JPH01315649A - Knocking controller of internal combustion engine - Google Patents

Abstract

PURPOSE:To make constantly accurate knocking decision level by setting a knocking decision level on the bases of both a cumulative % point of distributed approximate log conversion values of knocking intensity data and a value corresponding to the standard error of the approximate log conversion value of the knocking intensity data and using the decision level as a decision factor for the occurrence of knocking. CONSTITUTION:According to the output from a knocking sensor, a knocking intensity value as an effective knocking decision factor is detected by a detecting means B. This knocking intensity value is compared with a knocking decision level by a knocking decision means C to determine whether a knocking occurs or not. According to data obtained in the way above, a knocking control means D controls a knocking control factor. At this time a cumulative % point of distributed approximate log conversion values of the knocking intensity data is detected by a cumulative % point detecting means E and the value corresponding to the standard error of the approximate log conversion value is detected by a standard error detecting means F. Based on both the cumulative % point and the value corresponding to the standard error, a knocking decision level setting means G sets a knocking decision level.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention detects knock occurring in an internal combustion engine and controls knock control factors such as ignition timing and A/F (air-fuel ratio).
This relates to the so-called knock control system (KC3).

[Conventional technology]

Generally, the knock determination level (2) is created by multiplying the average value Vmoan of the knock sensor signals by a constant K.

The optimal value of KC3 changes due to manufacturing variations in the engine, knock sensor, etc., or changes over time.
There is a problem that accurate knock detection cannot be performed.

In order to solve this problem, a system was invented that automatically corrects the knock judgment level based on the distribution shape of the knock sensor signal (for example, Japanese Patent Laid-Open No. 60-243369
(Japanese Patent Application Laid-Open No. 62-267574).

[Problem to be solved by the invention]

However, even with this system, there is a problem in that knock determination is inappropriately performed until the determination level is automatically corrected. Therefore, the inventors of the present invention have considered why accurate knock detection is not possible with the current determination level creation method.

As shown in Japanese Unexamined Patent Publication No. 60-243369, the knock intensity value ■ (an amount effective for detecting knock, for example, the maximum peak value within a predetermined section of the knock sensor signal)
follows a lognormal distribution. FIG. 2 shows the distribution of knock intensity values ■ in a state where no knock occurs at all, drawn on log-normal probability paper. ■ and ■ in the figure are log (V)
This assumes a distribution with a small standard deviation (or it may be replaced with a variance) of , and a distribution with a large standard deviation. l
The standard deviation of og(V) naturally changes depending on manufacturing variations in the engine, knock sensor, etc., and the passage of time.

The reason for writing the standard deviation of ffiog (V) is that the standard deviation of the knock intensity value ■ has almost no meaning in a lognormal distribution. The same thing happens with an average value of V 11111! n and median VP. The same can be said. Mean value V 11111111 and median value ■. are completely different. Because ■□1 and ■. This is because the coincidence is limited to a normal distribution, and this is not the case for a signal that has a lognormal distribution such as a knock sensor signal.

The meaning is greater. Therefore, in the following explanation, the knock determination level will be created based on the more meaningful 5°. Now, VKS=KX■. (K=2
) and the judgment level, the probability of making a knock judgment in ■ is 1%, but in ■ it becomes 10%.

As a result, in an engine with a distribution like ■, there is almost no knock determination when knock does not occur, and in a system that controls ignition timing in response to knock, the ignition timing is controlled to the most advanced angle. However, in an engine with a distribution like ■, the knock determination cannot be ignored, and even though no knock occurs, the ignition timing is retarded, resulting in torque loss.

The above explanation assumes that no knocking occurs, but if knocking occurs, a problem arises in that the knocking cannot be determined.

In this way, ■o=×■,. , or ■. =KXV
In the determination level creation method for n5aa, log(V
) is different, knock detection cannot be performed accurately. That is, this method can simply cope with changes in only the magnitude of the knock sensor signal, and no consideration is given to the standard deviation of log(■).

On the other hand, if the technique disclosed in Japanese Patent Application Laid-Open No. 62-267574 is used, it is possible to cope with changes in the variance of the fog (V) distribution. In other words, this item has a knock judgment level ■. =AXDXVS. (D is a constant), and this A is a quantity that changes depending on the variance of the A o g (V) distribution. Therefore, changes in dispersion due to the engine, knock sensor, etc. can be absorbed.

However, since the update frequency of A is almost the same as the knock determination frequency, the update speed is slow, and as a result, there is a problem that it takes a long time until the knock determination level is corrected to the optimum value.

Therefore, the present invention aims to solve these problems.

