA'1') has been determined (
Step 82) in FIG. 5(A), here, the amount replacing VMF also includes the amount of change in throttle valve opening (ΔT V○).
and the amount of change in the steady flow rate Qo, which will be described later.

Therefore, if the response delay time is set in this way, even if it is determined that it is a transient period, the 8-transition time learning will not be performed immediately after that as the learning timing, but the cooling water temperature at that time ゛l The transient learning value (K
BTLRC) is executed (steps 82, 10
4.102, 103-109).

For example, when the water temperature is high as shown in Fig. 8, the lean request responds well immediately after acceleration (it appears, so the time when fq is determined at the time of acceleration (to)') is set as the learning timing after the elapse of the tRI time, and this On the other hand, when the water temperature is low as shown in Fig. FF19, the fuel evaporation rate decreases and the reduction in fuel temperature occurs with a delay, so the learning timing is set after the time LR2 (to2>tRI) has elapsed. Note that the reason why the learning timing extends over a predetermined period of time in both figures is because the average value is used. This point will be discussed later.

Here, the I'iT function is to learn the deviation in response to the period when the deviation from the target air-fuel ratio becomes larger during a transient period, and this improves the accuracy of the air-fuel ratio by learning, especially at low water temperatures. However, driveability and exhaust emissions in this temperature range are improved. On the contrary,
If the learning timing is the same regardless of the cooling water temperature, the learned value will be an appropriate value in the water temperature range that matches this timing, but it will not be possible to give an appropriate value in other temperature ranges. .

In addition, while the learning timing is set immediately after the transient determination based on the high water temperature, it is possible to stop learning itself at low temperatures where the response delay is large (and the learning accuracy is poor), but in this case, In this case, the learning performance is insufficient.In consideration of this, according to this example, learning is performed by reducing m even when the water temperature is low.

In other words, even if the learning itself does not stop when the water temperature is low, by shifting the learning timing depending on the cooling water temperature, the learning function can be fully utilized.

Next, calculation of learned values will be explained with reference to FIGS. 5 to 7. First, the steady state learning correction coefficient K B
Explaining from L k< C, this is the following equation (4^), (
4B). Note that, for convenience of explanation, "1-old" added before the symbol indicates the value at the previous calculation, and "new" indicates the value at the current calculation.

New KBLRC = Old KBLRC
=(i″FBYA/AFBYA) XLAMBDA (4B) where R is the learning correction coefficient shortage rate during steady state, and X is the rewriting rate during steady state.

Note that since there is a reciprocal relationship between the air-fuel ratio and the fuel-air ratio, the same control can be achieved even if the fuel-air ratio is used instead of the air-fuel ratio. For this reason, although the fuel-air ratio is adopted in this embodiment, it will be explained as the air-fuel ratio. Therefore, in the explanation of the following mathematical expressions, the fuel-air ratio may be meant.

These equations (4^) and (4B) are the effective pulse width T e
It is derived from the basic formula (3) which confirms the % formula %. In other words, during steady operation, the fuel delay correction amount K A i''HOS is almost equal to zero, so 1゛e (old 'l''e) calculated previously is equal to the old "I"
e= i' pX old KBLRCX old LAMBDA...
(4C). As a result, the actual air-fuel ratio A FB Y A is AFB
YA=Qa/old Te (where Qa is the air flow rate.

), so by substituting equation (4C) into this equation and rearranging for Qa, we get Qa=AFBYAXTPX old K B L RC x old LA
MBDA...(4D) is obtained.

Also, the target air-fuel ratio T F B Y A at the time of calculation this time is '
rFL3YA=(TpX!KBLRCX new LAMBDA
)/Qa (However, this time, assume that the air-fuel ratio has reached the target value and set the new LAMBDA=1.0.) Therefore, by transforming this formula, the new KBLRC becomes the new KBLRC=QaX T F B Y A/
Obtained from Tp-(4E).

