JPH04326036A - Knocking detection method - Google Patents

Abstract

PURPOSE:To enable a frequency analysis error due to input cut of a signal to be reduced in a knocking detection device which performs frequency analysis of a knocking sensor signal in real time and performs knocking detection. CONSTITUTION:In a knocking detection device with a means for performing frequency analysis of a knock sensor signal in real time and a means for judging presence or absence of knocking according to a size of an obtained spectrum value, an error in frequency analysis is reduced by performing weighting to a time-series data of a inputted knock sensor signal and then matching both signal levels within the frequency analysis section forcibly, thus enabling the error in frequency analysis due to input cut of signals to be reduced and hence detection accuracy of knocking to be improved.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for detecting engine knocking equipped with knocking detection means.

0002

2. Description of the Related Art A conventional knocking detection method, as disclosed in Japanese Patent Application Laid-open No. 79827/1983, involves Fourier transforming the signal of a knock sensor during a predetermined period.
There is a device that detects knocking.

0003

Generally, in frequency analysis such as Fourier transform, a signal within a predetermined period is extracted and analyzed, and the analysis result matches the spectrum of a signal in which the signal of the predetermined period is repeated. Therefore, if the first signal level and the last signal level of the analysis period do not match, the obtained spectrum will deviate from the spectrum of the original signal, which will cause deterioration in the knocking detection accuracy. Become.

[0004] The above-mentioned conventional technology does not take into account the influence of errors in frequency analysis accompanying signal extraction, and this may impair knocking detection accuracy.

An object of the present invention is to reduce errors in frequency analysis and ensure knocking detection accuracy.

[0006]

[Means for Solving the Problems] In order to achieve the above object, there is provided a means for detecting engine vibration, and a signal output of the vibration detecting means is captured as data at each predetermined sampling period, and each data of the captured signal output is provided. The present invention includes means for weighting the data, analysis means for frequency-analyzing the weighted data string to obtain a spectrum, and means for determining the presence or absence of knocking.

[0007]

[Operation] Engine vibration is detected by the vibration detection means, and its signal output is taken in as data at every predetermined sampling period. The captured data is divided into one or more data strings for each predetermined number of data, and after weighting each data in these data strings, frequency analysis is performed on each weighted data string, and the spectrum is calculated. The presence or absence of knocking is determined by comparing the magnitude of the obtained spectrum with a predetermined determination level.

[0008]

[Example] A first example of the present invention will be described. Figure 1 illustrates the effect of weighting vibration sensor signals, where (a) is the original vibration sensor signal and (b) is a part of the vibration sensor signal extracted for frequency analysis. . Generally, in frequency analysis, if the signal levels at both ends of the extracted waveform do not match, the frequency analysis result (spectrum) will have an error as shown in (c). In (d), the signal levels at both ends are forced to match by weighting the signal waveform; this is (b)
It is obtained by multiplying the waveform of and the weighting function of (e). The weighting function takes a value closer to zero as it approaches the periphery, and by weighting the signal using this weighting function and forcing the signal levels at both ends to match, the error in the spectrum can be reduced as shown in (f). can be reduced. Here, it is better for the weighting function to have fewer high-frequency components, but since there is a problem that the level of peripheral data decreases, it is necessary to select an appropriate weighting function.

FIG. 2 shows the configuration of the engine control device. The engine is taken into the combustion chamber through an air cleaner 1, a hot wire flow meter 2, a throttle valve 5, an intake manifold 6, and an intake valve. Water temperature sensor 17 near the combustion chamber
Then, install the knock sensor 18.

[0010]The engine crankshaft is equipped with a position sensor 12-1 for cylinder discrimination and a reference sensor 1.
Attach 2-2. The engine rotation speed is measured by the number of position signals per unit time. The knock sensor detects knocking vibrations within the combustion chamber as vibrations of the cylinder block, but in addition to this, a cylinder pressure sensor 19 that directly detects pressure vibrations within the combustion chamber may be used. The output voltage, reference signal, position signal, water temperature sensor signal, and knock sensor signal of the hot wire flowmeter are input to the control unit 9. The control unit calculates the fuel injection amount and ignition timing from these input signals, and the results are used as the drive signal for the injector 16 and the ignition coil 1, respectively.
It is converted into a control signal of No. 3 and output.

The ignition plug 15 ignites the air-fuel mixture in the combustion chamber by the high voltage generated by the ignition coil.

The signal acquisition operation in the control unit 9 will be explained with reference to FIG.

[0013] Figure (a) shows the configuration of the signal acquisition section. The POS counter 21 uses P as a reference signal.
The crank angle is measured by counting the number of OS signals. The CPU 20 writes A to the compare register.
/D conversion start angle is set. The A/D converter 23 converts the knock sensor signal into an A/D converter at a predetermined sampling period.
Convert. RAM 24 stores A/D converted data.

The operation timing will be explained with reference to FIG. When the POS counter is incremented and matches the value of the compare register corresponding to the start angle of A/D conversion, a match interrupt is generated in the CPU. A/D conversion is started upon occurrence of a coincidence interrupt, and the converted data is sequentially stored in the RAM 24 for frequency analysis. A
/D conversion ends after being performed a predetermined number of times. A/D here
The number of conversions is set to include the period during which knocking occurs.

