JPH0323332A - Air-fuel ratio control device for internal combustion engine - Google Patents

Abstract

PURPOSE:To perform accurate control of an air-fuel ratio by a method wherein based on at least two of the ratio between lean and rich signal generating times or the ratio between response times and amplitude of the output voltage of an oxygen concentration detector, deterioration of the oxygen concentration detector is accurately decided. CONSTITUTION:Based on an output signal from an oxygen concentration detector 17 provided in an engine exhaust passage A, a fuel feed amount is controlled by a means B so that an air-fuel ratio of fuel-air mixture is adjusted to a target value. A fuel feed amount is fluctuated at an estimated period, and an air-fuel ratio is fluctuated alternately to the rich side and the lean side based on a target air-fuel ratio by means of a means C. At least two of the ratio between a lean and a rich signal generating time or the ratio between response times and amplitude of the output voltage of an oxygen concentration detector 17 when an air-fuel ratio is fluctuated are measured by a means D. According to each measuring result, by using a fuzzy inference, a fuel feed amount is regulated by a means E so that an air-fuel ratio under control is adjusted to a target value.

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.

[Conventional technology]

The oxygen concentration detector (which has traditionally been placed in the engine exhaust passage)
2. Description of the Related Art An internal combustion engine is known in which the amount of fuel supplied is feedback-controlled based on an output signal from a sensor (hereinafter referred to as 02 sensor) so that the air-fuel ratio becomes a target air-fuel ratio. However, in such an internal combustion engine, if the 02 sensor deteriorates due to long-term use, the air-fuel ratio will deviate from the target air-fuel ratio. When the O2 sensor deteriorates, the time ratio between lean signal generation period and Lynch signal generation period of the Ox sensor output signal changes, or the response time of the lean signal and rich signal changes, or the output signal of the 02 sensor changes. Since the level changes, it can be determined from these changes that the O2 sensor has deteriorated.

Therefore, the 02 sensor deterioration detection device measures the generation time ratio between the lean signal generation time and the rich signal generation time of the output signal of the 02 sensor during feedback system 1 wholesale, and determines the deterioration of the 02 sensor from this measurement result. is publicly known (
Special Publication No. 57-32773 and Special Publication No. 57-32
(See Publication No. 774).

Furthermore, there is a known 02 sensor performance evaluation device that determines the deterioration of the 02 sensor from the ratio of the lean signal generation time to the rich time generation time when the air-fuel ratio is changed in a triangular waveform (particularly (Refer to Japanese Patent Publication No. 1983-124248).

In addition, air-fuel ratio control is performed in which the deterioration of the 02 sensor is determined from the maximum and minimum values of the output signal level of the 02 sensor during feedback control, and the fuel injection amount is feedback-controlled so that the air-fuel ratio becomes the target air-fuel ratio. The method is known (see JP-A-58-72646).

[Problem to be solved by the invention] However, when the Ot sensor deteriorates, O! It is impossible to predict whether the generation time ratio between the lean signal generation period and the rich signal generation period of the sensor will change, whether the response time of the lean signal and rich signal will change, or whether the output signal level of the 02 sensor will change.

Therefore, even if one of these changes is measured as described above, there is a problem in that it is difficult to accurately determine the deterioration of the 02 sensor.

[Means to solve the problem]

In order to solve the above problems, according to the present invention, the air-fuel ratio is set to the target based on the output signal of the oxygen concentration detector 17 disposed in the engine exhaust passage A, as shown in the block diagram of the invention in FIG. a fuel supply control means B that performs feedback control on the amount of fuel supplied so as to maintain the air-fuel ratio; an air-fuel ratio varying means C that alternately varies the air-fuel ratio to the rich side and lean side;
The output signal of the oxygen concentration detector 17 when the air-fuel ratio is varied by the air-fuel ratio variation means C. The generation time ratio of the lean signal generation time and the rich signal generation time, the response time ratio of the lean signal and the rich signal, and the oxygen concentration A measuring means D for measuring at least two of the amplitudes of the output voltage of the detector 17, and fuzzy inference from at least two of the generation time ratio, response time ratio, and amplitude during feedback control. It is equipped with a fuel supply amount increasing/decreasing means E for increasing/decreasing the fuel supply amount so that the air-fuel ratio becomes the target air-fuel ratio.

[Function] Since the deterioration of the 02 sensor is determined based on at least two parameters that represent the deterioration of the 02 sensor, the deterioration of the 02 sensor can be determined more accurately, and the air-fuel ratio can be more accurately adjusted to the target air-fuel ratio. The fuel ratio can be controlled.