[Means to solve the problem]

Therefore, as shown in FIG. 1, the present invention provides a knock sensor for detecting knock of an internal combustion engine, and a knock intensity value that is effective for detecting knock from the signal of this knock sensor.
knock determination means that determines the presence or absence of knock by comparing the knock strength value with a knock determination level; and knock control such as ignition timing or air-fuel ratio according to the determination result. a knock control means for controlling factors; a cumulative percentage point detection means for detecting a cumulative percentage point (2) of the distribution of substantially logarithmically transformed values of the knock intensity value V; and a standard deviation of the substantially logarithmically transformed value of the knock intensity value (■). The value S corresponding to
a standard deviation detection means for detecting the knock determination level; The cumulative percentage point ■.

and knock determination level creation means for creating a knock determination level based on the value S corresponding to the standard deviation.

Here, the knock judgment level VKII is V KD=S
' It is preferable to create it in the form of .

Also, the knock judgment level ■. ■I. = S' x V
It is even better to create it in the form of P4.

In addition, the value S corresponding to the cumulative 50% point VM and standard deviation
It is also possible to include a correction means for correcting the knock determination level created based on the above using a predetermined constant.

[Effect]

Thereby, the cumulative percentage point VP of the distribution of the approximately logarithmically transformed value of the knock intensity value ■ detected by the knock intensity value detecting means is detected by the cumulative percentage point detecting means, and the standard deviation of the approximately logarithmically transformed value of the knock intensity value ■ is detected by the cumulative percentage point detecting means. A value S corresponding to the value S is detected by the standard deviation detection means. Then, a knock judgment level is created by the knock judgment level creation means based on the cumulative percentage point VP and the value S corresponding to the standard deviation, and the knock judgment means compares this knock judgment level with the knock intensity value to determine the presence or absence of knock. judge.

〔Example〕

An embodiment of the present invention will be described below.

FIG. 3 shows an outline of the apparatus. 10 is a knock sensor that detects knock occurring in the engine; 20 is a filter that extracts only components related to knock from the signal of the knock sensor; 30 is an amplifier that amplifies the signal after passing through the filter; 40
50 is a peak hold circuit, 50 is a multiplexer that selects a human input signal to the A/D converter, 60 is an A/D converter that converts an analog signal to a digital signal, and 70 is a knock sensor and other various sensors 60 that output information. Microcomputer that calculates ignition timing, fuel injection amount, etc. based on 80
are the igniter, injector, etc. for realizing this calculation result. Moreover, 9o is various sensors such as a crank angle sensor, an intake air amount sensor, and a water temperature sensor, and AO is a signal processing circuit that processes these sensor signals.

FIG. 4 is a flow chart showing the knock determination and the creation of the knock determination level D. This routine starts from step S00, and the knock intensity ■ is detected at step Sol. This knock intensity value (■) is, for example, a peak hold value of a knock sensor signal within a predetermined crank angle. In step s02, a knock determination is performed. If ■≧
If it is vKD, it is determined that there is a knock. Depending on this result, in the knock control area (e.g. high load area),
Ignition timing etc. are controlled. In step SO3, it is determined whether the operating conditions of the engine satisfy specific conditions. This specific condition is defined by one or more of the following conditions, for example.

■The engine load is above the predetermined value, ■The engine speed is within the predetermined range, ■The rate of change in the engine load is below the predetermined value, and ■■8 is within the predetermined range. If YES in step SO3, go to step SO4, NO
If so, proceed to step SO5. In SO4, log
(V) Update the value S corresponding to the standard deviation of the distribution (details will be described later). In step SO5, ■ the median of the distribution ■ 9
(Details will be explained later).

In step SO6, the knock judgment level ■. Create 9 VKll-33XV using the following formula.

FIG. 5 is a detailed flowchart of step SO4. This routine starts from step SO4, and in step 5041 it is determined that ■≦■4. If YES, the process advances to step 5042; if NO, the process advances to step 5044. In step 5042, it is determined whether SxV≧■9, and if YES, proceed to step 5043; if NO, proceed to step 5044.
Proceed to ゛.

In step 5O43, S is decreased by ΔS.

In step 5044, the S update counter C3 is incremented by one. The step knob 5045 determines whether C3≧3, and if YES, the process proceeds to step knob 5046, and if NO, the process proceeds to step 3048. In step 3046,
Clear C3. In step 5047, S is set to ΔS
Make it bigger. This routine ends at step 8048.

