JP2018159332A - Knocking detector for internal combustion engine - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve knocking detection accuracy and thus inhibit a wrong determination during transient operation.SOLUTION: A CPU 9 includes: a frequency analysis section for extracting a plurality of frequency components from vibration detected by a vibration sensor 151 via a filter 39; a frequency switching section for switching a frequency band extracted in accordance with an operating state; a background level estimation section for acquiring a background level estimation value of each of the frequency components on the basis of the operating state; a smoothing section for calculating a background level BGLi by adding a background estimation value of each of the frequency components to each of the frequency components obtained by smoothing each of the frequency components obtained by subtracting the background level estimation value from each of the frequency components; a correction section for changing smoothing processing in the smoothing section when the frequency switching section switches the frequency band; a knocking index calculation section for calculating a knocking index on the basis of each of the frequency components and the background level of each of the frequency components; and a determination section for determining knocking by comparing the knocking index with a determination threshold value.SELECTED DRAWING: Figure 10

Description

  The present invention relates to a knocking detection device for an internal combustion engine that determines the occurrence of knocking from vibrations generated by abnormal combustion of the internal combustion engine.

  Knocking is a phenomenon in which the gas in the combustion chamber vibrates due to self-ignition of the unburned gas in the end portion of the combustion chamber of the internal combustion engine, and the vibration is transmitted to the engine body. For example, knocking causes a loss of energy generated by the engine (power reduction), an impact on each part of the engine, and a reduction in fuel consumption. Therefore, it is desirable to avoid knocking as much as possible. Therefore, it is essential to accurately detect the occurrence of knocking.

  Conventionally, as described in Patent Document 1, only the fundamental vibration frequency and harmonic components are separated from the output signal of the vibration sensor by using a bandpass filter, and depending on whether or not the output is greater than the background level. The occurrence of knocking was detected. Note that the background level is a weighted average of past values of the vibration sensor signal itself.

  The knocking detection device of Patent Document 1 detects the occurrence of knocking based only on the fundamental vibration frequency and the harmonic component. Therefore, if the background level becomes too high during high-speed rotation of the engine, accurate detection of knocking cannot be detected, or if the engine specifications change, the knocking resonance frequency also changes, so that accurate detection of knocking is detected. I could not.

  Therefore, Patent Document 2 discloses a knocking detection device including a vibration sensor that detects vibration generated in an engine, and a plurality of filters that are connected to the vibration sensor and have filter characteristics in different frequency bands in the knocking frequency band. It is disclosed.

  The resonance frequency of knocking generated in the engine varies depending on the operating state of the engine. For example, the resonance frequency of knocking varies depending on whether the engine speed is high or low. Therefore, by providing a filter circuit having a plurality of filters having filter characteristics in different frequency bands with respect to the resonance frequency of knocking, and selecting the output from the filter circuit according to the operating state of the engine, reliability can be improved. High knocking detection can be performed.

  In recent years, it is also known that a filter circuit is realized by a digital filter by software installed in a microcomputer, and further, a method of configuring a digital filter as a function of the microcomputer is also known.

  Patent Document 3 describes a method for solving a response delay caused by a delay filter used for detecting a background level. The knocking detection apparatus of the same document detects the operating state of the engine based on a change in the engine speed, and corrects the threshold for determining knocking based on the operating state.

  Further, conventionally, when the frequency band to be filtered is switched with respect to the knocking characteristic frequency which varies depending on the operating state of the engine, it is necessary to repeat the smoothing process until the background level suitable for the frequency band after the switching is reached. there were.

  In order to solve this problem, Patent Document 4 describes a method capable of calculating a background level suitable for a frequency band after switching in a short time. In this document, when the frequency band for filtering the knocking characteristic frequency is switched, the background level before switching is learned, and the value learned at the timing when the filter setting before switching is restored is reflected as the background level. ing.

JP 58-45520 A JP-A-56-000637 JP-A 63-295864 JP 2003-035194 A

  For example, attempts have been made to increase the compression ratio of the engine in response to demands such as improvement in fuel consumption and exhaust purification performance. On the other hand, when the compression ratio is increased, knocking is likely to occur, and it is expected that the detection accuracy of knocking will be further improved.

  As described in Patent Document 2, in the method of detecting knocking by extracting a plurality of resonance frequency components, when the detection accuracy of knocking is further increased, it is necessary to solve the following problems. That is, it is necessary to accurately calculate the average value of the input to the vibration sensor when knocking has not occurred. For example, when calculating an average value using a delay filter, it is necessary to apply the delay filter more strongly. On the other hand, the delay filter has a characteristic that the response is delayed when the operating state of the engine changes. When this response delay is large, an error in ignition timing due to erroneous detection of knocking occurs when the engine load fluctuates, and in particular, power performance, exhaust purification performance, fuel consumption performance, etc. may be adversely affected.

  As described in Patent Document 3, when the detection accuracy is improved by comparing the average value of a plurality of resonance frequencies and the corrected threshold value, the output of the vibration sensor that increases as the engine speed increases is also increased. It will be detected with high accuracy. Therefore, the risk of erroneous detection similar to the above may increase due to a sudden change in the throttle opening.

  As described in Patent Document 4, when learning the background level before and after switching the frequency band to be filtered with respect to the knocking characteristic frequency that changes according to the change in the engine speed, the engine load state and water temperature Dependent knocking characteristic frequency changes are not considered. Therefore, if the frequency band to be filtered is switched depending on the engine load state and the water temperature in addition to the engine rotation fluctuation, the background level to be stored becomes enormous, and an erroneous background level may be stored. Therefore, in the apparatus of Patent Document 4, there is a possibility of erroneous detection similar to Patent Documents 2 and 3.

