JPS62126244A - Knocking detecting device for internal combustion engine and control device using said detecting device - Google Patents

Abstract

PURPOSE:To enable the detection of a precise and actual knocking condition by detecting a mean knocking condition through the evaluation of a distribution form obtainable from the multi-cycle sampling of a knocking force value corresponding to a maximum peak value. CONSTITUTION:Within a control circuit 10 is provided a knocking force detecting means for taking out a knocking force value such as a rectification integral corresponding to the maximum peak value of signals from a knock sensor 4 for detecting the knocking of an engine 1. And the form of distribution available from the multi-combustion sampling of said knocking force value is evaluated to judge whether the distribution conforms to a predetermined form. Depending upon the result of the evaluation, a means knocking condition ranging over a plurality of cycles is detected and a knocking control factor is controlled correspondingly. According to the aforementioned constitution, a knocking condition can be detected independent of the dispersion, the aging and the like of the knock sensor 4 and the engine 1, and the engine 1 can be kept at a condition wherein a target knocking occurs among a plurality of cycles.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention provides a device for detecting the occurrence of knocking (hereinafter referred to as “knock”) that does not occur in an engine, and a device for detecting the knocking condition using this device and controlling the ignition. A knock control device (which controls knock control factors such as timing, air-fuel ratio, and intake pressure)
(hereinafter referred to as a knock control system).

[Conventional technology]

-i, the knock control system is a knock control system that uses a knock sensor to detect vibrations in the engine body caused by knock, and advances or retards the ignition timing according to the detection results to keep the ignition timing at the knock limit. This is to improve engine output and fuel efficiency.

In such a knock control system, as disclosed in, for example, Japanese Unexamined Patent Publication No. 58-7538, it is very important to accurately detect knock using a knock sensor.

In knock control systems that are currently being mass-produced, a type of knock sensor that mainly detects vibrations of the engine body is used from the viewpoint of cost and reliability. Since this type of knock sensor is affected by the mechanical vibration noise of the engine, the S/N ratio deteriorates in a high-speed rotation range where the vibration noise becomes large, making it extremely difficult to detect high-speed knock.

Further, the sensor output characteristics change due to manufacturing tolerances of the knock sensor and their changes over time, and the vibration transmission characteristics of the engine change due to individual engine differences and their changes over time, resulting in large variations in sensor output.

Therefore, it is difficult to detect high-speed knock in consideration of engine or sensor variations and their changes over time, and the current situation is that knock control is essentially cut off in the high-speed range.

In addition, in the low-to-medium speed range, even if extremely loud knocks that can damage the engine can be detected, due to the above-mentioned variations and changes over time, the knocking noise during control becomes louder (varies), making it difficult for the driver to hear. There were often cases where large knocks caused discomfort.

It is a well-known fact that, in order to solve the above-mentioned problems, sensors with a high signal-to-noise ratio, such as cylinder pressure sensors and combustion light sensors, are being studied at the laboratory level. It is also true that such sensors are currently very expensive and still have many reliability problems. However, even if the cost and reliability issues were resolved,
Issues of sensor manufacturing tolerances and aging remain.

In addition, it is true that such a sensor is not subject to the engine's mechanical vibration noise in principle, so it is expected that the S/N ratio in the high-speed range will be improved, but the vibration noise generated by the engine is not only mechanical noise. There is also a lot of pressure vibration noise and combustion light noise caused by combustion.

These combustion noises naturally change due to individual differences in engines, so basically they are caused by engine variations,
It cannot escape the influence of changes over time. That is, the above-mentioned problems cannot be fundamentally solved even if such a sensor is used.

[Problem that the invention seeks to solve]

In view of the above-mentioned problems, the present invention aims to provide a device that accurately detects knock without being affected by knock sensor or engine variations, changes over time, etc., and to provide a knock control device using the same. It is something.

[Means for solving problems]

In order to achieve the above object, a first aspect of the present invention provides a knock sensor for detecting engine knock;
Once the knock intensity value V is detected, a knock intensity value such as a rectified integral value corresponding to the maximum peak value of this signal is extracted from the signal of this knock sensor, and this knock intensity value V is obtained when sampling the pre-combustion cycle. A determining means is provided for determining whether the distribution shape is a predetermined distribution shape or not, and an average knock state over a plurality of cycles is detected according to the result of this determination.

In addition to the configuration of the first invention, the second invention further includes a control means for controlling knock control factors such as ignition timing, air-fuel ratio, and intake pressure according to the detection result of the knock state. We are prepared.

Furthermore, in addition to the structure of the first invention, the third invention includes a knock determination means for determining the presence or absence of knock for each combustion cycle based on the signal of the knock sensor, and a The apparatus includes a control means for controlling the knock control factors as described above, and a means for correcting a knock determination level for making a knock determination for each cycle according to the knock state detection result.

In addition, the fourth invention has the structure of the first invention,
a knock determination means for determining the presence or absence of knock for each combustion cycle based on a signal from a knock sensor; a control means for controlling the knock control factors as described above according to the knock determination result; and limiting means for adding a limit to the control value range of the knock control factor according to the detection result of the state.

[Effect]

Hereinafter, the outline of the present invention and the technical basis of the present invention will be explained in detail.

In order to solve the above problems, the present inventors previously proposed several methods and devices in Japanese Patent Application No. 59-99898. In this Japanese Patent Application No. 59-99898, when the maximum wave height value VIIAX in a predetermined section of the knock sensor is sampled many times in the non-knock state, it forms one approximately log-normal distribution, and is sampled in the knock state. In this case, the distribution is a combination of two approximately lognormal distributions with different variances.

Here, in the above invention, the maximum peak value V of the sensor signal,
4AX, and in order to detect the maximum peak value VMAX, a peak hold circuit or, instead, software of a microcomputer (hereinafter referred to as microcomputer) for detecting the peak is required. Among the knock control systems currently in mass production or under research and development, there are many systems that do not directly detect the maximum peak value v14AX, so in such systems, separate V and AX detection circuits or microcontrollers are required. It is necessary to add additional software, etc.

In view of this, the present inventors subsequently conducted detailed experiments and devised methods to detect the knock-specific frequency components (generally 6 to 9 KHz < 50kHz) without necessarily detecting the maximum wave height value V MAX. However, depending on the engine structure, it may exist on the lower frequency side or on the higher frequency side of 10KH2 or higher.) Indicates the strength of the knock vibration component corresponding to the maximum wave height value. It's the amount. For example, the value obtained by integrating the knock vibration output, the number of pulses in a pulse train obtained by comparing the knock vibration output with a predetermined level, the integrated value thereof, or the effective value of the knock vibration output (hereinafter referred to as the knock intensity value) ).

Furthermore, by comparing and examining the distributional properties of these knock intensity values, we discovered more general properties common to these knock intensity values. The present invention utilizes the common properties of knock intensity values to provide a device that always accurately detects knock conditions regardless of the type of knock sensor or method of extracting the knock intensity value, and a high-performance knock control using the same. Before explaining these knock intensity values V and their properties, we will first briefly explain the properties of the maximum wave height value V WAX on which the discovery was based.

In Figure 1, figure (1) shows 6 to 9 K! of the vibration of the engine body! (This is the knock sensor signal when only the frequency component of Z is extracted. If the maximum peak value V MAX of this signal within the engine combustion section (for example, 10 @ ~ 90 "ATDC) is sampled over multiple combustion cycles,
(2) A frequency distribution as shown in the figure is obtained. This V MAX
If we logarithmically transform and draw the frequency distribution again, if we sample with no knock at all, we will get a distribution that is close to a normal distribution as shown in Figure (3), and if we sample with no knock occurring to some extent, In the case of sampling, (4)
As shown in the figure, it deviates from the normal distribution on the high output side.