〔Example〕

Referring to FIG. 2, 1 is a cylinder block, 2 is a piston, 3 is a cylinder head, 4 is a combustion chamber, ? is the intake valve,
6 indicates an intake boat, 7 indicates an exhaust valve, 8 indicates an exhaust boat, and 9 indicates a spark plug.The intake boat 6 indicates a corresponding intake pipe 10.
are connected to the surge tank 11 through each intake pipe 10.
A fuel injection valve 12 is disposed inside each of the fuel injection valves. fuel injection valve 1
The fuel injection amount from 2 is controlled based on the output signal of the electronic control unit 20. Surge tank l1 is intake duct 1
3 to the air cleaner 14, and the intake duct 1
A throttle valve 15 is disposed within the throttle valve 3 . On the other hand, the exhaust boat 8 is connected to an exhaust manifold 16, and within this exhaust boat 8, a 0■ sensor 17 is arranged.

The electronic control unit 20 consists of a digital computer with ROMs interconnected by a bidirectional bus 21.
(read only memory) 22, RAM (random access memory) 23, CPU (microprocessor) 2
4, an input port 25 and an output port 26.

The air flow meter 14 generates an output voltage proportional to the amount of intake air, and this output voltage is input to the human power port 25 via the AD converter 27. A throttle switch l8 is attached to the throttle valve l5 to detect that the throttle valve 15 is at an idling opening, and the output signal of this throttle switch l is input manually to the input boat 25.

Further, the output signal of the 02 sensor 17 is manually inputted to the manual port 25 via the AD converter 28 . Furthermore, 25 human-powered boats
A rotation speed sensor 19 is connected to generate an output signal representative of the engine rotation speed. On the other hand, the output port 26 is connected to the fuel injection valve l2 via a drive circuit 29. In the embodiment shown in FIG. 2, the fuel injection time TAU from the fuel injection valve l2 is basically calculated based on the following equation.

TAll = TP・(1+KOS)・(1+α)・F
AP-F where TP: Basic fuel injection time KOS: Correction coefficient α due to deterioration of 02 sensor 17: Variation coefficient for varying the air-fuel ratio? AF: Feedback correction coefficient F: Other correction coefficients based on engine cooling water temperature, etc. The fuel injection time TAU is controlled by the feedback correction coefficient FAF, which changes based on the output signal of the 02 sensor t7, so that the air-fuel ratio becomes the target air-fuel ratio. Hereinafter, in order to make it easier to understand the present invention, an example will be described in which the target air-fuel ratio is set to the stoichiometric air-fuel ratio.

0.0 when the air-fuel mixture is greater than the stoichiometric air-fuel ratio, that is, when it is lean. It generates an output voltage of about 1 volt, and when the air-fuel mixture is smaller than the stoichiometric air-fuel ratio, that is, when it is rich, it generates an output voltage of about 0.9 volt.

FIG. 3 shows a control routine for the feedback correction coefficient FAF, and this routine is executed by interrupts at regular intervals. Referring to FIG. 3, first, in step 30, 0
■The output voltage of the sensor 17■ is a predetermined reference value VO
, for example, whether it is higher than 0.45 volts. When v>v0, the process advances to step 31, where it is determined whether or not V was less than Vo in the previous interrupt routine.

V<V. When this happens, the process proceeds to step 32, where a predetermined skip value S is subtracted from FAF. On the other hand, if it is determined in the previous interrupt routine that V>V0, the process proceeds to step 33, where a predetermined integral value K (K(S)) is subtracted from FAF.

On the other hand, if it is determined in step 30 that V<V0, the process proceeds to step 34, where it is determined whether or not V>v0 in the previous interrupt routine. V>V0
If so, the process proceeds to step 35, where the skip value S is added to FAF, and if V,<V, the process proceeds to step 36, where FAF is added.
Integral value K is added to AF.

Therefore, as shown in FIG. 4(A), when the air-fuel mixture becomes rich and V>V0, that is, when the 02 sensor 17 generates a rich signal, FAF is suddenly skipped {IIS is decreased, and then the integral value K As a result, it is gradually reduced. As a result, the fuel injection time TAU becomes shorter, and the mixture gradually changes to the lean side. On the other hand, when the mixture becomes low and reaches v0, that is, 02 sensor l
7 generates a lean signal, the FAF is suddenly increased by the skip value S, and then gradually increased by the integral value K. As a result, the fuel injection time TAU becomes longer, and the air-fuel mixture gradually changes toward the ridge side. In this way, the air-fuel mixture is maintained at approximately the stoichiometric air-fuel ratio, and at this time, the feedback correction coefficient FAF is 1. It fluctuates around 0.

