JP7726723B2 - Method and device for detecting abnormal engine noise - Google Patents

Description

This invention relates to abnormal noise detection technology for detecting phenomena such as misfires and abnormal combustion in engines, such as vehicle engines, based on engine sounds.

For example, attempts have been made in the past to detect engine misfires from changes in sound or vibration. For example, Patent Document 1 discloses technology in which engine vibration sounds are collected with a microphone, and sounds within a certain time range after a predetermined time has elapsed since the generation of an ignition pulse are sampled and analyzed for frequency, and the detected vibration waveform is compared with a normal vibration waveform for a specific frequency band to determine whether or not there is an abnormality.

Japanese Patent Application Publication No. 1-217218

With the above technology, it is necessary to obtain pulse signals indicating the ignition timing of each cylinder from the vehicle's engine controller in order to determine the sampling time frame. Therefore, it is not possible to easily check for misfires and other abnormalities in a short amount of time in places such as automobile repair shops or the finished vehicle inspection process on automobile production lines.

The engine abnormal noise detection method according to the present invention comprises:
Acquire the sound produced by the engine to generate sound data,
This sound data is converted according to frequency to generate frequency data,
determining a fundamental frequency associated with the rotational speed of the engine based on the frequency data;
Using this fundamental frequency, data on harmonics of the fundamental frequency is extracted from the frequency data.
Based on the harmonic data, it is determined whether or not there is an abnormal sound contained in the sound data.
The results of this determination are presented as information.
In one embodiment, the presence or absence of abnormal noise is determined from the change over time in the frequency of at least one harmonic.
In another aspect, the presence or absence of abnormal noise is determined from the time variation of the frequency and sound pressure of at least one harmonic.
In still another aspect, when the sound pressure of at least one harmonic tone decreases and returns to normal within a predetermined time range corresponding to a misfire, it is determined that the abnormal noise is due to a misfire.

According to this invention, by focusing on harmonics of the fundamental frequency, changes in sound caused by engine misfires, abnormal combustion, etc. become more noticeable and are easier to detect as abnormal noise. Therefore, it is possible to detect abnormal noises such as misfires without obtaining signals such as ignition timing from the vehicle.

1 is a functional block diagram of a first embodiment in which the present invention is applied to engine misfire detection; 4 is a flowchart showing the flow of processing in the first embodiment. FIG. 10 is an explanatory diagram of an example of a spectrogram displayed as frequency data. 10 is a flowchart showing the flow of a process for identifying a fundamental frequency. FIG. 10 is an explanatory diagram of a process for identifying a fundamental frequency. FIG. 10 is a functional block diagram of a second embodiment. 10 is a flowchart showing the flow of processing according to a second embodiment. FIG. 10 is a functional block diagram of a third embodiment. 10 is a flowchart showing the flow of processing according to a third embodiment. FIG. 10 is a functional block diagram of a fourth embodiment. 10 is a flowchart showing the flow of processing according to a fourth embodiment. FIG. 13 is a functional block diagram of a fifth embodiment. 13 is a flowchart showing the flow of processing according to the fifth embodiment. FIG. 4 is an explanatory diagram showing a display example on a display unit.

An embodiment of this invention applied to misfire detection in a vehicle engine is described below. This misfire detection embodiment is implemented, for example, as a misfire inspection to ensure that engine misfires do not occur during the finished vehicle inspection process, which is the final stage of an automobile production line. Typically, during the finished vehicle inspection process, an inspector test-drives the finished vehicle on a free roller, inspecting numerous items including the engine, meters, and brakes. During this test run on the free roller, the engine is accelerated so that the engine speed gradually changes from a low speed (e.g., 600 rpm) to a high speed (e.g., 6000 rpm) over an appropriate inspection time (e.g., approximately 10 to 20 seconds). A misfire inspection is performed based on the sound produced by the engine during this period. Hereinafter, the device of this embodiment will be referred to as the "misfire inspection device," and this misfire inspection device is configured as part of the inspection equipment used in the finished vehicle inspection process.

The engines to be inspected may be either gasoline engines (i.e., spark-ignition internal combustion engines) or diesel engines (i.e., compression-ignition internal combustion engines), and may be either four-stroke or two-stroke. In addition to in-line multi-cylinder engines, they may also be V-type multi-cylinder engines, etc.