  The present invention has been made in view of the above problems, and an object of the present invention is to provide a knocking detection device for an internal combustion engine that can improve detection accuracy of knocking during transient operation and suppress erroneous determination. Is to provide.

  An internal combustion engine knocking detection apparatus according to the present invention includes a frequency analysis unit that extracts a plurality of frequency components from a vibration detected by a vibration sensor of the internal combustion engine through a filter, and a frequency analysis unit according to an operating state of the internal combustion engine. A frequency switching unit that switches a frequency band extracted from the background, a background level estimation unit that obtains a background level estimation value of each frequency component of the vibration of the internal combustion engine based on an operating state of the internal combustion engine, and the frequency component Subtracting the background level estimation value of the frequency component, smoothing each frequency component after the subtraction, and adding the background level estimation value of the frequency component to the smoothed frequency component A smoothing unit for calculating a background level of each frequency component of the vibration of the internal combustion engine, and the frequency A correction unit that changes a smoothing process in the smoothing unit when the switching unit switches the frequency band, knocking of the internal combustion engine based on each frequency component in which the frequency band is switched, and the background level A knocking index calculation unit that calculates a determination index; and a determination unit that determines whether knocking has occurred in the internal combustion engine by comparing the knocking index with a determination threshold.

  According to the present invention, the followability when calculating the background level during transient operation is enhanced, and the smoothing process can be strongly applied. As a result, erroneous determination of knocking during transient operation can be suppressed, and the knocking detection accuracy can be improved regardless of the operating state of the engine.

The analysis result of the frequency component at the time of knocking non-occurrence | production which concerns on Example 1. FIG. The analysis result of the frequency component at the time of the knocking generation which concerns on Example 1. FIG. Explanatory drawing of the general principle which determines generation | occurrence | production of knocking which concerns on Example 1. FIG. FIG. 6 shows changes over time in throttle opening, engine speed, frequency background level, knock index, and knock determination signal during transient operation according to the first embodiment. FIG. 6 shows changes in the background level over time before and after switching the frequency band of the filter according to the first embodiment. 1 is a system configuration diagram of an engine ignition device according to Embodiment 1. FIG. FIG. 2 is a block diagram of a configuration of a control unit according to the first embodiment. The flow explaining the procedure which determines the presence or absence of the knocking which concerns on Example 1. FIG. The calculation chart explaining the procedure which determines the presence or absence of knocking of a prior art. 3 is a calculation chart for explaining a procedure for determining the presence or absence of knocking according to the first embodiment. The map data which described the estimated value of the background level which concerns on Example 1. FIG. The figure of the engine speed and load of the map data concerning Example 1. FIG. 5 is a flow of correction of smoothing processing after switching the frequency band of the filter according to the first embodiment. The comparison figure of the time change of the background level before and behind frequency band change of the filter concerning Example 1. FIG. The flow which demonstrates the procedure in which CPU20 which concerns on Example 1 calculates ignition timing. 10 is a flow of correction of smoothing processing after switching the frequency band of the filter according to the second embodiment.

  Hereinafter, some embodiments will be described in detail with reference to the drawings.

  First, the general principle of knock detection will be described.

<General principles of knock detection>
Engine vibration includes many vibration components generated by, for example, piston friction, crankshaft rotation, valve operation, and the like. Moreover, these vibration components change depending on the operating state of the engine. When knocking occurs in the engine, vibration specific to knocking occurs. The presence or absence of knocking is determined by separating the vibration unique to knocking from the vibration of the entire engine detected by the vibration sensor.

  FIG. 1 is a diagram illustrating the analysis result of the frequency component of the output from the vibration sensor when knocking does not occur according to the present embodiment.

  Here, the analysis results in the range up to about 20 KHz are illustrated. However, the target of the present embodiment is not limited to this. The same applies to the following drawings.

  FIG. 2 is a diagram showing analysis results f and g for each engine speed of the frequency component of the output from the vibration sensor when knocking occurs.

  As can be seen from a comparison between FIG. 1 and FIG. 2, when knocking occurs, each resonance frequency component is larger than when knocking does not occur. As shown in FIG. 2, for example, knocking occurs near the frequencies f10, f01, and f11.

  Further, as can be seen by comparing the analysis result f and the analysis result g in FIG. 2, the characteristic frequency of knocking changes according to the difference in the engine operating state (here, the engine speed). For example, knocking occurs near f10, f01, and f11 in the analysis result f, and occurs near g10, g01, g30, and g11 in the analysis result g.

  Here, by switching the setting of the frequency of the filter capable of detecting the characteristic frequency of knocking according to the operating state of the engine, the knocking determination with high reliability becomes possible. The filter may be a high-pass filter, a low-pass filter, or a band-pass filter.

  FIG. 3 is a diagram for explaining a general principle for determining whether knocking has occurred using the knocking determination index I. In FIG.

  In the following description of the principle, for the sake of convenience, the resonance frequency component ω10P (f10) of the resonance frequency f10 (6.3 KHz) and the resonance frequency component ω01P (f01) of the resonance frequency f01 (18.0 KHz) shown in FIG. An example in which the knocking determination index I is calculated will be described. However, the present invention is not limited to this, and the presence or absence of knocking may be determined using any two or more resonance frequency components.