If we delete Qa from equation (4D), <4E), we get new KBLRC = old KBLRCX (TFBYA x old LA).
MBDA/AFBYA) = Old KBL RCX R... (4F). That is, ) (= new KBLRC/old KBLRC, so R has a meaning as the deficiency rate of the learning correction coefficient. For example, if R = 1.0, it means that it matches the target value. After all, R = Rewrite the learning value so that it becomes 1゜0, line <, Therefore, the meaning of formula (4^) is (
l (-1) indicates the shortfall, and this shortfall X (from 0 to 1
This means that the old KBL RC will be rewritten by a constant between

Next, the transient learning correction coefficient K 131” 1., RC
is calculated using the following equations (411) and (41).

New K B TL RC = IL (K B ”rL RC
X(1+(R1″R1)XX KA r)...(41
1) RTR=(R(Tp+KATHO3)-Tp)/K
A T HOS ... (4I) However, R'l' R is the shortage rate of the transient learning correction coefficient,
XKAT is the rewriting rate during transition.

Here, equations (4u) and (41) are derived as follows. Since this time is a transition period, the shortage rate R is R = new Te
/ Old ``e... (4J),
However, new 'I'e = new 1'pX new K B L E < C + new K
A i” HOS X New K B i' L RC...
・(4K) Old 'I'e = Old 'rpX Old K A THOS Ten Old K A
l" HOS X Former K B i" L RC...
(4L).

Transforming these equations (4K) and (4L), we obtain the new K B
To organize T L RC, new KBTLRC=(R(old TpXIljKBLRC+
l [J K A T HOS X II K B
i” L RC)-New’1”pX New K B L
RC) /New K A'rHOS ・(4M).

Past? Insufficiency rate Ri' R of learning correction coefficient of li lL'7
is HT R = new K 131' L RC/old K B T
L RC Thealkara, by substituting the formula (4M) into this, R"I"■< is R'I' R= (RC1 day 1'
p X 1 day K B L RC + old K A
THOS x IEI K B T L RC) - New TpX New K B L RC) / (New K A T HOS
RC'I...(4N).

Here, 'l''9.K13LRC and K A THOS
Assuming that the new and old are almost equal, the following formula (411)
get.

RTR=(R(1 day l″pX1 day K B L
RC+ Ill K A T HOS X Old K B
T L RC) - old TpX old K B L [< C) / (old KA'rHO8X old KB1"LRC)... (4
P) Therefore, this formula (4P) is a strict formula for R′l″F<, and it is best to use this formula to find RT )?. However, in practical terms, the direction of learning is From the fact that they are the same and the relationship with the detection accuracy of the shortfall (RT)?-1), it is safe to set the old K B L RC and the old K B 1" L RC to approximately 1°0, and therefore it is not practical. The following equation (4Q) is obtained as the calculation equation for RT f< above.

R1' R= (k< (, old 'rp ten old K A 'r
'HOS)-formerly'l'p)/(formerly KATl-
i 0 S )...(4P) Calculate R and R1''R as above, and calculate these t:
It turns out that it is good to find the corresponding learning correction coefficient.

Learning simply means rewriting the learning map every learning cycle. The r56 diagram executes this.
This is the length blue routine shown in the PtS7 diagram, and the Pt55 diagram (A
) and the contents of steps 91 and 108 in FIG. 5(B). In other words, in steady state, the current N and Qc
Select the address corresponding to yL, read the data at that address as old learning data G TE I O, and use the following formula (4^-^) GTE I = 01'E I 0X
The new learning data G TE I is calculated as (1+(R-1)XX)...(4^-Δ), and this data G TE I is stored in the selected address (steps 121 to 124).

The same applies to the transient period (steps 131-134).
), old learning f-9G K A i' O Ka formula (411
-A) Rewritten to new learning data GKAT (step 133).

GKAT=GKATOX(1+(R1”R-1)XXK
AT) ... (4H-^) However, while the address selected during steady state is allocated between N and QCYL, the address selected during transient time is 'l'-
and Q CY L (Step 1)
21,131). This is because it is desired that the learned value be more dependent on the 6′ₛ control than on N during the transition.

