JPH01178773A - Knocking detection method for gasoline engine - Google Patents

Abstract

PURPOSE:To enhance the extent of reliability for knocking detection by taking knock information at a high frequency out of the much knock information, comparing the vibro-amplitude with a threshold value, and also comparing the compared result with that at the last time ignition in addition. CONSTITUTION:A knock detector 1 passes only a high frequency band of more than 10kHz in outputs from a knock sensor 1a with a band pass filter 1b, cutting a noise of valve seating or the like with a master circuit 1c, and inputs a peak value in one stroke into an electronic control unit 23 by way of a peak hold 1d. Another electronic control unit 27 compares vibro-amplitude of a peak value of knock information with a specified threshold value, and further compares the compared result with that based on the last time ignition, detecting the difference, and according to this difference, ignition timing is controlled for its delay or advance. Thus, such knocking detection and control as high in reliability can be performed.

Description

DETAILED DESCRIPTION OF THE INVENTION <Industrial Application Field> The present invention relates to a method for detecting knocking in a gasoline engine, and more particularly, to a method for detecting knocking that is more reliable than the conventional method.

<Prior Art> Knocking in a gasoline engine causes a decrease in output and thermal efficiency and damage to the engine, so it is necessary to prevent it. This knocking can be prevented by increasing the octane number of the fuel or by controlling the ignition advance angle of the engine. To explain this latter control a bit, a knock sensor such as a vibration acceleration sensor or a spark plug seat pressure detection sensor is attached to the engine to detect the occurrence of knock, while at the same time advancing the ignition timing as much as possible with the premise of suppressing the occurrence of knock. The engine output is adjusted to the appropriate angle, but when knock occurs, the advance amount is controlled to delay the ignition timing based on the knock information.

In other words, in conventional knock control, the amount of retard is calculated according to the knock information from the knock sensor, the basic advance amount is retarded by this retard amount to suppress the occurrence of knock, and when the knock is resolved, the amount of advance is changed back to the original value. It is controlled to return to

In the case of B, knock is identified. Since there are many types of engine noise in the output of the knock sensor, the noise level from this knock sensor is averaged, and vibrations exceeding this average value are considered to be knock. There is. However, since this noise level is unstable, erroneous identification due to noise fluctuations can be reduced to some extent by providing some dead band width for the noise level.

<Problems to be Solved by the Invention> However, even if the width of the dead band is set as above, knock identification is still insufficient.If the width of the dead band is increased, knock identification will be impaired, and if it is made smaller, erroneous identification will occur. It ends up being too many.

Therefore, setting the dead zone width is not a fundamental solution.

A band pass filter (BPF) is provided after the knock sensor to pass the knock frequency.
When the passband of F is set to a low frequency of 10k[iz or less,
Knock identification is extremely difficult due to the many other types of engine noise mentioned above, but on the other hand, the BPF passband is set to 10-
It has been found that when using a higher frequency, the output of this BPF is limited to valve seating noise in addition to knocking. In order to distinguish between this knock and valve seating noise, a method of measuring based on the crank angle may be considered, but in multi-cylinder engines, the timing of knocking and valve seating noise may coincide, making complete knock identification impossible. is not possible.

Therefore, in view of the conventional problems, the present invention provides a knocking detection method that is more accurate than the conventional method.

<Means for Solving the Problems> The present invention achieves the above-mentioned object by extracting knock information in a high frequency band from knock information provided by a knock sensor,
The vibration amplitude of this high-frequency knock information is compared with a threshold value,
This comparison result is further compared with the comparison result based on the previous ignition to detect the difference.

Points of focus and effects> Figure 2 shows the output waveform of a BPF of 10 positives or more, but as can be seen from Figure 2, even with fairly strong knocking, it only occurs once every few to several strokes. Moreover, it does not occur continuously in the same cylinder. On the other hand, mechanical noise such as valve seating noise has periodicity, and the increase and decrease of the noise is not rapid but continues in the same cylinder.

Therefore, knocking or other mechanical noise can be identified by discontinuities.

For this reason, it is possible to determine whether or not there is a knock by passing a high frequency wave through the BPF downstream of the knock sensor, then detecting a relatively large vibration and comparing it with the vibration detected in the previous process.