  The vibration sensor combines and detects vibration caused by the occurrence of knocking and background vibration (vibration generated by factors other than knocking). Therefore, the knocking determination index I is an index Ib corresponding to background vibration when knocking has not occurred, and an index Ib corresponding to background vibration and an index Ik corresponding to knocking when knocking has occurred. It is calculated | required by combining.

  The knocking determination index I can be expressed by the following formula 1 using main resonance frequency components. Here, ω is a real value determined by the engine speed. Furthermore, ω may take a binary value of 1 or 0. P is the vibration intensity (power spectrum) of each resonance frequency component.

  I = ω10P (f10) + ω20P (f20) + ω01P (f01) + ω30P (f30) + ω11P (f11) (Equation 1)

  As shown in FIG. 3, the knocking determination index Ib indicated by the resonance frequency component of the background vibration and the index Ik indicated by the resonance frequency component of the vibration due to knocking are different in direction and magnitude. As is apparent from a human auditory test, when there is no knocking, only the engine sound is produced. When there is knocking, for example, the sound is crisp. Therefore, the presence or absence of knocking can be heard according to the timbre.

  When the vibration due to the occurrence of knocking is added to the background vibration, the knocking determination index I exceeds the threshold value I02. Thereby, it can be determined that knocking has occurred. Hereinafter, in this specification, not only the five terms on the right side of Equation 1, but also the indices calculated by combining a plurality of resonance frequency components included in the output from the vibration sensor are all knock determination indices (hereinafter referred to as knocking indexes). Defined as I).

  As described above, since the knock index is calculated in consideration of the specific frequency component due to the occurrence of knocking in addition to the background vibration, it is possible to determine whether knocking has occurred even when the background vibration increases. .

<Basic concept of knock detection in this embodiment>
In the present embodiment, when calculating the knock index I, the background level is calculated by averaging the vibration detected by the vibration sensor for each frequency component, as in the prior art. Even if knocking does not occur, for example, if the vibration intensity suddenly changes due to an increase in engine vibration or combustion vibration, the calculation result of the background level cannot follow the latest state. As explained, knocking may be falsely detected.

  FIG. 4 shows an example of changes over time of the throttle opening TVO, the engine speed Ne, the background level BGLi of each frequency fi, the knock index I, and the knock determination signal (signal indicating the presence or absence of knock) during transient operation. is there.

  In FIG. 4, when the throttle opening TVO is changed from fully closed to fully open, the engine speed Ne actually rises after a delay time has elapsed. The background level (BGLi) is essentially a movement as indicated by a broken line. However, when BGLi is calculated by the averaging process, a delay as shown by the solid line occurs. The knock index I is obtained by calculating the S / N ratio (ratio of each frequency component to BGLi).

  According to the example shown in FIG. 4, the actual knock index I exceeds the theoretical knock index I even though knocking does not occur due to the delay of BGLi, and as a result, the knocking is erroneously determined. You can see that The reason for this will be further described.

  When the background level BGLi increases rapidly, the calculation result of the background level BGLi obtained by averaging is delayed from the actual value. When the ratio of noise to the background level BGLi is used as the knock index I, the calculation delay of the background level BGLi leads to the denominator of the knock index I being smaller than the actual value. Therefore, knock index I is calculated to be larger than the original value. In that case, in the transient operation state, the clearance between the knock index I and the knock determination threshold value I02 becomes small, and there is a possibility that it is determined that knocking is present even though knocking has not occurred. The example shown in FIG. 4 shows such a state.

  FIG. 5 is an example showing a change with time of the background level BGLi before and after switching of the frequency band of the filter.

  According to the example shown in FIG. 5, since the smoothing process is performed when the frequency band of the knocking detection filter is switched according to the operating state of the engine, the background level BGLi of the frequency band before the switching is referred to. Therefore, since it is delayed until the actual background level BGLi corresponding to the frequency band after switching, knocking may be erroneously detected.

  In this embodiment, as illustrated in FIG. 4, the calculation result of the background level BGLi suppresses the erroneous determination of knocking during the transient operation due to the fact that the actual background level BGLi cannot be followed and is delayed. To do. Furthermore, the present embodiment further improves the accuracy of the background level BGLi immediately after switching the frequency band to be filtered according to the difference in the operating state of the engine, as illustrated in FIG. As a result, erroneous detection of knocking during transient operation is prevented. As a result, the knocking detection accuracy is improved regardless of whether it is transient or steady.

<Device configuration>
FIG. 6 is a system configuration diagram of the engine ignition device.

  Air flows in from the inlet of the air cleaner 1, passes through the duct 3, the throttle body 5 having a throttle valve, and the intake pipe 6 and is sucked into the cylinder of the engine 7 as an “internal combustion engine”. The duct 3 is provided with a hot-wire air flow meter 2. The hot-wire air flow meter 2 detects the amount of intake air and outputs a detection signal representing the detection result to a control unit (CPU: Central Control Unit) 9.

  The fuel is injected from a fuel tank (not shown) through the injector 16, mixed with intake air in the intake pipe 6, and supplied into the cylinder of the engine 7. The air-fuel mixture is compressed in the cylinder of the engine 7 and ignited by the spark plug 15. The air-fuel mixture after the explosion is discharged from the exhaust pipe 8. An exhaust sensor 11 is provided in the exhaust pipe 8, and a detection signal representing a detection result by the exhaust sensor 11 is input to the control unit 9.