However, when the 02 sensor l7 deteriorates, the rich signal generation time T, . and the lean signal generation time TI, for example, the duty ratio (T, /
(T, +Tz )) deviates from the standard duty ratio shown in Figure 4 (A), or the response time T7 of the lean signal and the response time Tr of the rich signal differ as shown in Figure 4 (C). If the response time ratio (tf/t,) deviates from the standard response time ratio shown in FIG. 4(A), or if the response time ratio (tf/t,)
As shown in D), is the output signal V of the 02 sensor ■? The difference between the maximum value and the minimum value, that is, the amplitude ΔV. may deviate from the standard amplitude shown in FIG. 4(A). 0) If the sensor 17 deteriorates, one, two, or all of these phenomena will occur, but it is impossible to predict which one or which two will occur. In addition, when feedback control is performed using the deteriorated 02 sensor l7, as shown in Figure 4 (B), if the duty ratio (Tr/ (T, +Tt)) becomes larger than the standard duty ratio, the mixture is detected. If it becomes smaller, the mixture becomes richer, and as shown in Figure 4 (C), the response time ratio (T7/T
, ) is larger than the standard response time ratio, the mixture is on the lean side, and when it is smaller, the mixture is on the rich side, and as shown in Figure 4 (D), the amplitude Δ■1 is the standard. Although it is qualitatively known from experience that if the amplitude is smaller than the normal amplitude, the air-fuel mixture will be on the lean side, and if it is large, the air-fuel mixture will be on the rich side, but quantitatively it is not precisely known. Also, if two or all of these phenomena occur simultaneously, for example, if the duty ratio becomes larger than the standard duty ratio and at the same time the response time ratio becomes larger than the standard response time ratio. Although it is qualitatively known from experience that the air-fuel mixture will be on the lean side, quantitatively it is not precisely known. Therefore, in the embodiment shown in Fig. 1, the air-fuel mixture is alternately varied between the lean side and the rich side at a constant cycle during idling operation, and the deviation between the duty ratio at this time and the standard duty ratio, and the difference between the duty ratio at this time and the standard duty ratio are Find the deviation between the response time ratio and the standard response time ratio, and the deviation between the amplitude at this time and the standard amplitude, and use fuzzy inference from these deviations to determine whether the mixture will become lean during feedback control. It is determined whether the fuel is rich or not, and based on the result of this determination, the fuel injection amount is controlled so that the air-fuel ratio under feedback control becomes the stoichiometric air-fuel ratio.

Next, a routine for calculating the fuel injection time TAU will be explained with reference to FIG. This routine is executed, for example, every 360 degrees of crank angle.

Referring to FIG. 5, first, in step 40, the basic fuel injection time TP is calculated from the output signal of the air flow meter 14 and the engine speed. Then step 4l
For example, it is determined whether the engine is in an idling state based on the output signal of the throttle switch 18 and the engine speed. When it is not in the idling operation state, the process proceeds to step 42 where the coefficient of variation α is set to zero, and then step 4
Proceeding to step 3, the fuel injection time TAU is calculated. In this case, the fuel injection time TAU is the feedback correction coefficient FA.
Feedback control is performed based on F. On the other hand, if it is not in the idling state, the process proceeds to step 44 and the process proceeds to step 02.
The correction coefficient KOS due to the deterioration of the sensor l7 is set to zero.

Next, the process proceeds to step 45 to calculate the feedback correction coefficient F.
AF is fixed at 1.0, so feedback control is stopped at this time. The process then proceeds to step 43.

At this time, the coefficient of variation α is controlled by a routine to be described later so that the air-fuel ratio alternates between lean and rich.

Next, the air-fuel ratio control routine shown in FIGS. 6 to 8 will be explained with reference to the time chart shown in FIG. 9.

This air-fuel ratio control routine is executed by interrupts at regular intervals. Referring to FIGS. 6 to 8, first step 50
In this step, it is determined whether or not the engine is in an idling state based on, for example, the throttle switch 18 and the engine speed. If the vehicle is in the idling state, the process proceeds to step 5l, where it is determined whether a certain period of time has elapsed since the vehicle entered the idling state. If a certain period of time has elapsed, the process proceeds to step 52, where it is determined whether the measurement completion flag has been set. At this time, the measurement completion flag has been reset, so the process advances to step 53, and the counter C
The count value of is incremented by 1. That is, as shown in FIG. 9, when a certain period of time has elapsed at t0 after the engine enters the idling state, the counter C starts counting up.