EMBODIMENTS> The present invention will now be described in detail with reference to FIGS. 1 and 3 to 7. Figure 1 shows the electronic control unit (ECU)
) 23, its inputs and outputs are shown, and E
The main part of the CU23 includes a CPU27, and this CPU2
Interfaces 28, 29 and an A/D converter 30 are arranged on the input side of the engine 7, and a fuel injection driver 34 and an ignition driver 53 are arranged on the output side. Furthermore, E
A ROM 31 in the CU 23 that stores control program data and preset fixed data, a RAM 32 that can be updated and sequentially rewritten, and a BURAM 33 connected to the battery 24 are arranged.

On the input side of the ECU 23, a throttle sensor 14, a battery sensor 24, a water temperature sensor 19, an intake air temperature sensor 12 .
0□ sensor 17 and atmospheric pressure sensor 13 are connected to an interface 28, a key switch 26 and an idle switch 15 are connected to an interface 29, and a crank angle sensor 21 (engine speed (also serves as a rotational speed sensor to detect
The C sensor 22 and the air flow sensor 11 that is located in the intake passage and outputs intake air amount information are connected to each other.Cp ty 27
This is the input.

On the other hand, on the output side of the ECU 23, the injector driver 3
4 is connected to the injector 8, and the ignition coil 51 is 1 m long.
An ignition driver 53 is connected to an igniter 52 that continues to stir the engine using a power transistor. In addition, the ignition coil 51
is connected to the spark plug 18 via a distributor 50.

Now, the knock detector 1 connected to the interface 28 on the input side of the ECU 23 is the knock sensor la, 10k)b.
For example, B allows a knock frequency of 161 dtz to pass through.
PFlb, a mask circuit 1c that cuts noise other than the reason why knocking occurs.

It has a peak hold 1d that holds the peak value of the amplitude by the knock sensor 1a within one stroke, and this detector 1 outputs the peak value of the one stroke. When this peak value is read into the CPU 27, the peak hold 1d is reset. Further, the mask circuit 1c differentiates and discriminates the signal of the ignition driver 53 mentioned above, shapes it with a monostable multi-vibration brake, and performs masking for a predetermined period of time after ignition. Furthermore, the filter gain of BPFlb is changed by changing the gain depending on the engine rotation speed, and is switched so that the knock sensor output corresponds to the determination level.

Here, in Fig. 3, the sensor output (a), BPF (bl, mask output (C), peak hold output (dl), and reset signal tel are respectively shown. The peak value is extracted from the sine waveform in Figure 3 fb) from which the
It is held as shown in the figure (dl), and the hold is released by resetting. Therefore, the largest peak value in that stroke is input into the ECU 23.

Next, after this peak value is read into the ECU 23, the ignition reference signal interrupt process as shown in FIG. 4 (81) is performed. That is, by turning on the key switch 26 shown in FIG. The CPU 27 enters the control operation. Then, when the engine starts rotating, the CPU 27 enters an interrupt process in which a knock control routine is performed in place of the main routine every 1806 revolutions of the crank angle. Similarly, as shown in FIG. 4(b), Then, an interrupt process is started in which a timer routine is executed in place of the main routine every predetermined time period.

In the ignition reference signal interrupt process, ie, the knock control routine of FIG. 4 fal, cylinder identification process is performed in step A1.

This N process can determine the combustion cycle of each cylinder based on the signal from the crank angle sensor 21 shown in FIG. Step A
After cylinder identification in step A1, the peak value P of the Nth cylinder is determined in step A2. (1"I'l is read and at the same time, the difference ΔPH between this N-th cylinder's peak value P in the previous stroke and -1(N) is calculated.

Next, the process moves to step A3, and the peak value Pn(N) of the current process of the Nth cylinder is compared with the peak determination value xI. This peak judgment value xP is an engine rotation dependent setting value, and E
, CU 2.3 has a built-in map of peak judgment values xP corresponding to the engine speed. As a result of the comparison in step A3 of B, if Xp≦P, (goods), proceed to step A9 described later, but if P,,H>XP, step A
Move on to 4. In this step A4, the above-mentioned difference ΔP (goods) and the peak change determination constant Xdl are compared. In other words, it is used to determine whether the difference in peak value between the previous stroke and the current stroke of the Nth cylinder is greater than a certain value. If this is not likely, it is determined to be knocking because it is sudden.