In the process shown in FIG. 5, S becomes I! , the reason why the value corresponds to the standard deviation of the og(V) distribution will be explained using FIG. χ=median χ of the distribution of fog(V).

zH=ffiog(VM), point a is 1-, =lo
'io, where g(VM)-σX (σ8 is the standard deviation of the distribution of χ), becomes 34. This means that the probability that χ-1≦χ≦χ9 is 0.34. On the other hand, according to the process shown in FIG. 5, V? The probability that I/S≦■≦■9 is % (
S will be corrected so that it becomes approximately 33%). In other words, the expected value of the amount of change in S per cycle converges to a value that satisfies the following. Therefore, σX #1
! , og(S) holds true. Using this S, if we create VKD=S'X'J, then ■. is set at the 3σ point of the ffiog(V) distribution (because 3×σ on the logarithmic axis becomes S3 on the real axis).

Next, step SO5 will be explained in detail using FIG. The VP4 update routine starts at step 5050, and at step 3051, it is determined that ■≧■-, and YES.
If so, go to step 5052; if NO, go to step 305
Proceed to step 3. In step 5052, ■A is increased by Δ■. In step 5O53, it is determined that V<Vi, and if YES, the process proceeds to step 5054, and if NO, the process proceeds to step 5055. In step 5054, VP is set to Δ■
Make it smaller. In Stenbu 5O55, Δ■ is updated. This routine ends at step 3056.

In the above explanation, S and ①A are updated for each cylinder.

According to this embodiment, the knock determination level VXO is set to l o
g (V) Since it can be set at the +3σ point of the distribution, even if the variance of
, P, , tv, d V-0,0013, which is an almost negligible probability of misjudging a knock.

In this example, VKD was set to a value corresponding to the 3σ point of the 42 o g (V) distribution, but it is not limited to the 3σ point, but may be set to the 3.5σ point or 2.5σ point as necessary. You can change it. In this case, it is difficult to calculate Sn, but the following method can be used instead. It is the calculation result of S″ for a value S corresponding to a standard deviation in a certain range (for example, 1.0 to 1.5) (for n = 3°5, 1.0 to 1.5).
4.1) is written in the ROM in advance and the result of S''' is obtained in the form of a table search.This method requires a lot of time for the microcomputer to perform multiplication even if n is an integer value. This is a very effective method in some cases.

In addition, in this embodiment, the cent value of the S update counter is set to 3
Although S is made to correspond to the standard deviation of the log(V) distribution, it is not limited to 3 and may be any other number. For example, C3=4
Then, S is updated so that SVM P , v+d V=Va, which is 0.67σ from the normal distribution table.
, the quantity corresponding to (0,67σX= l o g (S)
). Therefore, if VP,=S'-"XVP, then the knock determination level can be set at the 3σ point.

Further, in this embodiment, the VKD was created using the median value ■9, but the value is not limited to the median value, and other cumulative °% point values ■ may be used. For example, if the cumulative 16% point is VI6, VKO
-3'X'J, , can be set to 3σ point.

Further, in order to change the control knock level depending on the engine state or cylinder, the determination level may be corrected using a constant. For example ■. =s' xcXVP (C: constant set depending on engine state or cylinder).

Furthermore, it is also effective to limit S within a predetermined range if necessary.

It is also effective to limit the knock determination level to within a predetermined range, if necessary.

Further, in this embodiment, the present invention was used without logarithmically converting ■, but the present invention can be used even when logarithmically converting. In that case, for example, the 3σ point is (VP+3XS), and the only difference is the form of the equation from the above case.

It is also effective to combine the present invention with the technique disclosed in Japanese Patent Application Laid-Open No. 60-243369. An example of this case will be explained using the flowchart of FIG. This is the fourth
This shows an alternative process to step S09 in the figure, and the other steps may be the same as described above.

This routine starts from step 5O700, and 5071
The variable A is calculated as A=S"+D. Step 5072
Then, it is determined whether the engine operating condition or the knock sensor signal satisfies a specific condition. If YES, the process proceeds to step 5074; if NO, the process proceeds to step 3078. These specific conditions are, for example, (i) engine load is above a predetermined value, (ii) engine speed is within a predetermined range, (iii) rate of change in engine speed is below a predetermined value, (iv) VM is within a predetermined range. , is. In step 5O74, it is determined that V≦VP/A, and if YES, the process proceeds to step 5075, and if No, the process proceeds to step 5076. In step 5075, D is increased by ΔD. In step 5076, ■≧■.

is determined, if YES, go to step 5O77, NO
If so, proceed to step 8078. In step 5077, D is reduced by 2×ΔD. This routine ends at step 5O7B. Here, D is not limited to a positive value, but may also have a negative value.