  The high voltage generated from the ignition coil 13 is distributed to each cylinder by the distributor 14 and supplied to the spark plug 15. The crank angle sensor 12 detects the rotational state of the engine 7. The crank angle sensor 12 outputs a Ref signal indicating an absolute position for each rotation and a POS signal indicating a position moved by a predetermined angle from the absolute position. The Ref signal and the POS signal are input to the control unit 9. A vibration sensor (combustion state sensor) 151 that detects vibration is attached to the engine 7, and a detection signal representing a detection result of the vibration sensor 151 is input to the control unit 9.

  FIG. 7 is a block diagram of the configuration of the control unit 9.

  The control unit 9 calculates a fuel supply amount, ignition timing, and the like based on signals from the sensors, and outputs control signals to the injector 16 and the ignition coil 13. The control unit 9 is roughly divided into a control block 34 and a knocking detection block 35.

  The control block 34 includes a CPU 20, an A / D converter 21, a ROM (Read Only Memory) 22, an input I / O 23, a RAM (Random Access Memory) 24, a DPRAM (Dual Port RAM) 25, and an output. An I / O 26 and a bus 37 are provided.

  The knocking detection block 35 includes a CPU 29, a port 27, a timing circuit 28, an A / D converter 30, a ROM 31, a RAM 32, a clock 33, an operational circuit 38, and a bus 36. Data exchange between the CPU 20 and the CPU 29 is performed through the DPRAM 25, for example.

  The intake air amount Qa detected by the hot-wire air flow meter 2 is converted into a digital value by the A / D converter 21 and taken into the CPU 20. The Ref signal and the POS signal detected by the crank angle sensor 12 are taken into the CPU 20 through the input I / O 23. The CPU 20 calculates in accordance with the program stored in the ROM 22 and outputs the calculation result via the output I / O 26 as a fuel injection time signal Ti for instructing the fuel injection amount and an ignition timing signal θign for instructing the ignition timing. Output to each actuator. The RAM 24 stores necessary data during the arithmetic processing.

  When the operational circuit 38 generates a top dead center (TDC) signal, the timing circuit 28 divides the periodic signal generated by the clock 33 in accordance with the content input to the port 27 by the CPU 20 to thereby obtain a sampling signal. Is generated. When the sampling signal is generated, the A / D converter 30 converts the output signal P of the vibration sensor 151 into a digital value.

  The CPU 29 stores the sampled digital value in the RAM 32 according to the program stored in the ROM 31, and determines whether knocking has occurred or not according to the flowchart described later with reference to FIG. Further, the output signal P converted into the digital value of the vibration sensor 151 is input to the knocking detection filter 39, and the output of the filter 39 is input to the CPU 29. The filter 39 can set an arbitrary frequency band by the CPU 29 and can extract a frequency component of the frequency band set by the CPU 29. The determination result of the presence / absence of knocking is notified to the CPU 20 via the DPRAM 25.

<Knocking determination procedure>
FIG. 8 is a flowchart illustrating a procedure for determining the presence or absence of knocking.

  This flowchart is executed for each explosion cycle of the engine 7 when the CPU 29 interrupts the CPU 29. Hereinafter, each step of FIG. 8 will be described.

(FIG. 8: Step S100)
The CPU 29 determines the operating state of the engine 7 based on the Ref signal and the POS signal detected from the crank angle sensor 12 and sets the frequency band of the filter 39.

(FIG. 8: Steps S101 and S102)
The CPU 29 takes in an A / D conversion value obtained by converting the detection signal P from the vibration sensor 151 by the A / D converter 30 (S101). The CPU 29 analyzes the knocking characteristic frequency of the captured A / D conversion value (S102). This frequency analysis is performed by a technique such as fast Fourier transform or Walsh Fourier transform.

(FIG. 8: Step S103)
The CPU 29 selects a plurality of frequency components analyzed in step S102 including the resonance frequency. For example, eight resonance frequencies are selected. The frequency component selected in this step is determined according to the filter 39 in step 100. However, for example, it may be determined in advance according to the specifications of the engine 7.

(FIG. 8: Step S104)
CPU29 calculates | requires S / N ratio showing vibration intensity for every frequency component selected by step S103. Specifically, the background level (BGL ₁ ,..., BGL _i ) corresponding to the selected frequency component (f ₁ ,..., F _i ) is obtained by smoothing processing, and the S / N for each frequency is obtained. The ratio SL _i = f _i / BGL _i is obtained. In step S103, if you select the eight frequency components _to determine the SL 1 to SL _8.

(FIG. 8: Step S104: Supplement)
The calculation of the background level BGLi in this step does not use the signal component detected by the vibration sensor 151 as it is, but subtracts the estimated value of the background level BGLi in advance, calculates the weighted average, and then calculates the background level. The estimated value of BGLi is added again. The ratio of _{f i} for that result is calculated as the SL _i in this step.

  If the frequency band of the filter 39 is switched in step S100, the weighted average correction is executed in this step. A specific procedure will be described with reference to FIGS. 10, 14, and 15 described later.

(FIG. 8: Step S105)
The CPU 29 extracts m components in the descending order of the S / N ratio obtained in step S104 from the frequency components selected in step S103, and obtains the knock index I by adding them. For example, the S / N ratios of the top five frequency components can be extracted and added.

(FIG. 8: Step S106)
The CPU 29 compares the determination threshold value I02 with the knock index I obtained in step S105 (S106). When knock index I is larger than determination threshold value I02 (S106: YES), the process proceeds to step S107, and otherwise (S106: NO), the process proceeds to step S109.