Also, considering the change characteristics of LAMBDA, Rlt('i
” The average values R and RTR of R are adopted.In other words, the method of determining the average value itself is the same for l <l' R, so to summarize, from the start of learning, at each control cycle Start the 7-step counter to count the number of control cycles, and calculate the operation p and the shortfall rate l(
Or L<i"l< is integrated and the counter value (number of control cycles) CNi"R or CN i"f<1'
R is the judgment level GAKN or GAKN TF, respectively.
(In this case, the integrated value of ■< (ΣR) or R'l'
Rcn integrated value (ΣR1'R) is converted to counter value CN 'r
l< *or CN TRT l? (Steps 83 to 87, 90, 103)
~106, 107). In addition, the subscript "-1" after the symbol
” indicates the value from the previous calculation. This subscript will also be used in the following description.

Note that whether it is a steady state or a transient state can be determined by fII by comparing the adhesion speed VMF (described later), which indicates the amount of change in engine load, and the transient determination level LKAT (step 82). Since there are operating ranges (such as idling) in which feed bank control is stopped during steady state, 1-normal learning is not performed during any operation (step 89). Also,
Learning is also stopped at 1 o'clock when the engine is cold (1'w is below the f1 constant level TwGAK), since this is a special operating range in which the air-fuel ratio is enriched (step 88).

Furthermore, at the time of transition, the counter value CN TR1゛l
In some cases, the rate shifts to a steady state before it reaches the judgment level G The time learning table is rewritten (steps 83, 111 and 112).

According to this type of learning, since the learned value is calculated using the main operating variables that directly affect the air-fuel ratio during steady and transient times,
Sufficiently high air-fuel ratio accuracy can be obtained despite differences in fuel properties, environmental changes, and changes over time.

In addition, the learning speed is determined by the accuracy of learning during transient periods, but as the learning accuracy improves, the learning speed increases accordingly, and depending on the operating range, the air-fuel ratio that causes $ lean combustion may suddenly change. It is also possible to make the learning speed sufficiently follow the changes.

In addition, X K A r l: Tsui Teha K B T L
RCf> When the value increases, the acceleration performance can be improved by increasing the value to respond more sensitively than when the value decreases.

Next, to outline the entire system, in the routine shown in Figure 3, the effective injection pulse width 'e is Te=(1''pXKBL
RC 1K A 'l' HOS X K B i' L R
C) XLAMBDA... Calculated in (3) (step 58). Here, 'l''e is given by the sum of the basic pulse width 1゛p given by the injector air MLQAIN J l:N and the fuel delay correction amount KATHO5. Learning correction coefficients (KBLRC and KBTLRC) are introduced.

Tp corresponds to the amount of fuel supplied from the injector, but to obtain the standard air-fuel ratio 'r' FB Y A, 'rp and the mass air fit of the injector (QAI N J
r: In addition, FA is an intake air temperature correction coefficient. ) is to have a proportional relationship between them. Therefore, '1゛p is 1''l)''(QA I N J XK TA) x 1''
It is sufficient if it is FB Y A x K - (5) (step 53). However, K is a coefficient based on the characteristics of the injection valve. The units of these parameters are, for example, I'
p (msec), QA I N J (Cc), K i
” A (mg/cc), K (msec/mg)
It is. Here, how accurately QA I N J is calculated determines the air amount detection accuracy.

First, in FIG. 3(A), the flow path surface 81t (A'``VO'') of the throttle valve section is determined from the throttle valve 111 degrees 1'' 0 by searching a two-dimensional table (A'rVO table), etc. (step 41 ), However, when the bypass passage 23 is opened, an error occurs by the passage area Auy of this passage 23, so in this case, the total passage area A (=A'rV O+ A
o y) must be adopted (step 43
). Here, since ATVO is uniquely determined by i''V O, the influence of intake pulsation accompanying intermittent intake is eliminated.

Next, from the value obtained by dividing this A by N (A/N) and the shilling volume V CY L, the linearization flow rate Q u o is determined by searching a three-dimensional table (Quo table), and this is corrected using the correction factor KFLAT. The steady flow rate Q ++ (=
Q n o X K F LΔ′l゛) is calculated (step 45.47). Here, Linearai χ style Fi Q
o.

refers to the steady flow rate per unit number of intakes that passes through the throttle valve section. Also, K F' L A T 1. t Q e
Since YL deviates slightly from the target air-fuel ratio depending on the operating condition, this value is introduced to eliminate this.
It is found by searching the three-dimensional table (KFLAT table) (step 46).The reason for adopting A/N is that if only A is used, there will be an operating range where A changes suddenly depending on changes in N. This is because the resolution decreases in this region.