In order to show this more clearly, we will use a log-normal probability paper. Generally, if the frequency distribution is a normal distribution, the cumulative frequency distribution will be an S-shaped curve with the cumulative 50% point as the inflection point, as shown in Figure 2 (1). Normal probability paper, as shown in Figure (2), is paper with a cumulative frequency scale so that the S-shaped curve peculiar to the normal distribution lines up exactly on a straight line when plotted.

This kind of normal probability paper (called lognormal probability paper when the characteristic values are logarithmic values) itself is already known in the field of statistical analysis, so a detailed explanation of the paper will be omitted, but it will be explained later. Only the properties of the paper that are directly relevant to the explanation will be described here.

The vertical axis of the normal probability paper, that is, the cumulative frequency scale, is 50%.
The further you move towards the scales at both ends of the point, the greater the distance from the 50% point (for example, the interval between the 70% and 60% scales is longer than the interval between the 60% and 50% scales), but the 50% The scale is vertically symmetrical with respect to the % scale. (For example, the 10% scale and 90% scale are equidistant from the 50% scale, and generally α
The % scale and the (100-α)% scale are equidistant from the 50% scale. ) Another important property of normal probability paper is that the slope of a straight line when plotted on this paper matches the reciprocal of the standard deviation σ (i.e., 1/σ). Therefore, the more the straight line lies sideways, the greater the dispersion of the distribution.

Now, if we plot the distribution of the maximum wave height value V HAX in the non-knock state on this log-normal probability paper, we can see the (
2) It becomes a straight line as shown by [No knock] in the figure. This means that the maximum wave height value VM□ without knocking forms one lognormal distribution.

On the other hand, the distribution of states with knock occurring at a certain frequency (i.e., a state where knock cycles and non-knock cycles coexist at a certain rate) is the same (2
) The line becomes a polyline whose slope changes midway, as shown by "With knock" in the figure.

That is, the lognormal distribution with small variance formed on the relatively low output side and the lognormal distribution with large variance formed on the relatively high output side overlap.

Now, examples of knock intensity values V other than the maximum wave height VMAX having the property of approximating approximately one log-normal distribution in a non-knock state are as follows.

The first is a predetermined section (for example, 10° to 90° ATDC)
This is the rectified and integrated value of the vibration output.

By changing the time constant of the integrator, this value represents the average vibration output of that cycle or the integrated value of the vibration output of that cycle, but in either case, it corresponds to the vibration intensity of that cycle. value.

The second is the number of pulse trains generated when the vibration output in a predetermined section is compared with a certain threshold value (for example, a level slightly above the vibration center), or the integrated value of the pulse trains.

This value also corresponds to the vibration intensity of that cycle.

This intensity value has a limited variation range compared to the range that the maximum wave height value VMAX can take, so it is slightly inferior in accuracy, but it has the advantage of being easy to handle.

A third example is the effective value (RMS) of the vibration output in a predetermined section. This RMS is JJ (t) ” d t where the vibration level at time t is x (t) and the time length of the predetermined section is T
/T, which also corresponds to the vibration intensity of the cycle.

From the above three examples, the strength value that generally corresponds to the vibration strength■
It is natural and reasonable to assume that the distribution in the non-knock state will be approximately logarithmic. For example, a certain intensity value V is the maximum wave height value VMAX that is a member of the same intensity value.
For approximately V=a ・(VMAX)' (a,
n is an arbitrary real number), it is clear that 1ogV = Ioga + n logVMax, and like %1ogVMAX, log V also has a normal distribution (where n is a lognormal distribution).

The present inventors next investigated what kind of distributional difference there is in the strength value ■ in the state with and without knocking.

FIG. 3 shows an example of the distribution when an in-cylinder pressure sensor is attached to one cylinder of the engine and the integral value Jv mentioned in the first example is extracted from this signal. The knock-specific frequency component (6 to 9Kllz) of the cylinder pressure sensor signal
) is taken out by a band-pass filter, and after half-wave rectification, it is integrated with a time constant of 5 m5ec from 10' ATDC, and the value is taken in at a timing of 90° ATDC.

As can be seen from Figure 3, the distribution without knocking is
Lognormal probability is represented by a straight line on paper. In other words, they form a lognormal distribution. On the other hand, the distribution in the state with knock is certainly a straight line on the low value side, but it becomes a curve that sharply bends at a certain point. This difference in the distribution is compared with the difference in the distribution of the maximum wave height value VNAX of vibration of the engine body.

Figure 4 shows the maximum wave height of engine vibration under the same conditions as Figure 3. This is the distribution when sampling 0. In other words, the signal from the piezoelectric knock sensor attached to the engine block has a center frequency of 7.5KH2 and a bandwidth of Q=
Filtered with a 20 dB bandpass filter, the maximum peak value V M between 10' and 90° ATDC of this signal
This is the distribution when AX is sampled only for the cylinders to which the above-mentioned cylinder pressure sensor is attached. Since data sampling is performed simultaneously by the cylinder pressure sensor and the vibration sensor, the knock conditions are exactly the same.

Comparing Figures 3 and 4, we find the following.

First of all, the cylinder pressure sensor integral value 5V in Fig. 3 is
The bending is clearer than the maximum wave height value V WAX of the vibration sensor in Fig. 4, and secondly, the integral value fv of the cylinder pressure sensor is bent in a curved line on the high output side, whereas the vibration The maximum wave height value VMAX of the sensor is a point that is bent almost linearly. As will be explained in detail later, this phenomenon means that the cylinder pressure sensor has a higher ability to distinguish between noise and knock, that is, has a higher signal-to-noise ratio.

However, the most important point in comparing Figures 3 and 4 is that although there is a difference in the way they bend as mentioned above, the cumulative percentage of bends occurring is fv, v,
Even with A×, there is almost no change (about 70% in this example)
)That's what it means.

This indicates that the cumulative % corresponding to the bending point can be a very effective and universal indicator for identifying knock conditions. This is because, even though knock sensors of completely different types and different knock intensity values were extracted, the cumulative percentages of the bending points are almost the same.

In fact, as will be explained later, the cumulative percentage corresponding to this turning point is the percentage of non-knock cycles (noise cycles) among all cycles, in other words, the remaining percentage (in this example, 1
00-70=30%) indicates the ratio of knock cycles.

This will be explained in detail using FIG. (11 in Figure 5)
The figure shows the frequency distribution ( (Illustrated in the upper row) and the corresponding cumulative frequency distribution (Illustrated in the lower row) on normal probability paper. The horizontal axis is on a logarithmic scale. In the frequency distribution in the upper row, the solid line shows the distribution of relatively small knock states where the ratio of noise to knock is 90:10, that is, 10% of the total is knock cycles, and the broken line shows the distribution of relatively small knock states where 40% of the total is knock cycles. This is a relatively large knock state distribution. In actual engine knock, the frequency of knocking increases as the knock gets louder, and the knock strength value itself also increases, so the distribution corresponding to the knock cycle among the broken line distributions is
In reality, it shifts toward the larger value. However, as the knocking condition increases, the frequency of knocking inevitably increases, so even if we discuss only the knocking frequency, the essence will not be lost.

If we plot the cumulative frequency distribution corresponding to each knock condition in the upper row (i.e., the distribution when the noise distribution and the knock distribution are considered as one group and accumulate from the smallest value) on log-normal probability paper, we get It will look like the bottom row. As can be seen from this figure, when viewed from the lowest value, it is almost a straight line up to a certain point, and then sharply bends into a downwardly convex curve. The cumulative percentage of points where bending occurs indicates the noise generation rate in any case. In other words, the remaining percentage indicates the knock occurrence rate.