Next, in step 54, it is determined whether the count value of count C has become larger than a predetermined constant value C0. C<C. If so, the process proceeds to step 55 where the coefficient of variation α is set to +α, and the process proceeds to step 57. On the other hand, C>C
．． If so, the process proceeds to step 56, where the coefficient of variation α is set to -α, and the process proceeds to step 57. In step 57, the counter C
It is determined whether the count value of has reached 2Co or not.

C=2C. Otherwise, jump to step 60. On the other hand, C=2C. When this happens, the process advances to step 58 where the counter C is cleared. Next, in step 59, the number of measurements N is incremented by 1, and the process proceeds to step 60. In this way, when C<C6, α is set to +α, and Co IC
During '-2C6, α is -α, so as can be seen from the routine shown in Figure 5, the fuel injection time TAU increases and decreases at a constant cycle, and the air-fuel mixture changes between lean and rich at a constant cycle. It will be repeated. Therefore, at this time, as shown in FIG. 9, the output voltage V of the 02 sensor 7 changes as the lean and rich cycles are repeated.

In step 60, it is determined whether the measurement start flag is set. At this time, since the measurement start flag has been reset, the process proceeds to step 61, where it is determined whether the number of measurements N is l. After L0 in FIG. 9, the count value of counter C is 2C. Since N=O until this point is reached, the processing routine is completed during this time. Then C=2C. When N=1, the process proceeds to step 62 and the 02 sensor l
Wait until the output signal 7 becomes higher than the reference voltage V0. When V≧v0, the process advances to step 63 and a measurement start flag is set. Therefore, in the next processing cycle, the process proceeds from step 60 to step 64.

In step 64, it is determined whether the output voltage (2) of the 02 sensor 17 is higher than the reference voltage V0.

■〉■. If so, the process advances to step 65 and V, <V in the previous processing cycle. It is determined whether or not it was. That is, it is determined whether ■ has exceeded v0 between the previous processing cycle and the current processing cycle. When ■ exceeds V0 between the previous processing cycle and the current processing cycle, the process proceeds to step 66, where it is determined whether the number of measurements N is 1 or not. When N=1, the process advances to step 67.

Also, in the previous processing cycle, when V > Vo, the process proceeds to step 67. In step 67, the count value of the counter Cl is incremented by 1, and then the process proceeds to step 69. Therefore, as shown in FIG. 9, the counter Cl is ■>■. It is counted up when
Therefore, it can be seen that the final count value of the counter Cl field represents the rich signal generation time T. On the other hand, V<
V. When this happens, the process proceeds from step 64 to step 68, where the count value of counter C2 is incremented by 1.
The process then proceeds to step 69. Therefore, as shown in FIG. 9, the counter C2 is V<V. It is understood that the final count value of the counter C2 represents the lean signal generation time TI.

In step 69, from the output signal V of the 02 sensor l7 in the current processing cycle to 0 in the previous processing cycle.
The output signal V of the two sensors 17 is subtracted, and the subtraction result (
It is determined whether or not (1-2) is larger than a predetermined constant value ΔV.

That is, it is determined whether or not the output signal of the Ot sensor 17 is in a transient state in which it changes from a lean signal to a rich signal.

When V-V,<Δ■, it is determined that there is no transient state in which the lean signal changes to the rich signal, and the process jumps to step 71. On the other hand, when V-Vl > Δ■, it is determined that it is a transient state in which the lean signal changes to a rich signal, and the process proceeds to step 70, where the count value of the counter C3 is incremented by 1, and then the process proceeds to step 71. ．． Therefore, the count value of the counter C3 is the response time t of the 02 sensor 17 when changing from a lean signal to a rich signal,
It can be seen that it represents

In step 7l, the output signal V of the 02 sensor 17 in the current processing cycle is changed to 0 in the previous processing cycle.
The output signal VI of the second sensor l7 is subtracted, and the subtraction result (
It is determined whether or not VV+) is smaller than a predetermined constant value -ΔV.