Next, Schilling air M Q c Y L is calculated using the following equation (6) using 2 (K2<1) as the delay coefficient (step 49).

(cy L =QoXK2 +Q CY L -I

For example, Q n shows the same change in response to a stepwise change in the throttle valve opening T VO , but the amount of air cannot flow in a stepwise manner to a cylinder located far away from the position where the throttle valve 21 is installed. QcyL responds with a first-order delay. Here, K2 is a value that determines how QcYL is made to follow Qll, and is obtained by searching a three-dimensional table (K2 table) or the like (step 48).

Note that this control routine is executed at predetermined intervals, so
The current value is determined by using the values of several times.

That is, in both formulas, the subscript "Q Cy"
-1'' means the previous value, and the current calculated value is sequentially determined using the calculated values several times. This allows for transient l
Even in this case, the Schilling air amount can be accurately calculated even though the air amount in the throttle valve section is being determined. Also,
During a transient period, a deviation occurs between Qo and QCYL, but this deviation is factored into the delay coefficient of 2, so that the amount of air to the shilling can be accurately determined even during a transient period.

Finally, the change in QCYL fl (Qcy and QcyL-1
) to T:1. Calculate ΔCM step 50),
By adding this addition amount ΔCM to QCYL, injector air +1QAINJ is obtained (step 51). That is, 60M”(QCYL QCyL-1)XKMAN I
O...(7) QAINJ:QCYLXvCYL+ΔCM...(8). Note that KMAN IO in step 50 is a manifold coefficient determined by the suction volume from the throttle valve to the intake port and the calculation cycle.

In the SPl system, the injection valve is installed farther away than the shilling, so QA I N J and QCYL do not necessarily match, and during a transient period, a predetermined amount (60M) corresponding to the pressure change in the intake pipe pressure changes. Misalignment occurs. Therefore, 8CM is introduced as a value that takes into account the deviation between the two in the transient OI, and has the meaning of the amount of air that compensates for the inside of the intake pipe as the pressure changes in the intake pipe. As a result, even on the basis of the amount of air flowing into the cylinder, the amount of air 'X flowing through the injection valve portion during a transient period can be accurately calculated. Calculating 'p' as the fuel amount using QAINJ means that it is possible to supply fuel according to the air amount at the fuel injection position, which allows the air-fuel mixture to be approximately at the target air-fuel ratio. Schilling can be supplied.

Next, the fuel delay correction fil K A i''HOS is calculated using the routine shown in FIG.
In the PI device, some of the injected fuel adheres to the inner wall surface of the intake pipe or intake boat before reaching the shilling, or the fuel Fi that is floating in the intake pipe without being inhaled (the amount of these fuels is referred to as [ The amount of adhesion is collectively referred to as 1.

) behavior determines fuel lag. Therefore, by quantifying the R amount, the fuel delay can be dealt with more efficiently.

Therefore, as shown in Figure 4, first 'I"w + QAIN
Using J and N, the equilibrium adhesion amount M! ') Find i (step 71). Here, the equilibrium adhesion amount refers to the adhesion amount that maintains an equilibrium state during steady operation. MF 1-1 prepares several 3-dimensional tables with QAINJ and N as parameters when 'I'- is constant, and combines searching these 3-dimensional tables with linear approximation interpolation calculation. It is determined by

Next, MFI-I and the predictive variable M of the adhesion amount at that time
Deposition rate VMF based on the deviation from F (MFH - MF)
is calculated (step 73). Here, the adhesion speed refers to the amount of adhesion per unit cycle. In other words, VMF is VM do = (MF'H - MF -1)XKMF
... (9) It is weighed at. When 1゛■0 changes stepwise,
MFH has been working hard to respond to this change, and has been working on the actual installation.
;The ti amount responds with a first-order lag. Here, the actual label, *
Although it is difficult to measure i, it is difficult to give a calculated value that approximates this first-order lag response, and for this reason, a deposition amount predictor variable MF is introduced. Therefore, during the transient change, the deviation from ME'II<ME'H-
Since it is understood that there is a deviation in the adhesion amount by MF), +i(VMI') is corrected based on this deviation as 'r;
). For example, at JJI + speed, a portion of the injected fuel is sucked into the A system until the adhesion amount is linked to the equilibrium state, and there is a shortage of 1 shilling, so increase the amount of that number of injection fuel. It is.