Next, let's assume a situation in which the same knocking occurs but the discrimination between knock and noise becomes extremely poor, that is, a case where a sensor with a low SN ratio is used or a knock intensity value with a poor SN ratio is selected. The solid line in Figure 5 (2) indicates the knock frequency of 10% (the noise and knock ratio is 90:
10) represents the distribution sampled using a sensor with a good SN ratio, and the broken line represents the distribution when a sensor with a poor SN ratio or a knock intensity value with a poor SN ratio is selected under the same knock condition. Furthermore, the corresponding cumulative distribution is shown in the lower row. From this figure, it can be seen that the distribution of a sensor with a good SN ratio bends in a curved line from a certain point, but when the SN ratio is relatively poor, it can be approximated by two straight lines that bend in the middle. It can be seen that the cumulative percentage of occurrence of bending is almost the same in both cases. This cumulative percentage also indicates the noise occurrence rate, and therefore the remaining percentage indicates the knock occurrence rate.

In other words, even if you select a sensor or knock intensity value with a somewhat poor S/N ratio, or a sensor with a relatively high S/N ratio or a knock intensity value, in either case, the cumulative percentage at which bending occurs will be accurately calculated. It is possible to indicate the occurrence rate of knocking and therefore the knocking condition.

For example, as shown in the frequency distribution indicated by the broken line in Figure 5 (2), the noise distribution and the knock distribution overlap in the shape B (
In other words, in a state where the signal-to-noise ratio is poor), it is impossible to individually identify whether each knock intensity value in the overlapping portion is noise or knock. In other words, the conventional method of determining a certain knock intensity threshold value and determining that an output above this level is considered knocking and anything below this level as noise cannot be identified at all. However, if attention is paid to the cumulative percentage of the bending points as described here, it is possible to accurately detect the knock state even in such a case.

As described above, a knock state can be accurately detected by focusing on the cumulative percentage of bends on the lognormal probability paper. If we carefully consider the properties described so far, we can understand this as a more general property.

In other words, the essence of the fact that the distribution in the presence of knock bends sharply on the lognormal probability paper is that the distribution is estimated only from data with relatively low output values, most of which can be considered as a noise group. On the other hand, data on the high output side deviates from the estimated value in such a way that it does not belong to the population of that distribution. By knowing from which point in the distribution the data deviates, it is possible to accurately detect the state of knocking.

In other words, essentially, the distribution does not need to be approximated as a substantially logarithmic distribution.

Of course, a state in which all of the sampled data is knocking, that is, an extremely large knocking state in which all cycles are knocking, cannot be identified by the method described above. However, such an oversized knock state can be sufficiently identified by conventional methods (for example, a method in which a knock is determined when the knock intensity exceeds a certain threshold value), so it can be used in parallel with such methods. If you do that, it won't be a problem at all.

Furthermore, in such an extremely large knock state, the data dispersion becomes extremely large, so if a method for monitoring dispersion is added, there is no need to concurrently use conventional methods.

Now, it is very easy to implement the method described above as a laboratory device (for example, a knock monitor measuring instrument). There is no need to explain that the function of sampling data, the function of logarithmically converting it, and the function of converting it into a cumulative frequency distribution and calculating the cumulative percentage of curvature can be realized by using a microcomputer.

However, in order to apply the present invention to an actual knock control system, further improvements are required.

That is, in an actual engine, it is necessary to detect a knock state more quickly, and furthermore, it is necessary to incorporate a control algorithm into the limited ROM and RAM of an on-vehicle microcomputer. To this end, it is desirable to more easily determine the cumulative percentage at which bending occurs, without taking the trouble to sample a large number of data and without performing logarithmic transformation. The following is an example of such a scheme. Of course, various other methods can be considered.

The method will be explained below using FIG. 6.

Figure 6 (1) shows the cumulative distribution of log V (illustrated on the log-normal probability paper in the upper row) and the frequency distribution at that time (illustrated in the lower row) when the knock is smaller than the control target. . In contrast, Figure (2) in the same figure shows the cumulative distribution and frequency distribution in a knocking state that is larger than the control target. Currently, knocking occurs at 10% of all cycles (therefore, the remaining 90% are non-knocking cycles)
Let be the knock state of the control target. At this time, the distribution of the cumulative frequency in a knock state smaller than the target bends at the output side higher than the cumulative 90% point on the lognormal probability paper, as shown in the upper part of the +11 diagram. Therefore, on the output side below the 90% point, it can be regarded as one straight line (that is, one lognormal distribution).

Now, let us assume the cumulative 50% point on the lognormal probability paper, that is, the median value M of the frequency distribution, and consider two threshold values VL and Vi that are equidistant above and below on the logarithmic axis with respect to this M. Equidistance on the logarithmic axis means a geometric relationship on the real axis (logVH-1ogM-1ogM-1o
gVL, that is, Vn/M"M/Vt). At this time, if the lower threshold value VL is set so that the knock intensity value falls below VL by the same number of frequencies as the control target knock occurrence frequency of 10%. In other words, V at the cumulative 10% point.
Suppose that L is set. Then, when the knock state is smaller than the target value as shown in Figure 6 (1), the upper threshold value VH is cumulatively 90% due to the properties of lognormal probability paper.
Match points. Therefore, when the knock is smaller than the target, the frequency at which the knock intensity value V falls below the lower threshold value VL is equal to the frequency at which it exceeds the upper threshold value VH. (10% in this case) However, in a knock state larger than the target as shown in (2) of the same figure, the bend occurs at a percentage lower than the cumulative 90%, so the upper threshold value V1
1 does not correspond to 90% cumulative, but corresponds to a lower cumulative %.

In other words, the frequency at which this upper threshold value (■8) is exceeded increases. (2) As shown in the upper part of the figure, the frequency increases by the amount indicated by the double-headed arrow. Therefore, when the knock is larger than the target, the frequency at which the knock frequency value ■ exceeds the upper threshold VH is greater than the frequency at which it falls below the lower threshold VL.

From the above, the knock intensity value V is monitored for a predetermined cycle, and the number of times NL that this V has fallen below the lower threshold value VL and the number of times N that this V has exceeded the upper threshold V1,
By comparing these, it can be determined whether the knocking state is larger or smaller than the target.

The remaining task is to set (1) and VH. The setting method will be described below.

Now, consider finding the approximate value 8 of the cumulative 50% point M without data sampling. Cumulative 50%
Point M is a point where the probability that the knock intensity value V exceeds M is equal to the probability that it falls below M. Therefore, every time a knock intensity value V is input, compare that V with a certain value ■8, and if ■ is larger, increase ■8 by ΔvM, and conversely, if it is smaller, increase ■i by the same amount. ■ If the VM is updated sequentially so that ΔVM decreases. is cumulative 50%
Converges to point M. That is, the cumulative 50% point M can be determined without taking the trouble of sampling data to determine the distribution.

Next, consider finding the lower threshold value VL.

■ is the cumulative 10% point, so the knock intensity value ■ is V
The probability of being below L is 1/10, and conversely, the probability of being above VL is 9/l 00 points. In other words, the respective probabilities are 1:9. Therefore, consider finding the cumulative 50% point M using a method similar to the method used to find it. Every time a knock intensity value ■ is input, compare that V with a certain value VL, and if ■ is larger, increase VL by ΔVL,
On the other hand, since it was small, VL was 19·Δ■, which is 9 times ΔVL.

If VL is updated sequentially so that VL decreases by
converges to the cumulative 10% point. Because the actual cumulative 10
This is because the expected value of increasing or decreasing VL is 0 only at the % point (9/10xΔVL 1/10x9
- ΔVt, = 0) - That is, by changing the ratio of the amount of increase to the amount of decrease, it is possible in principle to converge the value to an arbitrary cumulative percentage point.