In other words, is the output signal of the 02 sensor l7 a rich signal? It is determined whether the signal is in a transient state where it changes to a lean signal.
When V-V>1 ΔV, it is determined that there is no transient state in which the rich signal changes to the lean signal, and the process jumps to step 73. On the other hand, when V-V<1 ΔV, it is determined that the state is a transitional state in which the rich signal changes to a lean signal, and the process proceeds to step 72, where the count value of the counter C4 is incremented by 1, and then the process proceeds to step 73. move on. Therefore, the count value of the counter C4 is the response time t of the Ot sensor 17 when changing from a rich signal to a lean signal.
It can be seen that it represents l.

In step 73, the output voltage V of Ot sensor l7 is Vs+
It is determined whether or not it is larger than aκ, and if V>Vmaκ, the process proceeds to step 74 and V is set to Vtaax.
Therefore, V tax is O! It can be seen that this represents the maximum value of the output voltage ■ of the sensor 17. Similarly step 75
Then, it is determined whether the output voltage V of the 0■ sensor 17 is smaller than Vsin, and if it is less than VIlin, the process proceeds to step 76, where V is set to Vin. Therefore, V+sin
HaO! It can be seen that this represents the minimum value of the output voltage of sensor l7.

When the count value of the counter C reaches 2Co again, N==2 is set in step 59. Then ■ becomes ■ between the previous processing cycle and the current processing cycle. If it exceeds step 64 to step 65. The process proceeds to step 77 via step 66. In step 77, CI/(C
+ +Cz ), that is, the duty ratio Tr / (T
- +Tt) to ΣD, and the addition result is ΣD
Let it be r. Next, in step 7B, Ca/Cs, that is, the response time ratio tz/tr is added to ΣG, and the addition result is added to ΣG, . shall be.

Next, in step 79, the amplitude (Vmax Vain
) is added to Σ■, and the addition result is set as ΣV1.

Next, in step 80, each counter C1 . CzCx.
Clear Ca and set V coax and Vain to zero. Next, in step 81, it is determined whether the number of measurements N has reached a predetermined number N0, N=N. When this happens, the process proceeds to step 82, where a measurement completion flag is set, and then, in step 83, the number of measurements N is set to zero. That is, N=N. When it becomes, the duty ratio T,
/ (T, +TI), response time ratio tI/L, . and the amplitude (Vtaax-Vmin) have already been measured (N.-1) times, so ΣD1 . ΣG. ．．
, Σv1 are the duty ratio T r / (T r
+ T z ), response time ratio L z / tr and amplitude (Vn+ax −VIIIin ) of (N.−
1) It represents the sum of the items. When the measurement completion flag is set, in the next processing cycle, the process proceeds from step 52 to step 84, where the average value of the duty ratios ΣD,/(N.
-1) is the reference duty ratio, that is, the duty ratio when the 02 sensor 17 has not deteriorated. is subtracted, and the result of the subtraction is taken as the duty ratio deviation ΔD. Next, in step 85, the average value ΣG of the response time ratio
,/(NO-1) to the reference response time ratio, that is, 02
Response time ratio G1 when sensor l7 is not deteriorated. is subtracted, and the result of the subtraction is the deviation ΔG of the response time ratio Gr0.
It is said to be 7.

Next, in step 86, the reference amplitude, that is, the amplitude V when the 02 sensor 17 has not deteriorated. From the average value of amplitude ΣV. /(N6-1) is subtracted, and the result of the subtraction is taken as the amplitude deviation Δv1. In this way, the deviation ΔD of the duty ratio from the reference value, . , response time deviation Δ
C'r and the amplitude deviation Δ■1 are determined.

Then step 87. In 88, these deviations ΔD,
ΔG,, ΔV. Based on this, a correction coefficient KOS due to deterioration of the 02 sensor 17 is determined by fuzzy reasoning. As can be seen from Figure 5, as KOS increases, fuel injection time T
AU becomes longer. Therefore, when the air-fuel mixture becomes lean due to deterioration of the 02 sensor 17, the air-fuel ratio can be brought closer to the stoichiometric air-fuel ratio by increasing KOS.
It can be seen that when the air-fuel mixture becomes rich due to deterioration of the 02 sensor 17, the air-fuel ratio can be brought closer to the stoichiometric air-fuel ratio by decreasing KOS.

Next, fuzzy inference will be explained with reference to FIGS. 10 to 13.

First, let's take the duty ratio deviation ΔDr as an example. From experience, it is qualitatively clear that when the duty ratio deviation ΔD increases a little, the mixture becomes a little lean, and when ΔDr becomes considerably large, the mixture becomes very lean. I know.