(FIG. 8: Step S106: Supplement)
The determination threshold I02 used in this step may be determined in advance, or may be calculated based on the operating state such as the rotational speed of the engine 7, for example. As a calculation method, for example, a data map in which a correspondence relationship between the driving state and the determination threshold value I02 is defined in advance may be used.

(FIG. 8: Steps S107 and S108)
The CPU 29 determines that knocking has occurred (S107), and sets “1” to a knock flag indicating the occurrence of knocking (S108). This knock flag is used in a separately started ignition control task.

(FIG. 8: Step S109)
The CPU 29 sets the knock flag to “0”.

  The procedure for determining the presence or absence of knocking in the present embodiment has been described above. Hereinafter, details of the procedure in this embodiment will be described with reference to FIGS. 9, 10, and 12 to 14, comparing the procedure for determining the presence or absence of knocking in the prior art with this embodiment.

  FIG. 9 is a calculation chart for explaining the procedure for determining the presence or absence of knocking in the prior art.

  Each step is executed by the CPU 29. Hereinafter, processing of each calculation block in FIG. 9 will be described.

  The vibration sensor 151 detects the vibration of the engine 7 (S301), and the AD converter 30 converts the detection result into a digital signal (S302). The CPU 29 sets the frequency band detected by the knocking detection filter 39 according to the operating state such as the rotation speed and load of the engine 7 (S501). The CPU 29 applies, for example, a band-pass filter (BPF: Band-pass filter) to the vibration signal to thereby calculate the frequency components used to calculate the knock index I (here, three frequency components are illustrated). Is extracted (S303).

  The CPU 29 calculates a weighted average for each frequency component (S304). The CPU 29 converts the calculated weighted average into an appropriate index (for example, the ratio to the background level BGLi described in step S104) (S305). The CPU 29 obtains the knock index I by adding the indexes of the respective frequency components (S306). The knock determination threshold I02 can be obtained in the same manner as in step S106 (S307). The CPU 29 determines the presence or absence of knocking by comparing the knock index I with the knock determination threshold I02 (S308).

  In each calculation block of FIG. 9, a diagram schematically showing a calculation result at the time of transient operation is also shown. During transient operation, the frequency component increases with time. If the background level BGLi is calculated by the weighted average in S304, the background level BGLi is obtained as a smaller value than the actual value, and as a result, the knock index I becomes large, and there is a possibility that erroneous determination occurs. Further, when knocking or when a small level of noise or small knocking that does not detect knocking occurs, the detected value is reflected in the background level BGLi. As a result, the background level BGLi increases and the knock detection accuracy may decrease.

  FIG. 10 is a calculation chart illustrating the procedure for determining the presence or absence of knocking in the present embodiment.

  Each step is executed by the CPU 29. Hereinafter, the processing of each calculation block in FIG. 10 will be described.

  Steps S301 to S303 are the same as in the prior art. These correspond to S101 to S103 in FIG.

  In step S401, the CPU 29 estimates the operating state of the engine 7 based on, for example, the rotational speed and load of the engine 7. The ROM 31 stores in advance map data describing the correspondence between the operating state of the engine 7 and the standard background level BGLi in the normal operating state. The CPU 29 calculates an estimated value of the background level BGLi in the current driving state by referring to the map data based on the estimated driving state.

  FIG. 11 and FIG. 12 show an embodiment of step S401.

  FIG. 11 is an image diagram of map data describing an estimated value of the background level BGLi.

  The map data describes the correspondence relationship between the rotational speed (or rate of change) of the engine 7, the load (or rate of change) of the engine 7, and the background level BGLi (or a correction value for the specified level). In the example shown in FIG. 11, the background level BGLi tends to increase on the high rotation side and the high load side.

  FIG. 12 is a view of FIG. 11 viewed from the axis of the rotational speed of the engine 7 and the load of the engine 7.

  Since the characteristic frequency of knocking changes according to the operating state of the engine 7, background level estimation values in different frequency bands are combined into one according to the operating state in advance. By making one map in advance (three frequency components are illustrated here), it is possible to reduce the capacity of the ROM 31 as compared with the case of having a background level estimation value for each characteristic frequency of knocking. . The calculated background level estimated value is linked with the filter 39 in step S303, and is subtracted between the same frequency components in step S402 described later.

  As other methods for estimating the background level BGLi, the CPU 29 has a throttle sensor signal (a signal indicating the throttle opening), an intake air amount signal (a signal indicating the intake air amount for the engine 7), a fuel injection pulse signal (a fuel injection signal). A sensor signal indicating a load state such as an instructed pulse signal) and an intake pipe pressure signal (a signal indicating the pressure inside the intake pipe 6) may be acquired, and the operation state may be estimated based on these signals. The correspondence between these signals and the driving state can be described by, for example, the same map data as in FIGS. 11 and 12, or the driving state can be calculated by other appropriate methods. Furthermore, the map data of the rotation speed and load of the engine 7 illustrated in FIG. 11 may be combined. This step corresponds to the processing described in “S104: Supplement”.

  In step S402, the CPU 29 subtracts the estimated value of the background level BGLi of the corresponding frequency component from each frequency component extracted in S303. Thereby, since the fluctuation | variation of the background level BGLi by the change of a rotation speed or load is not input with respect to a subsequent weighted average, the tracking delay accompanying a weighted average process can be suppressed. However, the estimated value of the background level BGLi subtracted in this step is added back for each frequency component in step S403 after the weighted average. This step corresponds to the processing described in “Step S104: Supplement”.