Figure 1 shows a functional block diagram of the misfire detection device of the first embodiment. The misfire detection device of the first embodiment is composed of a sound measurement/acquisition unit 10, a frequency conversion unit 20, a fundamental frequency calculation unit 30, a harmonic identification unit 40, an abnormal sound detection unit 50, and a display unit 60.

The sound measurement and acquisition unit 10 includes a microphone that acquires engine sounds and converts them into electrical signals, i.e., sound data, and a recording unit that temporarily stores this sound data. The microphone is placed outside the vehicle so that it can collect engine sounds as the vehicle runs on free rollers. The microphone's directivity and frequency characteristics are selected to suit its position relative to the vehicle being measured and the frequency band of the engine sound. Typically, a microphone with directivity toward the vehicle is used. A microphone array or similar may be used to localize the sound source and obtain sound data from which noise has been removed. This microphone can also be an existing microphone installed to collect horn sounds during the finished vehicle inspection process, for example.

As mentioned above, the misfire test is performed while increasing the engine speed for, for example, 10 to 20 seconds, so sound data with a duration of, for example, 10 to 20 seconds is acquired.

The frequency conversion unit 20 generates frequency data by converting the acquired sound data according to frequency using frequency analysis techniques such as FFT (Fast Fourier Transform) or wavelet analysis. In particular, since it is necessary to ultimately determine the time change in frequency or sound pressure of a specific sound, frequency analysis is performed for each relatively short time domain to generate frequency data for each individual time.

When frequency data is expressed in two dimensions, for example, it is expressed as a frequency spectrum with frequency on the horizontal axis and power (or sound pressure) on the vertical axis. Furthermore, by overlaying the conversion results for each time in a time series, frequency data can be handled as a so-called spectrogram, which is three-dimensional data that includes time.

Figure 3 shows an example of a spectrogram display created by processing sound data obtained from a misfire inspection using FFT. In accordance with the general spectrogram format, the horizontal axis represents time and the vertical axis represents frequency. The total time length is approximately 15 seconds. Frequencies up to approximately 1000 Hz are included. The brightness and color of each point represent the power (dBA) (or sound pressure or amplitude) of a given frequency at a given time. While the attached image is a black-and-white image, an actual spectrogram display uses different colors, such as red for high power, yellow-green for medium power, and blue for low power, and the power is indicated by varying brightness within each color. The multiple lines in the spectrogram in Figure 3 represent frequencies that represent sound pressure peaks of several orders, including the fundamental frequency. In the illustrated example, the line in the lowest frequency band (designated F1) is the characteristic of the fundamental frequency, and several clearly visible lines (designated F2, F3, etc.) are the characteristic of frequencies corresponding to harmonics of the fundamental frequency. As mentioned above, the misfire test in one embodiment is performed while the engine rotation speed is increasing, so the fundamental frequency and the harmonics tend to increase over time (i.e., as the engine rotation speed increases).

Note that the graph shown below the spectrogram display in Figure 3 represents the original sound data, with amplitude on the vertical axis.

The fundamental frequency calculation unit 30 determines the fundamental frequency associated with the engine rotation speed based on the frequency data obtained by the frequency conversion unit 20. Here, the fundamental frequency is determined from the engine sound, i.e., frequency data, without obtaining a signal indicating the engine rotation speed from the vehicle being inspected. In an engine with multiple cylinders, sound is generated by intermittent explosions in each cylinder, and this is the fundamental frequency. Here, the fundamental frequency is estimated from the sound pressure peaks contained in the frequency data, while referring to information such as the number of strokes per cycle of the engine (i.e., whether it is a four-stroke or two-stroke engine), the number of cylinders, and the engine rotation speed range (for example, 600 to 6000 rpm that can be used during inspection, or the range from the lower limit rotation speed at idle to the upper limit rotation speed limited by a limiter).

For example, in a four-stroke, four-cylinder engine, two explosions occur during one rotation of the crankshaft, so the possible fundamental frequency range corresponding to a rotational speed range of 600 to 6000 rpm is 20 to 200 Hz. Therefore, the fundamental frequency is identified by searching for the frequency within this 20 to 200 Hz frequency range that shows the highest peak in sound pressure in the two-dimensional frequency spectrum or three-dimensional spectrogram described above. The identification of the fundamental frequency by the fundamental frequency calculation unit 30 will be explained further below.