In step A5, the difference ΔP(
N) is also equal to or greater than a constant value xd, the amplitude difference ■ determined to be the knock value is calculated. That is, V-PON-xP is calculated. Next, this difference ■ is multiplied by a constant K for calculating the retard amount to obtain KV, and KV is added to the negative of the previous retard to update the current retard amount. Then step A
7, the current retard amount ΔQ and the maximum retard amount ΔQ□.

8, and if ΔQ>ΔQ, x, ΔQ=ΔQ□ in step A8. 8.

After this, ΔQ≦ΔQ□ in step A8 and step A7.

8. P at step A3, (goods) ≦xP1 step A4
For each case of ΔP(goods)≦Xdp, the process moves to step A9 to calculate Q, .

In the process at step A9, the retard amount ΔQ at step 86 and each condition Q8 in electronic advance angle control. Q,,J
□p QA-r is added. These conditions QB, Q1.
IT and QA are calculated in the main routine. Here, an electronic advance angle block diagram is shown in FIG. In Fig. 5, the base advance angle Q is determined by the intake air amount A/N per stroke and the engine speed, the water temperature correction amount Q is determined by the water temperature,..., lT1 The intake temperature correction amount QA is determined by the lucid air temperature.
The advance angle amount Q, 9 is obtained by adding the retard amount ΔQ determined by the process up to step 86 in FIG. It is output to the device, and in step A11l, the previous peak value P of the Nth cylinder is rewritten by -1 (Nl), and in step A12, peak hold reset processing is performed.

FIG. 6 shows a second embodiment, and the same parts as those in FIG. In the knock detector 1, a half-wave rectification and amplification circuit 1f is arranged after the mask circuit IC, and this circuit 1f performs half-wave rectification and amplification.

Next, the averaging circuit 1g averages the noise and sets a threshold value. If this value is taken as a BGL value, the gain of this BGL value is switched and the BGL value is input to the comparator 1h. The comparator 1h compares it with the output of the mask circuit IC,
That is, the BGL value and the sensor waveform are compared, and signal values greater than or equal to the BGL value are integrated. On the other hand, the output of the comparator 1h is fed back to the half-wave rectifier circuit 1f, which cuts signals exceeding the BGL value to suppress an increase in the BGL value.

FIG. 7 shows waveforms of various parts of the knock detector 1 shown in FIG. Fig. 7(a) shows the knock sensor output waveform including noise, Fig. 7(b) shows the output waveform of BPFlb, Fig. 7(C
) is the mask signal, FIG. 7 (dl is the half-wave rectification and averaging output, FIG. 7 (el is the comparator output, FIG. 7 (fl is the reset signal, and FIG. 7 (gl is the integrated output signal). As a result,
In the hardware shown in FIG. 6, steps up to step A3 are performed up to comparing the peak value 1 shown in FIG. 4 (al), (N) and a constant value, here the average output. therefore,
The software starts with the step shown in FIG. 7 (taking the difference between the integrated output of fl and the output based on the previous ignition).

1 above. In the second embodiment, when knock is detected in one cylinder, knock retard is performed on the cylinder, but ignition timing due to knock is performed only on the cylinder in which knock is detected. You can.

<Effects of the Invention> As explained above, according to the present invention, in each cylinder,
By being able to pick up the sudden signal that is characteristic of a knock signal and distinguish it from noise, knocking can be detected more reliably than in the past.

[Brief explanation of the drawing]

1, 3 to 7 are explanations of embodiments of the present invention, in which FIG. 1 is a first configuration diagram, and FIG. 3 is a waveform diagram of each part of the knock detector 1 shown in FIG. Figure 4 is a flowchart of the ignition reference signal interrupt processing and the flowchart of the timer routine, Figure 5 is a basic block diagram of the electronic advance angle, Figure 6 is the second configuration diagram, and Figure 7 is the knock detector of Figure 6. 1 and 2 are waveform diagrams for explaining the principle of knock identification. In the figure, 1 is a knock detector, 1a is a knock sensor, 1b is a band pass filter, IC is a mask circuit, 1d is a peak hold circuit, 1f is a half-wave rectifier amplifier circuit, 1g is an averaging circuit, and 1h is a comparator. , 11 is an integrator, and A1-A12 are processing steps.

Claims (1)

[Claims] Among the knock information from the knock sensor, the knock information in the high frequency band is extracted, the vibration amplitude of this high frequency knock information is compared with a threshold value, and this comparison result is further compared with the comparison result based on the previous ignition to detect the difference. How to detect knocking in a gasoline engine.