However, if the total amount of deviation (MFl-1-Ml・゛) is used as the correction amount, phenomena such as overshoot will occur immediately after the transition, so the actual correction amount may be the total amount multiplied by a predetermined ratio. For this reason, the ratio K is introduced.
There is a MF"c (step 72). Therefore, given the KMF, CIIT sets Ml" to the MPH.
It is possible to make the object follow as quickly or as slowly as possible. Then, the correction amount (VM
F ) is corrected this time, so the next predictor variable is the current predictor variable M F 1. : The value is obtained by adding VMF (step 74).

Finally, VMF is multiplied by the correction factor G HF to obtain the fuel delay correction amount K A i')! Find the OS (step 76)
).

GHF is introduced as a value that takes into account differences in fuel properties. Note that GHFQCYL is the deceleration correction factor, GHFl
” B Y A is the air-fuel ratio correction factor.

Calculating K A 'rHOS in this way, the intake system! Since the lI amount is directly grasped, air-fuel ratio errors due to fuel lag are eliminated even in SPI systems where fuel lag is a particular problem.

In this embodiment, a wide-range air-fuel ratio sensor is shown, but an oxygen sensor may be used to move the stoichiometric air-fuel ratio to the target air-fuel ratio, and the fuel injection device may also be
It is not limited to PI devices.

(Effects of the Invention) As explained above, the present invention provides an air-fuel ratio control device that calculates a learning value during a transient period based on a feedback correction amount calculated from the deviation between the actual air-fuel ratio and the I15 air-fuel ratio. Set the response delay time according to the engine temperature,
Since we decided to set the learning timing to be after the response delay time has elapsed after the transient time has been determined, we will learn the deviation in response to the time when the deviation from the target air-fuel ratio becomes larger during the transient period. This makes it possible to improve transient drivability and exhaust emissions, especially in a temperature range where the cooling water temperature is low.

[Brief explanation of drawings]

Figure 1 is a conceptual configuration diagram of this invention, Figure 2 is the sP system 21.
The system diagram of the 8-thread control of this invention applied to Figures 3 to 7 are flowcharts explaining the calculation contents according to this invention.
Figures m8 and m9 are waveform diagrams showing the air-fuel ratio sensor output for detailed explanation of the present invention. The m10 diagram is a diagram showing the characteristics of the fuel delay time with respect to the cooling water temperature 'l'-. 1...Basic injection amount calculation means, 2...Actual air-fuel ratio detection-
L stage, 3... Feedback correction amount calculation means, 4...
・Engine temperature detection means, 5A response delay opening setting means, 6.
...Transient time determination means, 7...Learned value performance is a means, 8...
・Transient fuel injection amount calculating means, 21... Intake throttle valve, 22... Intake passage, 23... Bypass passage, 24
... Fuel injection valve, 25 ... Throttle valve opening sensor, 32
... Crank angle sensor, 33 ... Water temperature sensor, 34
...Air-fuel ratio sensor, 35...Control unit. Patent Applicant l] San Jidosha Co., Ltd. Agent Patent Attorney Go l 俟 Masaki, Patsu ¥,,,':'1゛1 (1 other person) tu-' ”-j Fig. 8 Fig. 9 to Time No. Figure 10

Claims (1)

[Claims] means for calculating a basic injection amount according to an operating condition signal, means for detecting an actual air-fuel ratio, means for calculating a feedback correction amount based on a deviation between the detected actual air-fuel ratio and a target air-fuel ratio, and an engine. means for setting a response delay time according to the temperature; means for determining whether or not a transient state is occurring; An internal combustion engine comprising means for calculating a learned value, and means for correcting the basic injection amount using the calculated learning value and the feedback correction amount to calculate a transient fuel injection amount. Air-fuel ratio control device.