The overtone identification unit 40 uses the fundamental frequency identified by the fundamental frequency calculation unit 30 to extract data on harmonics of the fundamental frequency from the frequency data. In theory, harmonics occur at frequencies that are integer multiples of the fundamental frequency. For example, if the fundamental frequency is 150 Hz, then the peaks at integer multiples of that frequency, such as 300 Hz, 450 Hz, 600 Hz, 750 Hz, etc., are harmonics of the fundamental frequency. Therefore, the overtone identification unit 40 detects the presence or absence of peaks near frequencies that could be harmonics based on the frequency data and identifies the harmonics. In reality, harmonics are not measured as exact integer multiples of the fundamental frequency, but exist within a certain range. Therefore, a sound pressure peak is searched for within a frequency range of, for example, approximately ±10% of the integer multiple frequency, and this peak is identified as a harmonic.

In the first embodiment, it is sufficient to identify one harmonic, so an appropriate harmonic order is extracted taking into account the characteristics of the engine, vehicle, etc. being inspected. In order to detect changes over time in frequency and sound pressure, as described below, it is preferable to extract a harmonic order of a relatively high order (e.g., 6th, 9th, etc.). Alternatively, multiple harmonic orders can be identified, and the sound pressures of each can be compared to select the harmonic with the highest peak sound pressure.

Harmonic data is time-series data that includes the frequency and sound pressure of a specific harmonic. As mentioned above, for example, the continuous line of high sound pressure points in the spectrogram display of Figure 3 corresponds to harmonic data. Figure 3 includes multiple harmonics, and the time-series data for each individual harmonic is extracted as harmonic data. Note that, as the first embodiment focuses on changes in frequency over time, as will be described later, the harmonic data may not include sound pressure information.

The abnormal sound detector 50 determines whether the approximately 15 seconds of sound data contains abnormal sounds caused by misfires, based on the harmonic overtone data extracted by the harmonic overtone identification unit 40. Specifically, the abnormal sound detector 50 determines whether the frequency of the specific harmonic under consideration changes over time, or more specifically, whether the frequency of the specific harmonic overtone drops and then returns to within a predetermined time range corresponding to a misfire.

That is, in this embodiment, the abnormal sound detection unit 50 includes a frequency change amount calculation unit 51 and a frequency restoration time calculation unit 52. The frequency change amount calculation unit 51 calculates the amount of frequency change when the frequency temporarily drops in the harmonic time series data. The frequency restoration time calculation unit 52 calculates the time it takes for the dropped frequency to restore. The abnormal sound detection unit 50 determines that an abnormal sound, or misfire, has occurred, for example, when there is a frequency drop of more than a certain threshold and the frequency restores within a certain time.

When a misfire occurs in a cylinder of an engine, the rotational speed of the crankshaft drops slightly, causing a change in the sound of the engine. However, if a misfire occurs under operating conditions with a fundamental frequency of 150 Hz, and the rotational speed drops by 2%, the frequency drop is only 3 Hz, and the rotational speed returns to normal when the next cylinder explodes, so the frequency drop is very brief. Therefore, not only is it difficult for an inspector to determine this by hearing, but even when processing data collected by a microphone, the influence of noise and other factors is significant, making it impossible to determine this without using an ignition pulse to limit the timing to a narrow range, as in Patent Document 1.

On the other hand, if a 3 Hz frequency drop occurs in the fundamental frequency as in the example above, focusing on the fourth harmonic, for example, a 12 Hz frequency drop occurs, making the frequency drop during misfire more noticeable than the fundamental frequency. This makes it possible to detect misfire from engine sound alone, even without signals such as ignition timing. Furthermore, by determining whether the frequency has recovered in a short period of time in addition to the frequency drop, it is possible to distinguish between abnormal noise caused by misfire and sound changes or noise due to other factors.