However, this is not a problem when calculating a cumulative percentage point where the amount of increase and the amount of decrease are almost equal, such as the cumulative percentage point, but it is not a problem when calculating a cumulative percentage point that is far away from the cumulative percentage point (for example, 10% or 90%). When determining the point), the ratio between the amount of increase and the amount of decrease becomes too large, and there is a fear of hunting near the convergence value. Therefore, the present invention proposes another method for determining VL.

Now, the knock intensity value V, a certain value VL, and the predetermined number of cycles C
Continuously compare only between If the knock strength value V falls below VL even once during that period, set VL to Δ■,
■ Only if V continues to exceed VL during this period.

By increasing ΔVH, the predetermined number of cycles C
Consider updating the VL one after another.

Since the amounts of increase and decrease are the same, ■ converges to the point of the following probability relationship. That is, the knock intensity value V is V
Let P be the probability of falling below the convergence value of L, (1-P) c
=1-(1-P) converges to the point shown by the c(7) relationship. In order to converge VL to the cumulative 10% point, P =
0.1 and the number of cycles C should be about 7 cycles. With this method, the cumulative percentage far away from the cumulative 50% point
Point values can be determined sequentially with high accuracy.

In this way, the cumulative 50% point VH and the lower threshold VL could be determined, so the upper threshold
As VL=A,■□=■I ・A(=VMz/VL)
can be easily found.

The method described above makes it possible to easily obtain the cumulative percentage points of bends without data sampling and without logarithmic transformation.

〔Example〕

The apparatus of the present invention and its operation will be described in detail below according to embodiments.

FIG. 7 is a configuration diagram showing an embodiment of the present invention.

In the figure, 1 is a 4-cylinder 4-stroke engine, 2 is a reference angle sensor for detecting the engine's reference crank angle position (for example, top dead center), and a crank angle sensor that generates an output signal at every fixed crank angle of the engine. 3 is an igniter and a coil that receives an ignition timing control signal output from the control circuit 10 and cuts off power to the ignition coil.

Reference numeral 4 denotes a knock sensor for detecting engine knock vibration corresponding to the engine knock phenomenon using a piezoelectric element, 5 an air flow meter that detects the intake air amount of the engine, and outputs a signal in accordance with this. a throttle valve connected to a throttle angle sensor; 7 an injector for injecting fuel into the intake manifold based on the fuel injection timing and fuel injection period determined by the control circuit 10;
8 is a turbocharger for supercharging; 9 is a 02 sensor that generates an output signal depending on whether the air-fuel ratio of exhaust gas is rich or lean compared to the stoichiometric air-fuel ratio; 10
is a control circuit for controlling the ignition timing and air-fuel ratio of the engine according to the state of input signals from each of the sensors.

Next, the detailed configuration and operation of the control circuit IO will be explained with reference to FIG. In Fig. 8, 10-1 is a central processing unit (CPU) for calculating ignition timing and fuel injection amount.
A microprocessor with an 8-bit configuration is used. 1
0-2 is a read-only storage unit l-(ROM) for storing control programs and control constants necessary for calculations.
), 10-3 is a temporary storage unit (RAM) for temporarily storing calculation data during operation of the CPU 10-1 according to a program. 10-4 and 10-5 are waveform shaping units for shaping the output signals of the reference angle sensor 2-1 and crank angle sensor 2-2 (a magnetic pickup is used in this embodiment) built into the distributor 2. It is a circuit.

10-6 is an interrupt control unit for causing the CPU to perform interrupt processing based on an external signal or an internal signal; 10-7 is a CP
This is a 16-bit timer configured so that the count value increases by one at each clock cycle, which is the basic cycle of U operation. The engine speed and crank angle position are input to the CPU by the timer 10-7 and the interrupt control section 10-6 in the following manner. That is, the reference angle sensor 2-
Each time an interrupt occurs due to an output signal of 1, the CPU reads the count value of the timer. Since the count value of the timer increases every clock cycle (for example, 1 μs), the timing of the reference angle sensor signal can be determined by calculating the difference between the count value at the current interrupt and the count value at the turning interrupt. The interval, that is, the time required for one rotation of the engine can be measured.

In this way, the engine speed is determined.

The crank angle position can be determined by using the top dead center signal of the reference angle sensor 2-1 as a reference, since the signal from the crank angle sensor 2-2 is output at every fixed crank angle (for example, 30° CA). can be determined in units of 30'CA. This crank angle signal every 30'CA is used as a reference point for generating an ignition timing control signal.

10-8 is a multiplexer for timely switching a plurality of analog signals and guiding them to an analog-to-digital converter (A/D converter) 10-9;
It is controlled by a control signal output from -12. In this embodiment, an intake air amount signal from the air flow meter 5 and a knock intensity detection circuit 10-1 are used as analog signals.
A knock intensity value signal starting from 0 is input. (In addition, water temperature sensor signal, battery voltage, etc. are input.)10
-9 is A/A to convert analog signal to digital signal.
It is a D converter. 10-10 is a knock intensity value detection circuit for receiving the signal from the knock sensor 4 and extracting the knock intensity value V for each cycle. Reference numeral 10-11 is a manual port for digital signals, and the input signal from the 0□ sensor 9, the idle signal and power signal from the throttle sensor 6, etc. are manually input to this port.

10-12 are output ports for outputting digital signals. This output port outputs an ignition timing control signal for the BUNITA 3, a fuel injection control signal for the injector 7, a control signal for the multiplexer 10-8, and a control signal for the knock intensity value detection circuit 10-10. 10-13 is a CPU bus, and the CPU carries control signals and data signals on this bus signal line to control peripheral circuits and send and receive data.

Next, the knock intensity value detection circuit 10-10 will be explained using FIG. 9. In FIG. 9, 10-10-1 is a filter such as a band pass or bypass for selecting and extracting only the knock frequency component from the output signal of the knock sensor 4, and 10-10-2 is a half-wave rectifier for the output of this filter. Alternatively, a rectifier for full-wave rectification, 10-10-3 is an integrator for integrating the output of the rectifier to obtain the knock strength value for each cycle, and 10-10-4 is for controlling the integration interval of the integrator. This is a gate circuit.

The operation of this knock intensity value detection circuit will be explained using FIG. 1O. In Figure 10, (Figure 11 is the knock sensor output after passing through the filter, and Figure (2) is the knock sensor output after passing through the filter.
This is the half-wave rectified output. (3) The diagram shows the control circuit 10
This is an output signal of a gate circuit 10-10'-4 that operates according to a control signal output from an output port 10-12 in the gate circuit 10-10'-4. That is, it rises at about 10' ATDC and rises at about 90' ATDC.
This is the signal that falls immediately after the knock strength is taken in by CPU10-1 at 6ATDC, and this is the signal that falls immediately after the knock strength is captured by CPU10-1.
Determine the integral interval of 3. (4) The figure shows the output signal of the integrator 10-10-3 which is set and reset by the gate circuit.

In this embodiment, the integrator has a time constant of approximately 5 m5ec.
has been adjusted to. When the gate circuit rises, integration of the signal after half-wave rectification (Figure (2)) is started, and when it falls, the integrated value is reset. The gate circuit is CP
Since Ul0-1 falls immediately after taking in this integral value through the multiplexer 10-8 and the A/D converter 10-9, the integral value for each cycle, that is, the knock intensity value, is successively taken into the CPU.

Next, knock detection and the operation of the knock control system will be described in detail using flowcharts shown in FIGS. 11 to 15.

FIG. 11 shows an interrupt routine for setting the ignition timing and fuel injection time in the timer of the microcomputer. Interrupt 1
00, in step 200, the basic ignition timing θ8SE and the basic injection time τ83. is calculated.

This basic ignition timing θ, 56 and basic injection time τ85. teeth,
Engine rotation speed N and engine addition Q/N (Intake air IQ measured by an air flow meter is
(The value divided by is proportional to the engine load) is stored in the microcontroller's ROM as a two-dimensional map. At this step 200, the ignition timing and injection time are simultaneously corrected based on various sensor signals other than the knock sensor.