Therefore, in order to maintain the air-fuel ratio at the stoichiometric air-fuel ratio even when the Ot sensor 17 has deteriorated, the KOS should be increased slightly when ΔD becomes a little large, and the KOS should be made large when ΔD becomes large. It would be better to make it bigger. However, it is not known quantitatively how much KOS should be increased when ΔD increases. Therefore, first of all, a membership function is used to quantitatively express to some extent that ΔD is a little large or quite large.

Figure 10 (A) shows an example of the membership function given to the duty ratio deviation ΔD. In FIG. 10(A), the horizontal axis shows ΔD, and the vertical axis shows the membership value (0 to 1). Also, Figure 10 (
Each character shown in A) is called a fuzzy label and has the following meaning. P B (Positive Big) Positive and quite large PM (Positive Medium)
Positive and somewhat large P S (Positive Sm
all) Positive and slightly larger ZO (Zero)
The reference value NS (Negative S
tsa) Negative and slightly larger NM (Negative
Mediaua+) Negative and somewhat larger NB (Ne
gative Big) Negative and quite large sokuchi,
ΔD,=0.1 is intuitively perceived as ΔD2 being a little large, and therefore 0. 1 is supported by PS,
Also, ΔD,=0.2 is intuitively perceived as ΔD, being somewhat large, and therefore 0. 2 is supported by PM. ΔDr is 0. When the membership value deviates from 2, it becomes smaller than 1, which means that even if it is somewhat large, the degree to which it can be perceived intuitively becomes smaller. For example, as shown in FIG. 10(B), ΔDr is 0. If it is about 12, the membership value is 0. This means that the degree to which ΔD is intuitively perceived as being slightly large (ps) is about 80%. In other words, the fitness for a slightly larger ΔD (ps) is 80%. In this case, ΔD,
．． The area shown by the hatching below the line drawn horizontally from the intersection of the
S) represents the possibility of being perceived intuitively. Therefore, as shown in Figure 10 (C), if ΔD = 0.1, there is a 100% possibility that ΔD is intuitively perceived as being slightly larger (PS). On the other hand, as shown in Figure 10 (CD), when ΔD, straddles the membership functions representing PS and PM, there is a possibility that ΔD1 can be intuitively perceived as slightly larger (PS) by 5.
0 percent, and ΔD is somewhat large (PM
), the probability of being able to perceive it intuitively is about 30 percent. Other deviations ΔGr+ΔV. Membership functions are similarly given to the correction coefficient KOS, and these are the 11th
As shown in the figure. FIG. 11(A) shows the membership function for the duty ratio deviation ΔD1, which has already been explained using FIG. FIG. 11(B) shows the number of memberships relative to the response time ratio deviation ΔG1. As can be seen from FIG. 11(B), for example, if ΔG=0.2, there is a 100% possibility that ΔGr is intuitively perceived as being slightly larger (PS).

FIG. 11(C) shows the amplitude deviation ΔV. shows the membership function for . As can be seen from FIG. If (C), for example, ΔV. = 0.2, Δ■1 is somewhat larger (
There is a 100% possibility that it can be perceived intuitively as PM). FIG. 11(D) shows the member symp function for the correction coefficient KOS due to deterioration of the 02 sensor 17. 11th
As can be seen from Figure (D), setting KOS to about 0.04 intuitively means slightly increasing KOS (PS), and setting KOS to about 0.08 intuitively means KOS This means increasing (PM) somewhat.

By the way, for example, ΔD is almost at the standard value (ΔD,=Z
○), ΔG, is a little larger (ΔGr=PS) or Δ■
1 is a little larger (Δv.=PS), if KOS is made a little larger < (KOS=PS), O! It is intuitively known from experience that even if the sensor l7 deteriorates, the air-fuel ratio will remain approximately at the stoichiometric air-fuel ratio. ΔDr+Δ
The relationship between G1, ΔV1 and KOS has been established as a rule based on experience, and this rule is shown in FIG.

That is, as shown in FIG. 12(A), the X, Y, and Z axes have deviations ΔD1 and ΔGrrΔV1, respectively, and the deviations ΔD,
, ΔGr. PB, PM, PS of ΔVs. Z○,N
S. KOS is PB against NM and NB. PM, PS, Z
○． It is determined whether it corresponds to NS, NM, or NB.