In the spectrogram shown in Figure 3, which is an example, time changes in frequency corresponding to misfires (i.e., the frequency drops below the threshold and then recovers within a short period of time) can be seen at three locations indicated by arrows labeled "mf." Therefore, three misfires, indicated as MF1, MF2, and MF3, were detected within the approximately 15-second inspection period. As is clear from Figure 3, even when such misfires occur, there is almost no frequency change in the time series data for the fundamental frequency (F1 in Figure 3).

The display unit 60 is a display means for displaying the results of the judgment to relevant parties such as workers near the inspection device, inspectors driving the vehicles, managers, and data scientists who will use the data. It may include, for example, an LCD display or an organic EL display. Furthermore, if any alarm sounds or audio are involved, it may also include a sound source, amplifier, speaker, etc. for generating and emitting the sound. The images, videos, sounds, and audio displayed here are generated by image generation means and sound/audio generation means included in the display unit 60. Specific examples of what is displayed on the display will be explained later.

Figure 2 is a flowchart showing the processing flow of the misfire detection device of the first embodiment described above. First, the sound of the engine of a vehicle being test-run on a free roller is collected by the microphone of the sound measurement and acquisition unit 10 and acquired as sound data (Step 1). Next, the frequency conversion unit 20 described above uses frequency analysis techniques such as FFT (Fast Fourier Transform) and wavelet analysis to convert the sound data according to frequency and generate frequency data. More specifically, frequency analysis is performed for each relatively short time domain, and frequency data for each time is generated (Step 2).

Next, the fundamental frequency is identified using the frequency data as described above (step 3), and then appropriate harmonic time series data is extracted based on the fundamental frequency (step 4).
Next, it is determined whether the frequency of this harmonic has dropped by more than a threshold (step 5). For example, it is determined whether the frequency of the target harmonic has dropped by more than 10 Hz in one second. Note that "more than 10 Hz in one second" is just one example, and the appropriate threshold will vary depending on, for example, the order of the target harmonic. In other words, it is sufficient if the value is appropriate for detecting a sudden frequency drop caused by a misfire.

If the frequency drops by more than the threshold, it is further determined whether the frequency drop has recovered in a short time (step 6). For example, it is determined whether the original frequency has returned within 0.02 seconds. Note that "0.02 seconds" is merely an example, and will vary depending on the number of cylinders in the engine, etc. In other words, an appropriate value is set to identify whether the frequency drop is due to a misfire. As an example, "0.02 seconds" corresponds to the time interval (two crankshaft revolutions) between a misfire in a cylinder at 6000 rpm in a four-stroke engine and the next explosion in the same cylinder. Normally, the frequency will recover due to an explosion in another cylinder.

If the determinations in both steps 5 and 6 are YES, proceed to step 7, determine that a misfire has occurred, and display a message to that effect.

If the determination in either step 5 or step 6 is NO, proceed to step 8, determine that the engine is normal (no misfire), and display a message to that effect.

Finally, in step 9, it is determined whether retesting is necessary for some reason. If retesting is necessary, return to step 1; if not, end the test.

Next, the determination of the fundamental frequency in the fundamental frequency calculation unit 30 will be further explained with reference to Figures 4 and 5. Figure 4 is a flowchart showing the process for determining the fundamental frequency. First, engine specification data is obtained from a database included in the finished vehicle inspection device (step 101). This specification data includes at least information on the number of strokes per cycle of the engine (i.e., whether it is a four-stroke engine or a two-stroke engine), the number of cylinders, and the engine rotational speed range (for example, 600 to 6000 rpm that can be used during inspection, or the range from the lower limit rotational speed at idle to the upper limit rotational speed limited by a limiter). Next, the frequency range of the fundamental frequency corresponding to the engine rotational speed range is calculated (step 102). For example, for a four-stroke, four-cylinder engine, the possible fundamental frequency range corresponding to a rotational speed range of 600 to 6000 rpm is 20 to 200 Hz.