For example, the ignition timing and injection time may be modified based on water temperature, or the injection time may be modified based on battery voltage (taking into account invalid injection time).

Next, in step 300, it is determined whether the current operating conditions are knock control execution conditions. For example, it is generally known that when the load on the engine is light, knocking is likely to occur, and that if knock control is forcibly performed under this light load, output, fuel efficiency, etc. will actually decrease. Therefore, in this embodiment, knock control is executed only when the engine load (determined by Q/N) is equal to or higher than a predetermined value. (In addition, it is also possible to set a limit based on the engine speed.) If it is determined that the knock control execution condition is met, the ignition timing and injection time are corrected by knock control in step 400. The amount of correction of the ignition timing by knock control (in this embodiment, it is set as a retard angle iR from the basic ignition timing θBEE) is calculated in another routine described later, but the final ignition timing θ is determined in step 400 by θ=θBSt. It is determined as R.

Further, the amount of correction of the fuel injection time is determined based on this retard amount R. In other words, when the retardation amount R is large, the ignition timing is retarded and the exhaust temperature may become high (too high). Therefore, in this case, it is necessary to take a long injection time to maintain the air-fuel ratio. For example, the injection time correction amount may be determined in proportion to the retard IR.

The final ignition timing θ and final injection time thus determined are set in the timer of the microcomputer at step 500. After being set, the program returns to the main routine (step 600).

FIG. 12 shows the knock state detection and knock control routine that are the main focus of the present invention. In this embodiment, this interrupt routine is executed for each ignition cycle near the top dead center (TDC) of the engine.

Upon entering the interrupt routine at step 700, step 8 is performed to detect an average knock condition over a relatively large number of cycles in the manner described in the operation of the invention.
Execute 00. Details of this step 800 will be described later. Next, in step 900, it is determined whether or not the current knock control execution condition is present, and if the knock control execution condition is not present, the process returns to step 1400.

If it is under the knock control execution condition, the steering knob 1000 calculates a knock determination level ref for determining the presence or absence of knock for each cycle.

Although knocking can be detected separately using the method of the present invention, it requires a relatively large number of cycles to detect, so if knocking occurs frequently within a very short period of time, such as when the engine suddenly accelerates, In consideration of this, the conventional knock control system method of making a knock determination every cycle and retarding the ignition timing every cycle was also used.

This knock judgment level Vref is determined by the constant K determined for each cylinder (a value that converges to the median of the frequency distribution, details will be described later) obtained for each cylinder in step 800 (however, this The value will also be modified in the appropriate direction in the steps described below.)
It is created with a relationship of -VH to ref=. Therefore, although the knock determination level differs for each cylinder, knocks in all cylinders can be detected more accurately in this way. Note that the initial value of K is stored in the ROM as a two-dimensional map of engine speed N and cylinders.

Next, in step 1100, the knock intensity value ■ taken in as the value of the immediately preceding ignition cycle is compared in magnitude with the knock determination level V ref to perform a knock determination for each ignition. Depending on the result, the ignition timing is retarded and the IR
The step of calculating is 1200 (details will be described later). Next, in step 1300, it is determined whether the current knock determination level is appropriate or not, and this is constantly corrected in the appropriate direction (details will be described later).

After this, the process returns to step 1400.

Next, the stamp 800 for detecting a knock state will be explained in detail using FIG. 13. In step 801, the knock intensity value ■ taken in this time is compared with the above-mentioned upper threshold value VH corresponding to that cylinder, and V>VHI.
If so, increase the number of times NM exceeded by one (step 802).

Otherwise, the process moves to step 803, where the knock intensity value ■ is compared with the lower threshold value ■1. V<VH
In the case of step 804), the number of times the number has fallen below is increased by one (step 804), and the process moves to step 805. In other cases, the process directly proceeds to step 805. Next step 805
, the cumulative 50% point value VM is updated. The update method is as described above. In other words, if the knock intensity value ■ exceeds the current VM corresponding to that cylinder, (V
>VN ), Vs is increased by ΔVH, and conversely V<V
If it is H, decrease it by the same amount of Δ■A as ■8. By doing this, ■o is sequentially updated so as to always follow the cumulative 50% point.

Next, in step 806, it is determined whether it is time to detect a knock condition. That is, detection of the knock state is performed at predetermined intervals. In this embodiment, this interval is approximately 0.7 seconds. There are two reasons why we use time intervals instead of cycles. The first reason is to detect the knock state in a manner as close as possible to human sensory evaluation. Generally, when humans evaluate the state of knocking, they do so based on how many knocks occur within a certain period of time. In other words, the passage of engine cycles is an irrelevant concept to the driver of the car, and the only evaluation standard known to the driver is the time between them. By detecting the knock state on an hourly basis, it is possible to extract the best output and fuel efficiency without causing discomfort to the driver.

The second reason is the threshold value ■, which is executed after this.

■Related to update 8. As described above, in order to check whether the knock state is larger or smaller than the target value, it is necessary to set the lower threshold value VL to the same cumulative percentage point as the knock occurrence frequency in the target knock state. be. However, the target knock state is generally set smaller as the engine speed increases. Even if a relatively large amount of knock occurs in the low speed range of the engine, the engine will not be damaged, and on the contrary, output and fuel efficiency will be improved. However,
At high speeds, even in relatively infrequent knock conditions,
This may cause pre-ignition and damage the engine. For this reason, it is necessary to suppress the knock state to a relatively small level in the high speed range.

Therefore, as the speed increases, it is desirable to set the lower threshold value (2) to a smaller percentage accumulation point. This ■.

In order to set C to a small percentage, it is sufficient to increase the number of VL update cycles C, as already described in the operation section. If you update the VL every hour.

The number of cycles C during that time increases as the speed increases, so that the above-mentioned objective can be achieved.

Next, in step 807, it is determined whether engine conditions have suddenly changed during the detection interval. That is, it is checked whether the engine speed and the engine load have changed by a predetermined value or more within 0.7 seconds. In this embodiment, for example, the allowable change in engine speed is set to ±30° rpm, and the allowable change in engine load is set to ±153 nunHg in terms of pressure. If it is determined in step 807 that the state is quasi-steady, a knock state is detected in step 808.

Detection of this knocking state is performed by comparing the number of times N and NL for each cylinder. i.e. N8 and NL
If they are almost equal, it is determined that the knocking state of that cylinder is smaller than the target value, and N is determined.

If is still larger than NL even after considering the error, it is determined that the knock state is larger than the target value. This detected knock state is stored as a flag for each cylinder, and a knock state detection completion flag indicating that a knock state has been detected is set.

Next, in step 809, the upper threshold value V and the lower threshold value VL are updated. The update method is as follows. First, ■□ and VL are VH/
The relationship is Vs = VM /Vt. Assuming that this geometric number is A (■M/v+ = VM /v, =A), by updating A, it is possible to update ■A and VL. V
Since H=VM/A, if the knock intensity value ■ falls below VL even once during this detection interval (that is, if the NL counted in step 804 is 1
In the above case), by increasing A by ΔA, V
By decreasing L by ΔVL and, if NL=0, decreasing A by ΔA, VL is decreased by Δ■.

increase the height. By doing so, VL can be set to the required cumulative percentage point. And the upper threshold value ``A'' can be made by ``Work=A-VH4''. In this way, ■□ and VL are updated for each cylinder.

Next, in step 810, clear N, NL (N□
= 0, NL = 0), then the detection interval is 0.7
Clear the counter that generates sec (step 8
11) Do.

Next, using FIG. 14, a method for calculating the ignition timing retard amount R in step 1200 of FIG. 12 will be explained in detail.