FIG. 12(B) shows Δv. = indicates the ZOO surface, and the first
Figure 2 (C) shows the ΔV,=PSO plane. For example, in FIG. 12(B), ΔV. =ZO, ΔD, =PS, Δ
If G,=PS, then KOS=PS. These ΔD,
In principle, the relationship between ΔG,, ΔV1 and KOS is ΔD,,
ΔGrlΔV, all PB, PM. P.S., ZO. N
Although the combination is determined for the combination of S, NM, and NB, the combination portion that has no possibility is set as a blank space. Note that Figure 12 (B.) and (C) show typical KOS fuzzy labels, and the blank areas do not necessarily indicate KO.
This does not mean that a fuzzy label for S has not been determined.

Next, referring to FIG. 13, ΔD, ΔG. ．． , ΔV, will be described below. Note that FIG. 13 shows ΔDr, ΔG1, ΔV. is shown in Figure 13 (A). (B
) and (C) are taken as an example.

Figure 13 (A), (B). As shown in (C), ΔD, belongs only to zO, and ΔG, belongs to ZO and P
ΔV, belongs to both ZO and PS. Therefore, the following four combinations are considered, and the possible KOS for each combination is determined.

(i) ΔD,=Z○ and ΔG,=ZO and ΔV. =Z○ (ii) ΔD, = ZO and ΔG, = ZO and Δv. = ps (iii) ΔD, = ZO and ΔG, = PS and ΔV. =Z
O (iv) ΔD,=ZO and ΔG,=PS Δv. = ps Next, these (i) to (iv) will be explained in order.

(i) When ΔD,=ZO and ΔG,=ZO and ΔV,=0 In this case, from the relationship shown in Figure 12(B), KOS=Z
○, and in this case, Fig. 13 (A), (B). (C
), KOS is obtained by the method shown in (D).

That is, first of all, the possibility that ΔD,=ZO, ΔC,=
Find the possibility that ZO and the possibility that ΔV,=ZO. This possibility is shown in Figure 13 (A), (B), (C).
Each figure is indicated by hatching.

In this example, as can be seen from Figure 13, the probability that ΔD1 is ZO is about 80%, the probability that ΔG, is ZO is about 50%, and Δvll is ZO.
The probability that it is ○ is about 20%. In this case, the possibility that ΔDI"I ΔG, is ZO is relatively high, but the possibility that ΔV. is ZO is low, and therefore, from the overall perspective, the possibility that KOS will be ZO is low. This is explained, for example, above. In the example of , for example, when ΔV. is almost zero, it is clear that there is almost no possibility that KOS will take ZO.In other words, the possibility that KOS will take ZO is ΔDr+ ΔG.., ΔV, is ZO. It will match the lowest possibility. Therefore, the 13th
As shown in Figure 13 (D), the possibility that KOS takes ZO is considered to be equal to the possibility that ΔV1 is Z○, so the line 1 drawn at the same position as in Figure 13 (C). The region below is the possibility that KOS can take ZO.

(ii) ΔDW. =ZOteΔC, =Z○andΔV.
= PS In this case, from the relationship shown in Figure 12 (C), KOS = P
S, and in this case, Fig. 13 (E), (F), (G
), KOS is calculated as shown in (H). In this case, ΔG, is least likely to be ZO, so the first
The area below line l2 drawn at the same position as in Figure 3 (F) is KO.
This is the possibility that S can take PS.

(Mountain) When ΔD,=Z○ and ΔG,=PS and Δ■1=ZO In this case, from the relationship shown in Figure 12 (B), KOS=P
S, and in this case, Fig. 13 (I), (J), (K
). KOS is determined as shown in (L). In this case, the possibility that ΔV, is Z○ is the lowest, so the first
The area below line l3 drawn at the same position as in Figure 3 (K) is KO.
This is the possibility that S can take PS.

(iv) ΔD,=ZO and ΔG,=PS and ΔV,=
In the case of PS In this case, from the relationship shown in Figure 12 (C), KOS=P
S, and in this case, Fig. 13 (M), (N) . (0
). KOS is calculated as shown in (P). In this case, the possibility that ΔGr is PS is the lowest, so the first
The lower eMM of the line 7!4 drawn at the same position as in Figure 3 (N) is the possibility that KOS can take PS.