Next, proceed to step 103, where FFT frequency analysis is performed on the sound data shown in Figure 5(a) (original sound data with a duration of approximately 15 seconds) for each relatively short time region (first between T1 and T2) to generate frequency data. This results in a two-dimensional frequency spectrum for the period T1 to T2, as shown in Figure 5(b). All frequencies in this frequency spectrum that exhibit convex peaks in sound pressure or power are then extracted (step 104). From these multiple peak frequencies, the peak frequency that is within the aforementioned possible fundamental frequency range (e.g., 20 to 200 Hz) and represents the maximum sound pressure peak is extracted (step 105), and this is determined as the fundamental frequency for the period T1 to T2 (step 106). Figure 5(c) shows a portion of the spectrogram display mentioned above, and as shown in Figure 5(c), the fundamental frequency F1t1 for the period T1 to T2 is determined.

Once the sound data between T1 and T2 has been processed in this way, the target time region is incremented in step 108, and steps 103 to 106 are performed for the next period T2 to T3. In other words, the sound data between T2 and T3 is frequency analyzed using FFT, multiple peak frequencies are extracted from the obtained frequency spectrum, and the peak frequency that is within the possible fundamental frequency range and has the maximum sound pressure peak is extracted from that and determined as the fundamental frequency between T2 and T3. As a result, the fundamental frequency F1t2 between times T2 and T3 is determined, as shown in Figure 5(c).

By repeating steps 103 to 106 for each individual time domain, the fundamental frequency is determined for each time domain. In step 107, it is determined whether the entire time range of the sound data has been processed. Once all processing of the sound data, which has a duration of approximately 15 seconds, has been completed, the identification of the fundamental frequency is complete. This results in the fundamental frequency being obtained in the form of time-series data.

The individual fundamental frequencies (F1t1, F1t2, F1t3, etc.) calculated for each time region basically have continuous characteristics. If there are any discontinuous areas, you can simply ignore the values in that section and make the previous and next values continuous. Alternatively, you can proceed with identifying harmonics from the values of areas that are generally continuous.

The fundamental frequency may also be determined by other methods. For example, if there are multiple frequency peaks (convex mountain-shaped) in a two-dimensional frequency spectrum, the smallest peak frequency is the fundamental frequency. Alternatively, if two or more peak frequencies are detected, the frequency difference between two adjacent peaks may be determined to be the fundamental frequency.

As described above, the misfire detection device of the above embodiment uses a microphone placed outside the vehicle to collect engine sounds and process the signals, making it possible to detect misfires without obtaining signals such as ignition timing from the vehicle. In particular, it focuses on the harmonics of the fundamental frequency, the range of change of which increases due to misfire, and determines misfires based on the temporary drop and recovery of frequency associated with misfire, thereby ensuring reliable misfire detection even in environments where other noise is present. Furthermore, misfire detection can be efficiently performed during test runs on free rollers during the completed vehicle inspection process.

Next, a misfire detection device according to a second embodiment will be described with reference to Figures 6 and 7. The following mainly focuses on the differences from the first embodiment. The misfire detection device according to the second embodiment determines whether abnormal sounds corresponding to misfires are contained in sound data based on a drop in the sound pressure of overtones, instead of a drop in the frequency of overtones.

As shown in the block diagram of Figure 6, the misfire detection device of the second embodiment, like the first embodiment, is configured to include a sound measurement/acquisition unit 10, a frequency conversion unit 20, a fundamental frequency calculation unit 30, a harmonic identification unit 40, an abnormal sound detection unit 50, and a display unit 60. Here, the abnormal sound detection unit 50 includes a sound pressure change amount calculation unit 53 and a sound pressure restoration time calculation unit 54 instead of the frequency change amount calculation unit 51 and frequency restoration time calculation unit 52 of the first embodiment. Note that in the second embodiment, the harmonic data generated by the harmonic identification unit 40 includes at least sound pressure, and may not include time series information on frequency.

Sound pressure is represented by the brightness and color of each point in the spectrogram display of Figure 3, for example, and by the height of the peak in a two-dimensional frequency spectrum. When a misfire occurs in an engine, a temporary drop in sound pressure occurs because the energy of the explosion is not added. However, this drop in sound pressure due to a misfire is impossible to detect in sound data that contains a mixture of many frequencies and noise. In contrast, as in this embodiment, by extracting data on specific harmonics (time-series data on the sound pressure of harmonics) and focusing on the time changes in the sound pressure of these harmonics, it is possible to detect changes in sound pressure due to a misfire.