First, in step 1201, if the knock intensity value V exceeds the knock determination level K. >.

This ΔR5 is also used in general knock control, and is normally about 0.5' to 2° CA, but in this embodiment, it is set to 1' CA. Next, in step 1203, a counter for the number of knock determinations is incremented. This counts the number of times it is considered a knock cycle, and is used as a basis for determining the knock determination level correction described later.

- In step 1201, if V≦K-VH, it is considered a non-knock cycle, and the process moves to step knob 1204. In step 1204, only when the non-knock cycle continues for a predetermined period of time (generally about 1 sec), the ignition timing is advanced by decreasing R by ΔR1. In other cases, the current retard amount R is maintained as is.

Next, in step 1205, it is determined in step 808 of FIG.
Check with the knock state detection completion flag. If it has not been detected yet (that is, during the above-mentioned detection interval), the process moves to step 1213, but if the detection completion flag is set, the magnitude of the knocking state is checked in step 1206. That is, if the flag indicating that the knock state is greater than the target value is set, step 1 is executed.
At 207, the retard angle IR is increased by ΔR2. This ΔR2 is different from the above-mentioned ΔR1 which determines knock and retards the ignition timing every cycle, but is a correction amount for correcting the ignition timing according to the average state of knock, so it is a smaller value than ΔR, for example 0. Approximately 5'CA is good. As a result, the retard amount R is slowly corrected in accordance with the average knocking state, and is also corrected in small increments based on this in order to quickly avoid knocking depending on whether or not knocking occurs in each cycle.

When the retardation amount R is set for each cylinder (that is, in the case of cylinder-specific ignition timing control), the retardation angle ff1R of that cylinder may be increased or decreased according to the knock state of that cylinder, as described above. In the case of all-cylinder ignition timing control, there is only one retard amount R, and in this case, it is sufficient to increase or decrease it by comprehensively determining the knock state of each cylinder. For example 4
In the case of a cylinder-based engine, it is possible to determine that the overall knock state is large when the knock state is large in two or more cylinders. Alternatively, instead of having the aforementioned NL and N for each cylinder, the overall knocking state can be directly known by counting the number of NL and N for each cylinder.

Even in this case, two threshold values VH and VH are provided for each cylinder, and comparisons are made for each cylinder and only the comparison results (i.e., NL
, N11 only) It is desirable to use it in common for all cylinders, but if it sacrifices the accuracy of detecting the knock state,
It is also possible to make VH and V□ common to all cylinders.

Next, in step 1208, the retard angle iR is compared with the maximum retard amount limit Rmax. In general, the retardation amount R is usually limited in order to avoid excessive ignition timing fluctuations due to knock control and to avoid an increase in exhaust temperature, but conventionally, this limit value Rmax is a predetermined fixed value. there were. However, in this example, this is learned based on the average knocking state. That is, the retardation amount R is set by the minimum retardation amount limit Rmin and the maximum retardation amount limit Rma.
Keep the ignition timing within the range x to prevent the ignition timing from being unnecessarily advanced or retarded. This is then learned to absorb seasonal changes, differences in gasoline properties, engine variations, etc. If the ignition timing is too retarded (that is, if R becomes too large)
This is a problem as the exhaust temperature rises, but if knocking occurs frequently, the engine may be damaged unless the engine is further retarded. In this case, priority is given to engine protection. It is also possible to detect this state and turn on the abnormality lamp.

Step 1208 is a step for determining whether the control range of the retard amount R should be modified to be more retarded. That is, if R is required to be larger than the maximum retard angle 51 Rmax and the average knock state is large, Rmax is set to ΔRmax in step 1209.
(for example, about 0.2'CA), and at the same time R
min to ΔRmin (this is also, for example, 0.2°CA
The limit range of R is corrected to a more retarded direction by increasing the angle by a certain amount (degree). For example, this control range is preferably about 56~7°CA (Rmax-Rmin#5°CA~7''CA)
.

Now, if it is determined in step 1206 that the knocking state is smaller than the target, the above-mentioned operation is reversed. That is, in step 1210, R is decreased (ie, corrected in the advance direction) due to the knock state, and in steps 1211 and 1212, the control range of the retard amount R is corrected in the advance direction.

In this way, the process moves to step 1213, where the retardation amount R is limited in the winter. In other words, if R>Rmax, replace R=Rmax, and if R<Rmin, replace R
= Replace with min.

Next, using Figure 15, step 130 of Figure 12
The process of automatically correcting the knock determination level used for knock determination for each cycle in 0 in the appropriate direction will be described.

First, in step 1301, it is checked by the knock state detection completion flag whether or not a knock state has been detected. If detected, the magnitude of the knocking state is checked in step 1302. If it is determined that the knock state is larger than the target, step 13
In step 03, the number of knock determinations in each cycle so far is checked (this number of determinations is determined in step 1203 in FIG. 14).
).

If the number of knock determinations is less than a predetermined value, it is determined that the knock determination is too high, and K is decreased by Δ in order to lower the knock determination level (step 1304). That is, when the knocking state is large and the number of knock determinations is small, the knock determination level is too high and the knock determination is incorrect. In this case, if the number of knock determinations is relatively large, K is held as is because the knock determination is normal.

If it is determined in step 1302 that the knocking state is smaller than the target, it is checked in step 1305 whether the number of knock determinations is too large.

If the number of knock determinations is too large, K is increased by Δ to correct the knock determination level to a higher level (step 1306).

In other words, if the knock state is small and the number of knock determinations is too large, the knock determination level is too low and the noise cycle is erroneously determined to be a knock. If the number of knock determinations is small, this simply means that knock has not actually occurred in the engine, so it is unclear whether the knock determination level is appropriate (it may be appropriate or it may be too high).

Therefore, in this case, K is held as is.

This modification of the knock determination level is performed for each cylinder.

That is, based on the knock state of each cylinder and the number of knock determinations for each cylinder, K corresponding to each cylinder is corrected in an appropriate direction.

Of course, using the same concept as described in the calculation method of the retard amount R, it is also possible to correct K by using one common value instead of having it for each cylinder. Also,
Determination of the number of knock determinations (steps 1303 and 130)
5) may be judged as "0", rlJ-wise, as if there was a knock judgment during that time or not.

Next, in step 1307, the number of knock determinations is cleared, and then the knock state detection completion flag is also cleared (step 1308>).

As described above, knock control is performed.

In the above embodiment, the average knock condition is detected, and based on this, the ignition timing is manipulated, the ignition timing control range is modified, and the knock determination level for determining knock for each cycle is modified. All can be effective alone. For example, for detecting a knock state, this can be applied as is to a non-common system or a gasoline octane number detection system. As an example applicable to actual cars, it can also be used to display a warning light when knocking becomes unusually likely due to gasoline properties or some other cause.

Regarding the operation of the ignition timing, in the above embodiment, the ignition timing is operated for each ignition according to the knock judgment result for each cycle, and the ignition timing is operated at a longer cycle depending on the average knock condition. Adopts a double loop with. This is in consideration of the transient state of the engine in which a sudden change in the knock state is expected. However, the average knock state is as in this embodiment, for example l sec
Since the ignition timing can be calculated within 100 seconds, it is also possible to correct the ignition timing only depending on the knock state, without having to correct the ignition timing for each ignition. In this case, there is no need for a knock determination level for determining the presence or absence of knock for each cycle.

It is also possible to configure a simple knock avoidance system that detects knock conditions over a longer period of time and thereby switches the ignition timing characteristics to two or more stages using multiple maps. .