  In step S304, after the estimated value of the background level BGLi is previously subtracted from each frequency component in S402, a weighted average of each frequency component is calculated again. In step S403, the CPU 29 adds the estimated value of the background level BGLi subtracted in step S402 to each frequency component with respect to the result of the weighted average of each frequency component. As a result, the signal level is returned to the level before the subtraction.

  Furthermore, the characteristic frequency of knocking changes during transient operation. When the frequency component of knocking is weighted and the background level BGLi is calculated, the previous value is referred to even after the frequency band is switched. There is a possibility that erroneous determination will occur. Therefore, when the frequency band detected by the knocking detection filter 39 is switched in S501, the weighted average value in S304 is corrected. The weighted average correction process will be described with reference to FIG.

  FIG. 13 is a diagram illustrating a correction flow of the smoothing process after the frequency band of the filter 39 is switched.

(FIG. 13: Step S600)
The CPU 29 calculates the weighted average calculation BGLi (current value) by subtracting the background level estimated value (basic BGL) in S401 from the signal component detected by the vibration sensor 151 in advance.

(FIG. 13: Step S601)
The CPU 29 determines whether the frequency band to be extracted by the band pass filter has been switched in S501 according to the operating state of the engine 7.

(FIG. 13: Step S602)
If it is determined that the frequency band has been switched in S601 (S601: YES), the CPU 29 sets the background level BGLi (previous value) corresponding to the frequency component selected in S601 to the BGLi (current value) calculated in S600. Update. Furthermore, the background level BGLi for each filter frequency band corresponding to the operating state may be stored, and the weighted average value may be updated with the stored value.

(FIG. 13: Step S603)
The CPU 29 performs a weighted average process represented by the following equation.
BGLi = (BGLi (previous value) × Wp + BGLi (current value) × Wc)
At this time, Wp + Wc = 1, Wp> = 0, and Wc> = 0.

  Hereinafter, the correction effect of the weighted average process described in FIG. 13 will be described with reference to FIG.

  FIG. 14 is a timing chart showing correction of the weighted average process.

  The upper diagram of FIG. 14 shows a case where correction is not performed, and a tracking delay occurs until the actual background level BGLi is reached after the frequency is switched.

  The lower diagram of FIG. 14 shows a case where the weighted average is corrected. As a result of performing the weighted average correction process shown in FIG. 13, a value suitable for the background level BGLi after switching the frequency band is obtained, so that knocking detection error after switching of the characteristic frequency band of knocking during transition is performed. Can be prevented. These steps correspond to the processing described in “Step S104: Supplement”.

  Returning to FIG. Step S305 is similar to the prior art, and corresponds to step S104. However, the denominator in calculating SLi, which is the S / N ratio, is the result of step S403, and the numerator is the frequency component extracted in step S303. Prior to reaching this step, the transient fluctuation of the background level BGLi is removed through steps S304 and S403, so that erroneous determinations associated with the transient fluctuation can be suppressed unlike the conventional technique. . Step S306 and subsequent steps are the same as in the prior art, and correspond to step S105 and subsequent steps.

  FIG. 15 is a flowchart for explaining the procedure by which the CPU 20 calculates the ignition timing.

  The CPU 20 is activated periodically (for example, every 10 msec). Hereinafter, each step of FIG. 15 will be described.

(FIG. 15: Step S201)
The CPU 20 reads the engine speed N and the intake air amount Q from predetermined registers set in the RAM 24.

(FIG. 15: Step S202)
The CPU 20 calculates an intake air amount Q / N (basic fuel injection amount) per unit rotational speed, and obtains a fuel injection time width Ti from the intake air amount Q / N. The CPU 20 obtains the basic ignition timing θbase based on the basic ignition timing map stored in the ROM 22. The basic ignition timing map is a data map that describes the correspondence relationship between the intake air amount Q / N, the rotational speed N, and the basic ignition timing θbase.

(FIG. 15: Step S203)
The CPU 20 determines whether or not knocking has occurred according to the contents of the knock flag (which the CPU 29 has obtained in accordance with FIG. 8). If knocking has occurred (S203: YES), the CPU 20 proceeds to step S213, and if not (S203: NO), the CPU 20 proceeds to step S204.

(FIG. 15: Step S213)
The CPU 20 subtracts a predetermined retardation amount Δθret from the ignition timing θadv. By this subtraction, the ignition timing is retarded.

(FIG. 15: Step S214)
The CPU 20 initializes the count value A and proceeds to step S208. The count value A is a variable for counting the number of occurrences of knocking. The count value A is used to determine whether it is time to recover the ignition timing θadv retarded by the occurrence of knocking by the advance amount Δθadv. How to use the count value A will be described in later steps.

(FIG. 15: Steps S204 and S205)
The CPU 20 increments the count value A by one (S204). The CPU 20 determines whether or not the count value A has reached a predetermined value (for example, 50) (S205). CPU20 progresses to step S206, when it has reached | attained (S205: YES), and when not having reached (S205: NO), it skips to step S208.

(FIG. 15: Steps S206 and S207)
The CPU 20 adds a predetermined advance amount Δθadv to the advance value θadv (S206). By this addition, the ignition timing retarded in step S213 is recovered. When this flowchart is activated every 10 msec, 0.5 seconds have elapsed since the count value A was initialized when the count value A reached 50. That is, this step recovers the ignition timing every time 0.5 seconds elapse after the ignition timing is retarded due to the occurrence of knocking. The CPU 20 initializes the count value A (S207).