Figure 7 is a flowchart showing the processing flow for misfire testing in the second embodiment. After appropriate harmonic time series data is extracted in steps 1 to 4, step 5A determines whether the sound pressure of that harmonic has dropped by more than a threshold. For example, it determines whether the sound pressure of the target harmonic has dropped by 3 dB or more in one second. Note that "3 dB or more in one second" is merely an example.

If there is a drop in sound pressure equal to or greater than the threshold, it is further determined whether the drop in sound pressure is restored in a short period of time (Step 6A). For example, it is determined whether the sound pressure returns to its original level within 0.02 seconds. Note that "0.02 seconds" is merely an example.

If the determinations in both Steps 5A and 6A are YES, proceed to Step 7, determine that a misfire has occurred, and display a message to that effect.

If the determination in either step 5A or step 6A is NO, proceed to step 8, determine that the engine is normal (no misfire), and display a message to that effect.

Next, a misfire detection device according to a third embodiment will be described with reference to Figures 8 and 9. The misfire detection device according to the third embodiment determines whether abnormal sounds corresponding to misfires are contained in sound data based on both a drop in harmonic frequency and a drop in harmonic sound pressure. Specifically, a misfire is determined to have occurred when a drop in harmonic frequency and a drop in sound pressure occur substantially simultaneously.

As shown in the block diagram of Figure 8, the misfire detection device of the third embodiment, like the first embodiment, is configured to include a sound measurement/acquisition unit 10, a frequency conversion unit 20, a fundamental frequency calculation unit 30, a harmonic identification unit 40, an abnormal sound detection unit 50, and a display unit 60. Here, the abnormal sound detection unit 50 includes a frequency change amount calculation unit 51 and a frequency restoration time calculation unit 52, as in the first embodiment, and a sound pressure change amount calculation unit 53 and a sound pressure restoration time calculation unit 54, as in the second embodiment. Note that in the third embodiment, the harmonic data generated by the harmonic identification unit 40 includes both frequency information and sound pressure information.

Figure 9 is a flowchart showing the processing flow for misfire testing in the third embodiment. After appropriate harmonic time series data is extracted in steps 1 to 4, step 5B determines whether the harmonic has experienced a frequency drop of more than a threshold and a sound pressure drop of more than a threshold. For example, it determines whether the frequency of the target harmonic has dropped by more than 10 Hz in one second, and whether the sound pressure has dropped by more than 3 dB in one second.

If there is a frequency drop or sound pressure drop that is equal to or greater than the threshold, it is further determined whether these frequency drops and sound pressure drops are restored in a short period of time (Step 6B). For example, it is determined whether the original frequency and sound pressure are restored within 0.02 seconds.

If the determinations in both Steps 5B and 6B are YES, proceed to Step 7, determine that a misfire has occurred, and display a message to that effect.

If the determination in either step 5B or step 6B is NO, proceed to step 8, determine that the engine is normal (no misfire), and display a message to that effect.

In this embodiment, by requiring both a decrease in frequency and a decrease in sound pressure, misfire detection becomes more strict and false detections due to noise, etc. are reduced.

Next, a misfire detection device according to a fourth embodiment will be described with reference to Figures 10 and 11. The misfire detection device according to the fourth embodiment determines whether an abnormal sound corresponds to a misfire based on a drop in frequency and a drop in sound pressure of multiple harmonics. Specifically, a misfire is determined to have occurred when a drop in frequency and a drop in sound pressure occur substantially simultaneously for multiple harmonics.

As shown in the block diagram of Figure 10, the misfire detection device of the fourth embodiment, like the third embodiment, is composed of a sound measurement/acquisition unit 10, a frequency conversion unit 20, a fundamental frequency calculation unit 30, a harmonic identification unit 40, an abnormal sound detection unit 50, and a display unit 60. Here, the harmonic identification unit 40 extracts harmonic data (time series data of frequency and sound pressure) for all extractable harmonics. For example, if the fundamental frequency is 150 Hz, harmonics exist near integer multiples of that, such as 300 Hz, 450 Hz, 600 Hz, 750 Hz, etc., and all of these harmonics are identified as much as possible and time series data of their frequencies and sound pressures is generated.