In addition, in the above embodiment, ignition timing is treated as a factor for controlling knock, but this is due to the air-fuel ratio, intake pressure (
For example, it can be widely applied to knock control factors such as turbocharging pressure, EGR, anti-knock agent, etc. In this case, immediate knock control factors such as ignition timing are controlled according to the knock judgment results for each cycle, and the average knock control factors such as boost pressure are controlled according to the detection results of the average knock state. Various arrangements such as controlling the factors that influence the state become possible. Of course, only one of the ignition timing, air-fuel ratio, intake pressure, EGR, anti-knock agent, etc. may be controlled.

Further, as for modifying the control range of knock control factors such as ignition timing, it is also possible to perform only this in accordance with the knock state. That is, the control can be performed only according to the knock determination result for each cycle, and only the control range can be modified according to the average knock state.

Further, regarding the modification of the knock determination level, it is not necessarily necessary to commonly use the determined signal determined for each cycle and the knock intensity value for detecting the knock state (that is, the signal integral value in the embodiment). For example, the maximum peak value VNAX of the signal is used as the signal to be judged for knocking, and the integral value rVNAX is used to detect the average knock state.
You may also use In this case, the VMA that becomes the signal to be determined
There is no need to know the value of X directly. In other words, only information as to whether V MAX is larger or smaller than the knock determination level is required using a comparison circuit or the like. However, it is more efficient to use them in common as in the above embodiment.

Further, in the above embodiment, two threshold values VH and V
Although the relationship of □ is set to be approximately geometrical with respect to the median value M of the distribution (50% point of the cumulative distribution), some error occurs in the approximate value ■A that should converge to the median value M. Therefore, VL=VM4/A.

v, = (A+D) - In some cases, it is better to add some offset D as shown in ■9. Since this offset D can be used as an element for finely adjusting the target knocking sound, it can take either a positive or negative sign, but it is essentially desirable that D be a positive value. This is because, in order to obtain vM, the knock intensity value V is increased or decreased by the same amount Δ■8 to converge to the median value M, but even if it converges, the median value M
It fluctuates by the same amount around . Therefore, if you look at this on a logarithmic axis, there will inevitably be a large amount of deviation to the lower side. Therefore, it is desirable to make D positive in order to correct this. Or, to correct this from the beginning, raise the VH a little more than cumulative 50% (for example, cumulatively 55%).
%) position can also be converged by self-loss. In this case, the ratio of increase/decrease in V14 may be set to, for example, 55:45.

Further, in the above embodiment, the lower threshold value ■.

is corrected by applying a feedbank so that the cumulative percentage point is the same as the knock occurrence frequency in the target knock state, but if the distribution of knock intensity values can be predicted to some extent, or if the knock state can be determined with such accuracy. If there is no need for detection, a fixed value may be used. Therefore, in this case, V□ is also a fixed value, and there is no particular need to find the median value M of the distribution. However, it is clear that the above embodiment is significantly superior.

In addition, in the above embodiment, the average knocking state is detected by comparing the number of times the upper threshold value ■8 is exceeded and the number of times the lower threshold value VL is lowered. It is also possible to detect a knock condition.

That is, calculate the cumulative α% point, Lα, when accumulating from the lowest value, the cumulative α% point, I]α, when accumulating from the highest value, and the median value VM of the distribution. However, by examining the magnitude relationship between VM/Lα and Hα/■□, it is possible to determine whether the knock state is greater or less than the target knock occurrence frequency α%.

That is, in a small knock state where the knock occurrence frequency is lower than α%, Lα, VH, and Hα are aligned on a straight line on the lognormal probability paper. Therefore, in this case, Lα and Hα are ■. Since there is a geometric relationship with respect to Vs/Lα and Hα/
■Oh is almost equal.

On the other hand, in a large knock state where the knock frequency is higher than α%, this geometrical relationship breaks down, and Hα/■ is better for VM.
/Lα. In addition, how to find Hα and Lα is as follows:
As is clear from the disclosure of the operation part of the invention, the knock strength value V is compared with Lα and Hα during a predetermined number of cycles C, and if
α is decreased by △Lα, and if it is less than the predetermined number of times, Lα
increase by △Lα, and at the same time, if ■ exceeds Hα by the predetermined number of times or more, increase Hα by △Hα,
If the number of times is less than the predetermined number, Lα and Hα are converged to a desired cumulative percentage point by decreasing Hα by ΔHα.

Further, in the above embodiment, a method and apparatus that do not use a logarithmic converter have been shown, but the use of a logarithmic converter or software that performs logarithmic conversion instead is acceptable even considering cost and other constraints. If so, further various methods and devices can be considered. For example, it is also possible to detect the knock state based on the magnitude relationship between the average value and the median value of the logarithmically transformed value log V of the knock intensity value ■. That is, when the knock state is small, the distribution of log V is approximately a normal distribution, so the median value and the average value of the normal distribution match. However, as the knocking state increases, the average value becomes larger than the median value, making it possible to identify the knocking state.

Further, in the above embodiment, the rectification and integral value (2) in a predetermined section of the sensor signal is used as the knock intensity value (2), but other values corresponding to the maximum wave height value may be used. This has already been explained in detail in the operation of the invention.

Furthermore, in the above embodiment, a type of sensor that detects engine block vibration was used as the knock sensor.
As already explained in detail, a microphone, cylinder pressure sensor, combustion light sensor, etc. can be used for this purpose.

Furthermore, as already explained, the assumed distribution is not necessarily a lognormal distribution, but can be approximated as a lognormal distribution by applying appropriate transformations such as a binomial distribution or a gamma distribution. The above embodiment can be applied as is to any distribution that poses a problem, and it is easier to handle a distribution that is a normal distribution instead of a logarithm.

In other words, it is sufficient to replace the geometric relationship with an arithmetic relationship.

Further, as long as the calculation time and equipment cost are acceptable, it is possible to identify the knock state for any distribution. That is, it is sufficient to assume a single knock-free distribution using only relatively low-value data, and see where the high-value data deviates from this distribution.

〔Effect of the invention〕

As described in detail above, the first invention of the present application detects the average knock state by determining the distribution shape when sampling the knock intensity value V corresponding to the maximum wave height value over multiple cycles. Therefore, the true knock condition can be detected with high accuracy without being affected by variations in the knock sensor or engine, changes over time, etc., which have been difficult in the past.

Further, in the second invention, knock control factors such as ignition timing, air-fuel ratio, and intake pressure are controlled according to the result of the average knock state, so that knock control factors such as ignition timing, air-fuel ratio, and intake pressure are controlled. The engine can be maintained in the target knocking state.

Furthermore, the third invention corrects the knock judgment level according to the detection result of the average knock state, and controls the knock control factors according to the presence or absence of knock in each combustion cycle. Not only can the knocking state be maintained, but also the knocking property can be suppressed with good responsiveness even when sudden knocking occurs during engine transient conditions.

Further, the fourth invention controls the knock control factor according to the presence or absence of knock in each combustion cycle, and limits the control term range of the knock control factor according to the detection result of the average knock state. Therefore, it is possible to prevent knock control factors from changing unnecessarily.

[Brief explanation of drawings]

Figure 1 is a diagram showing the maximum value of the knock signal and its distribution state, Figure 2 is a diagram showing the cumulative distribution of the maximum value of the knock signal and this distribution on lognormal probability paper, and Figure 3 is a diagram showing the maximum value of the knock signal and its distribution state. Figure 4 is a diagram showing the cumulative distribution of the integral value on log-normal probability paper. Figure 4 is a diagram showing the cumulative distribution of the maximum value of the knock signal on log-normal probability paper under the same knocking condition as Figure 3. A diagram showing the distribution of knock strength values between knock cycles and non-knock cycles. Figure 6 is a diagram for comparing the distribution of knock strength values when the knock is smaller and larger than the target. Figure 7 is a diagram showing the distribution of knock strength values according to the present invention. FIG. 8 is a detailed configuration diagram of the control circuit in FIG. 7, FIG. 9 is a detailed configuration diagram of the knock intensity value detection circuit in FIG. 8, and FIG. 10 is a detailed configuration diagram of the knock intensity value detection circuit in FIG. 8. Signal waveform diagrams for explaining the operation of the value detection circuit, and FIGS. 11 to 15 are flowcharts showing knock detection and control procedures in the control circuit. 1...engine, 3...igniter 1 ignition coil. 4...knock sensor, 7...injector, 8...
・Turbocharger, 9...0□sensor, 10...
Control circuit, 10-1...CPU, 10-2...
ROM, 10-3...RAM, 10-10...
Knock strength value detection circuit. 10-10-2... Rectifier, 10-10-3... Integrator.