(FIG. 15: Step S208)
The CPU 20 calculates the ignition timing θign by adding the advance value θadv to the basic ignition timing θbase.

(FIG. 15: Step S209)
The CPU 20 reads the maximum advance value θres from the maximum advance value map stored in the ROM 22. The maximum advance value map is a data map describing a correspondence relationship between the intake air amount Q / N, the rotational speed N, and the basic ignition timing θres.

(FIG. 15: Steps S210 and S211)
The CPU 20 determines whether the ignition timing θign exceeds the maximum advance value θres (S210). If not exceeded (S210: NO), the CPU 20 skips to step S212. If it has exceeded (S210: YES), θign has advanced too much, so the CPU 20 sets the ignition timing θign to the maximum advance value θres in step S211 (S211).

(FIG. 15: Step S212)
The CPU 20 outputs the delay time td, the frequency division ratio ts, and the sampling point number ns to the port 27 according to the state of the engine 7. The frequency division ratio ts determines the sampling period of the digital value of the output from the vibration sensor 151, and the sampling point ns determines the sampling point.

<Summary>
A CPU (knocking detection device) 9 according to the present embodiment includes a frequency analyzer, a frequency switching unit, a background level estimation unit, a smoothing unit, a correction unit, a knocking index calculation unit, and a determination unit. Prepare. The frequency analyzer extracts a plurality of frequency components from the vibration detected by the vibration sensor 151 of the engine 7 through the filter 39. The frequency switching unit switches the frequency band extracted by the frequency analyzer according to the operating state of the engine 7. The background level estimation unit acquires a background level estimation value of each frequency component of vibration of the engine 7 based on the operating state of the engine 7. The smoothing unit subtracts the background level estimation value of each frequency component from each frequency component and then smoothes each frequency component after the subtraction, and the background estimation of each frequency component to each smoothed frequency component The background level of the vibration of the engine 7 is calculated by adding the values. The correction unit changes the smoothing process in the smoothing unit when the frequency switching unit switches the frequency band. The knocking index calculation unit calculates a knocking determination index of the engine 7 based on each frequency component and the background level of each frequency component. The determination unit determines whether knocking has occurred in the engine 7 by comparing the knocking index with a determination threshold value.

  Thereby, the followability at the time of calculating a background level at the time of transient operation increases, and a smoothing process can be applied strongly. As a result, erroneous determination of knocking during transient operation can be suppressed, and an accurate background level can be calculated regardless of whether the engine is in steady operation or transient operation. Can be suppressed.

  In particular, the background level estimation value is subtracted from the frequency component of the vibration sensor 151 in advance and then subjected to a weighted average, and then the background level estimation value is added back. Thereby, in the weighted average process, the transient fluctuation of the background level BGLi is excluded, so that it is possible to suppress the calculation delay associated with the transient fluctuation and the knocking erroneous detection caused by this. By updating the weighted average value after switching the frequency band and calculating the background level BGLi, the accuracy of the background level BGLi is improved and the accuracy of knock detection can be increased. Therefore, it is possible to optimally control the ignition timing in all operating states.

  Furthermore, since the correction unit changes the smoothing process in the smoothing unit every time the frequency switching unit switches the frequency band, the followability of the background level BGLi with respect to the change in frequency is increased, and the smoothing process is strongly applied. Thus, erroneous determination of knocking during transient operation can be suppressed.

  Furthermore, since the correction unit replaces each frequency component that is changed before the frequency band is switched with each frequency component that is smoothed after the frequency band is switched, the background level BGLi can be updated even during transient operation. The background level BGLi can follow the change in frequency.

  Furthermore, since the background level estimated value is calculated based on a map made up of a plurality of frequency components, the estimated value of the background level BGLi can be easily calculated simply by referring to the map.

  A CPU (knocking detection device) 9 according to the second embodiment will be described focusing on differences from the first embodiment.

  FIG. 16 is a diagram illustrating a flow of correction of the smoothing process after switching the frequency band of the filter 39 according to the second embodiment.

(FIG. 16: Step S700)
The CPU 29 calculates BGLi (current value) for weighted average calculation by subtracting the background level estimated value (basic BGL) in S401 from the signal component detected by the vibration sensor 151 in advance.

(FIG. 16: Step S701)
The CPU 29 determines whether the frequency band to be extracted by the band pass filter has been switched in S501 according to the operating state of the engine 7.

(FIG. 16: Step S702)
If it is determined in S701 that the frequency band has been switched (S701: YES), the CPU 29 updates the background level BGLi (previous value) corresponding to the frequency component selected in S701 to 0. Further, the weighted average value to be updated is set to a value shorter than the value before the convergence time at which the background level BGLi converges after frequency switching is updated.

(FIG. 16: Step S703)
The CPU 29 determines whether a predetermined time has elapsed after the frequency is switched. The predetermined time may be determined in advance according to the operating state of the engine 7, or may be a value corresponding to a weighted average weight.

(FIG. 16: Step S704)
When it is determined in S703 that the predetermined time has not elapsed (S703: NO), the CPU 29 substitutes the weight Wp of BGLi (previous value) into Wheavy and BGLi (current value) Wc into Wlight. At this time, Wheavy> Wlight, Wp + Wc = 1, Wp> = 0, and Wc> = 0.