The operation of the CPU 20 will be explained using the flowchart shown in FIG. In step 100, the knock sensor signal is captured as data and stored in the RAM. In step 110, each data is corrected using a weighting coefficient by the following calculation.

data i = w i ×
data i i =1~L
...(Equation 1) w i: Weighting coefficient i: Data number L: Number of analysis points Here, the weighting coefficient is set in advance as a table as shown in Figure (b) corresponding to each data, and is The more peripheral the data, the closer the weighting coefficient is to zero. As a result, the levels of the first point and the last point of the analysis section match, and errors in frequency analysis can be reduced.

In step 120, frequency analysis is performed on the corrected data to obtain a spectrum for each frequency. In step 130, for example, several spectra corresponding to characteristic frequencies of knocking are selected as an index representing the magnitude of knocking, their sum is set as S, their average value is set as N, and the S/N is determined. In step 130, the S/N
The presence or absence of knocking can be determined by comparing the magnitude of the noise with a predetermined determination level.

By weighting the knock sensor signals as described above, the error in frequency analysis can be reduced and the accuracy of knocking detection can be improved.

Next, a second embodiment will be explained.

When frequency analysis is performed using a microcomputer, it may not be possible to analyze the captured data all at once due to processing speed limitations. In that case, Fig. 5 (a
), there is a method of dividing the analysis interval into a plurality of parts, performing frequency analysis on each analysis interval, and calculating the sum of the spectrum values of each analysis interval. In this case as well, as in the first embodiment, an error in frequency analysis occurs due to the signal levels at both ends of the analysis section not matching, but as a countermeasure, if each analysis section is weighted as shown in the first embodiment,
As shown in FIG. 5(b), the signal level in the peripheral area of each analysis section decreases, which has the disadvantage of deteriorating the detection accuracy of knocking that occurs in a short period of time. This embodiment takes measures against signal reduction when the analysis section is divided. As a measure against signal reduction, the analysis sections are set to overlap each other as shown in FIG. 5(c). Thereby, the decrease in the signal is compensated for when calculating the sum of the frequency analysis results, and it is possible to prevent deterioration of the knocking detection accuracy.

The configuration of the engine control device to which this embodiment is applied and the configuration of the signal acquisition section are the same as in the first embodiment.

FIG. 6 shows the processing procedure of the CPU 20 in this embodiment.

At step 200, the knock sensor signal is taken in as data and stored in the RAM. In step 210, the number m of analysis sections and the number L of analysis points are set. Here m
is set to sufficiently include the period during which knocking occurs throughout the analysis interval.

[0024] Below, steps 220 to 240
The steps up to this point are executed repeatedly for the number of analysis intervals.

In step 220, the starting point P of the analysis interval is
Set k (k is the number of the analysis section). Here, Pk is
Adjacent analysis sections are set to overlap, and the degree of overlap is approximately 1/2 to 1/4 of the analysis section. In step 230, the data corresponding to the analysis interval is weighted in the same manner as in the first embodiment, and the weighted data is stored. In step 240, frequency analysis is performed on the weighted data to obtain a spectrum.

After the frequency analysis has been completed for the m analysis sections, in step 250, the sum of the spectra of each analysis section is determined for each frequency. In step 260, the total value of the spectrum is S, the average value is N, and S/N is
seek. In step 270, the presence or absence of knocking is determined by comparing the obtained S/N with a predetermined determination level. If knocking is detected here, the ignition timing is retarded in step 280 to avoid future knocking.

[0027] Through the above processing, the accuracy of knocking detection can be ensured by making the analysis intervals overlap in time and compensating for the decrease in signal due to signal weighting.

[0028]

As an effect of claim 1, by weighting the knock sensor signal, it is possible to reduce errors in frequency analysis and improve knocking detection accuracy.

As an effect of claim 2, in a knocking detection device having a plurality of analysis sections, by making the analysis sections overlap, a decrease in signal level due to signal weighting is compensated for, and a decrease in knocking detection accuracy due to signal weighting is avoided. can be prevented.

[Brief explanation of drawings]

FIG. 1 is an explanatory diagram of the principle of the present invention.

FIG. 2 is a system configuration diagram.

FIG. 3 is a configuration diagram of a signal acquisition section.

FIG. 4 is a diagram illustrating the processing procedure of the first embodiment.

FIG. 5 is a diagram showing a method of setting an analysis interval.

FIG. 6 is a diagram illustrating the processing procedure of the second embodiment.

[Explanation of symbols]

1... Air cleaner, 2... Hot wire air flow meter, 5... Throttle valve, 6... Intake manifold, 9... Control unit, 13... Ignition coil, 15... Spark plug, 16... Injector, 17... Water temperature sensor, 18... Knock sensor , 20...CPU, 23...A/D converter, 24...
RAM.

Claims (2)

[Claims] [Claim 1] Engine knocking detection that detects the presence or absence of knocking by capturing the signal output of a vibration detection means for detecting engine vibration as data at every predetermined sampling period, and frequency-analyzing the data to obtain a spectrum. A knocking detection method, characterized in that, in a device, each data of the captured signal output is weighted, and the presence or absence of knocking is determined by performing frequency analysis on the weighted data. 2. The knocking detection device according to claim 1, comprising:
After dividing the captured data into multiple temporally overlapping data sequences and weighting each data sequence, frequency analysis is performed for each data sequence to obtain a spectrum. , a knocking detection method characterized by determining the presence or absence of knocking.