Claims (17)

[Claims] (1) A knock sensor for detecting engine knock; a knock intensity value detection means for extracting a knock intensity value V corresponding to the maximum peak value of this signal from the signal of this knock sensor; The present invention is characterized by being equipped with a discriminating means for discriminating whether the distribution shape obtained when cycle sampling is a predetermined distribution shape or not, and detecting an average knocking state over a plurality of cycles according to the result of this discrimination. Knocking detection device for internal combustion engines. (2) The above-mentioned knock intensity value V is a value obtained by rectifying and integrating knock vibration output containing frequency components unique to knock within a predetermined interval, or a value obtained by comparing the knock vibration output and a predetermined level within a predetermined interval. Claim 1, characterized in that the number of pulses in a pulse train, the integrated value of the pulse train, or the effective value of knock vibration output.
Knocking detection device as described in . (3) The predetermined distribution has a relatively small knock intensity value V
A knock state is detected by determining the shape of the distribution according to the cumulative percentage from which the actual cumulative frequency distribution starts to differ from the cumulative frequency distribution of this distribution. A knocking detection device according to claim 1 or 2, characterized in that: (4) The knocking detection device according to any one of claims 1 to 3, wherein the predetermined distribution is a lognormal distribution. (5) The determining means is configured to select a first threshold value V_L set to a relatively low side and a second threshold value V_L set to a relatively high side within a range of values that the knock intensity value V can take. Threshold V
_H, compares the knock intensity value V and each threshold value over a plurality of cycles, and determines the number of times the knock intensity value V falls below the first threshold value V_L and the knock intensity value V
The number of times that the distribution exceeds the second threshold value V_H is counted, and the distribution shape is substantially determined based on the magnitude relationship of each counted value. The knocking detection device according to any one of Item 4. (6) With respect to the cumulative 50% point M of the distribution obtained when the knock intensity value V is sampled multiple times,
The first threshold value V_L is set to be smaller than the cumulative 50% point M, and the second threshold value V_L is set to be smaller than the cumulative 50% point M.
6. The knocking detection device according to claim 5, wherein H is set to be larger than this point M. (7) The determining means has means for estimating the cumulative 50% point M based on the values of the knock intensity values V that are input one after another, and for the estimated value V_M of this point M, the first
The threshold value V_L and the second threshold value V_H are in a nearly geometrical ratio (V_M/V_L≒V_H/V_M), or a relationship in which V_H is set to a slightly larger side than the geometrical ratio (V
_M/V_L+D=V_H/V_M, D>0) The knocking detection device according to claim 5 or 6, wherein (8) The estimated value V_M is a variable that is updated sequentially every cycle, and each time the knock intensity value V is input, V and V
If the knock strength value V is larger than _M, then V
Increase _M by ΔV_M, and if it is smaller, V_M
By decreasing V_M by the same amount ΔV_M, V_M is sequentially updated so as to converge to the cumulative 50% point M, and the first threshold value V_L using this value of V_M and a predetermined geometric ratio A is set. V_L=V_M/A, the second threshold V
_M as V_H=V_M・A or V_H=V_M・(
The knocking detection device according to claim 6 or 7, wherein the knocking detection device is set according to the relationship A+D) (D>0). (9) The geometric ratio A is a variable that is updated every predetermined number of cycles C, and during this number of cycles C, the knock intensity value V and the first threshold value V_L (=V_M/A) are compared, and if the knock strength value V is below V_L by a predetermined number of times or more, the geometric ratio value A is slightly increased to correct V_L to a lower value, and if it is less than the predetermined number of times, V_L is lowered by a small amount.
In order to correct L to a higher value, the geometric ratio A is slightly decreased so that the probability that the knock intensity value V falls below the first threshold value V_L becomes approximately equal to the desired probability. 9. The knocking detection device according to claim 8, wherein the geometric ratio A is adjusted. (10) The cycle number C is changed according to the engine conditions, so that the probability that the knock intensity value V falls below the first threshold value V_L changes depending on the engine conditions. A knocking detection device according to claim 9. (11) The cycle C takes on a larger value as the engine speed increases, so that the probability that the knock intensity value V becomes substantially lower than the first threshold value V_L is
11. The knocking detection device according to claim 10, wherein the knocking detection device is configured to gradually decrease as the engine rotational speed increases. (12) The knocking detection device according to claim 11, wherein the number of cycles C is determined so that the cycle elapsed time is substantially constant. (13) The discrimination means is configured to determine that the knock intensity value V is a cumulative 50% of the distribution obtained when sampling the knock intensity value V in multiple cycles.
The method of calculating the value V_M at point M using the knock intensity values V that are input one after another, and the method that corresponds to the desired cumulative α% when the values are accumulated from the lowest value using the same knock intensity values V. The means for determining the cumulative α% point Lα for
means for estimating point Hα, and V_M/Lα and Hα/
The knocking detection device according to any one of claims 1 to 4, characterized in that the distribution shape is substantially determined based on the magnitude relationship with V_M. (14) The cumulative α% points Lα and Hα are updated every predetermined number of cycles C, and the knock intensity value V is compared with Lα and Hα during the predetermined number of cycles C.
If Lα falls below Lα a predetermined number of times or more, Lα becomes ΔLα
If the number of times is less than a predetermined number, Lα is increased by ΔLα. At the same time, if the knock intensity value V exceeds Hα by the predetermined number of times or more, Hα is increased by ΔHα. 14. The knocking device according to claim 13, wherein is converged to a desired cumulative percentage point by decreasing Hα by ΔHα. (15) a knock sensor for detecting engine knock; a knock intensity value detection means for extracting a knock intensity value V corresponding to the maximum peak value of the signal from the signal of the knock sensor; A discriminating means for discriminating whether the distribution shape obtained when cycle sampling is a predetermined distribution shape; 1. A knocking control device comprising: control means for controlling knock control factors such as ignition timing, air-fuel ratio, and intake pressure. (16) a knock sensor for detecting engine knock; a knock intensity value detection means for extracting a knock intensity value V corresponding to the maximum peak value of the signal from the signal of the knock sensor; A determination means for determining whether the distribution shape obtained when cycle sampling is a predetermined distribution shape, and a method for detecting an average knock condition over multiple cycles according to the determination result and making a knock determination for each cycle. a knock determination means for determining the presence or absence of knock based on the knock determination level for each cycle based on the signal of the knock sensor; 1. A knocking control device comprising: control means for controlling knock control factors such as engine ignition timing, air-fuel ratio, and intake pressure. (17) a knock sensor for detecting engine knock; a knock intensity value detection means for extracting a knock intensity value V corresponding to the maximum peak value of the signal from the signal of the knock sensor; a determining means for determining whether a distribution shape obtained when cycle sampling is a predetermined distribution shape; a knock determining means for determining the presence or absence of knock for each cycle based on the signal of the knock sensor; Depending on the judgment result, the engine's ignition timing or air-fuel ratio,
A knocking device comprising: a control means for controlling a knock control factor such as an intake pressure; and a limiting means for adding a limit to a control value range of the knock control factor according to a determination result of the discrimination means. Control device.