(FIG. 16: Step S705)
When it is determined in S703 that the predetermined time has elapsed (S703: YES), the CPU 29 updates the weight Wp of BGLi (previous value) to Wlight and the BGLi (current value) Wc to Wheavy.
When setting the weighted average weight in S704 and S705, the weight may be changed over time.

(FIG. 16: Step S706)
The CPU 29 performs a weighted average process represented by the following equation.
BGLi = (BGLi (previous value) × Wp + BGLi (current value) × Wc)

  According to the present embodiment, the correction unit reduces the weight of each frequency component smoothed after the frequency band is switched from the frequency component smoothed before the frequency band is switched until a predetermined time elapses. Since the smoothing process in the smoothing unit is executed, noise can be suppressed and the knocking detection accuracy can be improved.

<Modification>
In addition, this invention is not limited to an above-described Example, Various modifications are included. For example, the above-described embodiments have been described in detail for easy understanding of the present invention, and are not necessarily limited to those having all the configurations described. Furthermore, a part of the configuration of a certain embodiment can be replaced with the configuration of another embodiment, and the configuration of another embodiment can be added to the configuration of a certain embodiment. Furthermore, it is possible to add, delete, and replace other configurations for a part of the configuration of each embodiment.

  In the above embodiment, each step for detecting knocking by the CPU 29 (corresponding to a frequency analysis unit, a frequency switching unit, a background level estimation unit, a smoothing unit, a knocking index calculation unit, and a determination unit) is executed. Explained the implementation. However, an equivalent function may be implemented in hardware such as a circuit device. For example, one or more of frequency analysis, smoothing, processing for obtaining an estimated value from map data, calculation of a knock index I, and determination of a threshold value I02 may be implemented in a circuit device. .

  The output from the vibration sensor 151 attached to the cylinder block has a characteristic that increases with an increase in the rotational speed or load of the engine 7, for example. This is due to mechanical noise such as piston sludge in the engine 7 and combustion mode change. Therefore, in step S103, a different frequency component can be selected for each cylinder.

  In order to acquire injector noise resulting from the operation of the injector 16, the vibration sensor 151 sets a period for operating the injector 16 as a sampling window, and acquires the injector noise within the window. Since this injector noise is unnecessary when calculating the knock index, it is desirable to remove it. Therefore, data describing the assumed injector noise may be stored in advance in the ROM 22 or the like, and the CPU 29 may subtract the injector noise from the detection result of the vibration sensor 151 in the sampling window. As a result, the CPU 29 can accurately calculate the knock index without being affected by the injector noise.

  Since the standard background level is described in the map data in which the estimated value of the background level is described, it is desirable not to describe the transient fluctuation of the background level in the map data. For example, when the rotation speed of the engine 7 or the engine load increases rapidly, it is assumed that the background level also increases rapidly. However, such a rapidly changing background level BGLi is not described in the map data. Specifically, it is desirable to describe in the map data only those with a fluctuation rate of the background level BGLi within a certain range.

  It is not always necessary to extract all the frequency components extracted from the detection result of the vibration sensor 151. For example, only representative frequency components can be extracted, and other frequency components can be supplemented by interpolation processing. An arithmetic expression for the interpolation processing can be determined in advance based on, for example, a test result. Thereby, the calculation load for extracting a frequency component can be suppressed.

  In the above embodiment, the ratio (S / N ratio) of the frequency component to the background level BGLi is calculated as a knock index for the frequency component. However, the difference with respect to the background level may be used as a knock index instead of the ratio.

  The CPU 9 does not need to include a plurality of CPUs 20 and 29, and may be a single CPU in which one CPU also serves as the other CPU.

7: Engine, 9: Control unit, 39: Filter, 151: Vibration sensor

Claims (5)

A frequency analysis unit that extracts a plurality of frequency components from a vibration detected by a vibration sensor of the internal combustion engine through a filter;
A frequency switching unit that switches a frequency band extracted by the frequency analysis unit according to an operating state of the internal combustion engine;
A background level estimation unit that obtains a background level estimation value of each frequency component of vibration of the internal combustion engine based on the operating state of the internal combustion engine;
After subtracting the background level estimation value of each frequency component from each frequency component, each frequency component after the subtraction is smoothed, and the background estimation value of each frequency component is added to each smoothed frequency component. A smoothing unit that calculates a background level of each frequency component of the vibration of the internal combustion engine by adding together,
A correction unit that changes a smoothing process in the smoothing unit when the frequency switching unit switches the frequency band; and
A knocking index calculating unit that calculates a knocking index of the internal combustion engine based on each frequency component and a background level of each frequency component;
A knocking detection apparatus for an internal combustion engine, comprising: a determination unit that determines whether or not knocking has occurred in the internal combustion engine by comparing the knocking index with a determination threshold value. The correction unit changes the smoothing process in the smoothing unit each time the frequency switching unit switches the frequency band.
The knock detection device according to claim 1. The correction unit replaces each frequency component smoothed before switching the frequency band with each frequency component smoothed after switching the frequency band,
The knock detection device according to claim 1. The correction unit reduces the weight of each frequency component that is smoothed after switching of the frequency band to be lower than the frequency component that is smoothed before switching of the frequency band until a predetermined time elapses. Performing the smoothing process in the unit,
The knock detection device according to claim 1. The background level estimate is calculated based on a map composed of the plurality of frequency components.
The knock detection device according to claim 1.

Images