Next, as shown in Figure 13 (Q), the maximum possibilities for each fuzzy label of KOS are superimposed and the KO
Create the possibility distribution of S. In this case, for ZO, the maximum is shown in Fig. 13 (D), and for PS, the maximum is shown in Fig. 13 (H).
The shape will be as shown in Q). The reason for superimposing the maximum possibility for each fuzzy label is as follows. For example, if we consider KOS=PS, K as shown in Figure 13 (H), (L), and (P)
The possibility that an OS can take a PS represents the suitability of each rule. In other words, it is most likely to be shown in Figure 13 (H), but this is also shown in Figure 13 (E).
, (F), (G), (}{), that is, ΔD,=ZO, ΔG,−ZO and ΔV. This means that the rule that KOS=PS when =PS is most suitable. This means that if the probability that ΔDr becomes ZO, the probability that ΔG1 becomes PS, and the probability that ΔV becomes Z○ are each 100%, then in other words, if the rules are completely compatible, KOS is This is obvious if you consider that PS is always required. Therefore, as mentioned above, KO
If the maximum possibilities for each fuzzy label of S are superimposed, this represents the probability distribution of KOS. In the probability distribution shown in Figure 13 (Q), PS
The possibility that it is ZO is greater than the possibility that it is ZO, so KOS is closer to PS than ZO. In this case, the value of KOS becomes the center of the possibility distribution, that is, the center of gravity of the hatched area in FIG. 13 (Q). FIGS. 13(R) and (S) show a method for determining the center of gravity of the possibility distribution. That is, first, as shown in FIG. 13(R), the outline of the hunting region in FIG. 13(Q) is expressed in the form of a function F (KOS). Then the 13th
As shown in Figure (S), G0 obtained in this way
becomes the correction coefficient KOC. Each membership function shown in Figure 11 and Figure 12 (A
) are stored in the ROM 22 in advance. Therefore, in step 87 of FIG. 6, each deviation ΔDl'l ΔGr+ is calculated from these membership functions and rules.
Calculate the function F (KOS) based on Δ■,. Next, in step 88, the center of gravity is determined from the number F (KOS) and set as KOS. Next, in step 89, the counter C
are cleared, ΣDr, ΣGr. Σ■1 is assumed to be zero. Next, in step 90, the measurement completion flag is reset, and then in step 91, the measurement start flag is reset.

On the other hand, if it is not in the idling state or if a certain period of time has not elapsed since the idling state, the process advances to step 92 and each counter C. CI, C2, C3, C
4 is cleared, VmaxVmin + ΣDr,
ΣGr, ΣV, , N are assumed to be zero. Next, in step 91, the measurement start flag is reset. As described above, the duty ratio deviation ΔD, the response time ratio deviation ΔG1, and the amplitude deviation ΔV. is 02 sensor 17
of two of these parameters, e.g. ΔD. ．． Deterioration of the 02 sensor l7 can be known quite accurately by only ΔG and ΔG.

[Brief explanation of drawings]

Fig. 1 is a diagram of the structure of the invention, Fig. 2 is an overall view of the internal combustion engine, Fig. 3 is a flowchart for controlling the feedback correction coefficient, Fig. 4 is a diagram showing changes in the output voltage of the 02 sensor, etc. Figure 5 is a flowchart for calculating fuel injection time, Figures 6 to 8 are flowcharts for controlling air-fuel ratio, Figure 9 is a time chart, and Figures 10 and 11 are membership functions. 12 is a diagram showing the rules, and FIG. 13 is a diagram for explaining the fuzzy inference method. 5...Intake valve, 7...Exhaust valve, 12...
-Fuel injection valve, 15...throttle valve, 17...02 sensor.

Claims (1)

[Claims] a fuel supply control means for feedback controlling the amount of fuel supplied so that the air-fuel ratio becomes a target air-fuel ratio based on an output signal of an oxygen concentration detector disposed in an engine exhaust passage; When the air-fuel ratio is varied by the air-fuel ratio variation means, the air-fuel ratio is varied by the air-fuel ratio variation means for alternating the air-fuel ratio to the rich side and the lean side with respect to the target air-fuel ratio by varying the supply amount at a predetermined period. At least two of the following: the ratio of the lean signal generation time to the rich signal generation time of the output signal of the oxygen concentration detector, the response time ratio of the lean signal to the rich signal, and the amplitude of the output voltage of the oxygen concentration detector. Increase or decrease the amount of fuel supplied so that the air-fuel ratio during the feedback control becomes the target air-fuel ratio using a measuring means to measure and fuzzy inference from at least two of the occurrence time ratio, response time ratio, and amplitude. An air-fuel ratio control device for an internal combustion engine, comprising means for increasing and decreasing fuel supply amount.