〔Effect of the invention〕

The present invention produces the following effects.

(2) Since the knock determination level is created based on the standard deviation S of the ffiog(V) distribution, accurate knock determination can be made with respect to changes in the dispersion of V due to changes in the engine, knock sensor, etc.

■It is possible to increase the update frequency of S (in the example, 3
(once per cycle), the determination level is quickly corrected to an appropriate value.

(2) By devising the updating method of S, if (2) A deviates from the true value during a transient period, S moves to correct the deviation, so that accurate knock detection can be performed even during a transient period.

The present invention is disclosed in Japanese Unexamined Patent Publication No. 60-243369, Japanese Unexamined Patent Publication No. 62
- This is an improvement on Publication No. 267574, but the effects ■ and ■
This is an effect unique to the present invention that is not found in these prior applications.

In particular, regarding effect ■, supplementary explanation using Figure 9 (b
), <' VM becomes smaller than the true value (in steady state, as shown in Figure 9 (a), ■ A is the same as the true value)
. For example, now ■o is the median true value at the cumulative 40% point), so in order to satisfy S□1iP(V)dV=33%, S″P(V)dV=7S S will be corrected. Since the cumulative 40% point and 7% point correspond to -0625σ point and -1.47 point, respectively (calculated from the normal distribution table), S will be corrected by fog (
V) It will be corrected toward the amount corresponding to -0,25+1.47=1.22σ of the distribution, and will take a larger value than the true value. Therefore, since S corrects the deviation from the true value of VH, erroneous knock determination can be prevented. Conversely, during deceleration, VH becomes larger than the true value.

For example, if ■o is now at the cumulative 60% point, S is corrected to become 27) as above. Cumulative 60%
The point and the 27% point are +0.25σ point and -0°6, respectively.
Since it corresponds to the 1σ point, S is eventually corrected to an amount corresponding to +0.25 + 0°61 = 0.86σ, so it becomes a value smaller than the true value, and the deviation from the true value of ■9 is corrected.

As described above, the present invention can produce excellent effects that are not found in JP-A-60-243369 and JP-A-62-267574.

[Brief explanation of the drawing]

Fig. 1 is a complaint correspondence diagram of the device of the present invention, Fig. 2 is a distribution characteristic diagram of the knock intensity value ■ on corresponding normal probability paper, Fig. 3 is a block diagram showing an embodiment of the device of the present invention, Fig. 4, 5, 7, and 8 are flowcharts for explaining the operation of the apparatus shown in FIG. 3, and FIGS. 6 and 9(a) to 9(C) are distribution characteristic diagrams of knock intensity values. IO...Knock sensor, 70...Microcomputer, 80...Injector, igniter. Representative Patent Attorney Takashi Okabe ■ Figure 2 Figure 4 Figure 5 Figure 6 Figure 7-CL
-ノ＾CL

Claims (6)

[Claims] (1) A knock sensor for detecting knock of an internal combustion engine, a knock intensity value detection means for detecting a knock intensity value V effective for knock detection from a signal of the knock sensor, and a knock determination based on the knock intensity value. knock determination means for determining the presence or absence of knock by comparison with the knock level; knock control means for controlling knock control factors such as ignition timing or air-fuel ratio according to the determination result; and approximately the logarithm of the knock intensity value V. Cumulative % of distribution of converted values Detecting point V_P
a point detection means, and a standard deviation detection means for detecting a value S corresponding to a standard deviation of a substantially logarithmically converted value of the knock intensity value V;
The knock judgment level V_K_P is set as the cumulative percentage point V_P.
and knock determination level creation means for creating a knock determination level based on the value S corresponding to the standard deviation. (2) The knock control device for an internal combustion engine according to claim 1, wherein the knock judgment level creation means creates the knock judgment level in the form of V_K_D=S^n×V_P (n is a constant). (3) The knock control device for an internal combustion engine according to claim 1, wherein the cumulative percentage point V_P is a cumulative 50% point V_M. (4) The standard deviation detection means updates the value S corresponding to the standard deviation so that the probability that V_M/S≦V≦V_M becomes 1/3. A knock control device for an internal combustion engine as described above. (5) The knock control device for an internal combustion engine according to claim 3 or 4, wherein the knock determination level creation means creates the knock determination level V_K_D in the form of V_K_D=S^3×V_M. (6) Any one of claims 3 to 5, further comprising a correction means for correcting the knock judgment level created based on the cumulative 50% point V_M and the value S corresponding to the standard deviation by a predetermined constant. The knock control device for an internal combustion engine according to claim 1.