As in the third embodiment, the abnormal sound detection unit 50 includes a frequency change amount calculation unit 51, a frequency restoration time calculation unit 52, a sound pressure change amount calculation unit 53, and a sound pressure restoration time calculation unit 54, and further includes a harmonic number calculation unit 55 that calculates the number of harmonics in which time changes in frequency, etc. appear substantially simultaneously. The abnormal sound detection unit 50 determines that a misfire has occurred when the number of harmonics in which time changes in frequency, etc. appear substantially simultaneously is equal to or greater than a predetermined number.

Figure 11 is a flowchart showing the processing flow for misfire testing in the fourth embodiment. After identifying the fundamental frequency in steps 1 to 3, time-series data for multiple (all extractable) harmonics is extracted in step 4. Then, in step 5C, it is determined whether these multiple harmonics have experienced a frequency drop of more than a threshold and a sound pressure drop of more than a threshold. For example, it is determined whether the frequency of the targeted harmonic has dropped by more than 10 Hz in one second and the sound pressure has dropped by more than 3 dB in one second. Note that the threshold may be different for each individual harmonic.

If there is a frequency drop or sound pressure drop that is equal to or greater than the threshold, it is further determined whether these frequency drops and sound pressure drops are restored in a short period of time (Step 6C). For example, it is determined whether the original frequency and sound pressure are restored within 0.02 seconds.

If there is a phenomenon in which at least one harmonic has experienced a frequency drop and sound pressure drop of more than the threshold value and has recovered within a predetermined time, then in step 10, it is determined whether such a phenomenon (frequency drop and sound pressure drop and their recovery) occurred substantially simultaneously for a predetermined number or more of harmonic overtones, for example, two or more.

If the above phenomenon occurs simultaneously for two or more harmonics, proceed to step 7, determine that a misfire has occurred, and display a message to that effect.

If there is a drop in frequency and sound pressure above the threshold and no overtones are restored within the specified time, or if only one overtone is detected, proceed to step 8, determine that the condition is normal (no misfire), and display a message to that effect.

In this embodiment, by using the occurrence of time-dependent changes in frequency and sound pressure in multiple harmonics as a weighting condition, misfire detection becomes more stringent and false detections due to noise, etc. are reduced.

In the above example, both the time change in frequency and the time change in sound pressure are detected for multiple harmonics, but it is also possible to determine only the time change in frequency, as in the first embodiment, or only the time change in sound pressure, as in the second embodiment.

Next, a misfire detection device according to a fifth embodiment will be described with reference to Figures 12 and 13. Like the fourth embodiment, the misfire detection device according to the fifth embodiment determines whether a change in frequency, etc., has occurred over time in multiple harmonics, and checks whether the amount of frequency drop in each harmonics correlates correctly with the order of the harmonic. Specifically, if the amount of frequency drop in multiple harmonics correlates correctly with the order of each harmonic, it is determined to be a misfire. However, even if a drop in frequency and sound pressure has occurred in multiple harmonics, if the amount of frequency drop does not correlate correctly with the order, it is determined to be noise or the like, and not a misfire.

For example, if the third harmonic causes a frequency drop of 10 Hz, then the sixth harmonic should cause a frequency drop of 20 Hz. If this correlation is not maintained, there is a possibility that some kind of noise is present. Therefore, in the fifth embodiment, in such cases, a misfire is not determined.

As shown in the block diagram of Figure 12, the misfire detection device of the fourth embodiment, like the third embodiment, is configured to include a sound measurement/acquisition unit 10, a frequency conversion unit 20, a fundamental frequency calculation unit 30, a harmonic identification unit 40, an abnormal sound detection unit 50, and a display unit 60. Here, like the fourth embodiment, the harmonic identification unit 40 extracts harmonic data (time series data of frequency and sound pressure) for all extractable harmonics.

As in the fourth embodiment, the abnormal sound detection unit 50 includes a frequency change amount calculation unit 51, a frequency restoration time calculation unit 52, a sound pressure change amount calculation unit 53, a sound pressure restoration time calculation unit 54, and a harmonic number calculation unit 55 that calculates the number of harmonics for which time changes in frequency, etc. appear substantially simultaneously. Furthermore, the abnormal sound detection unit 50 includes a harmonic order/change amount calculation unit 56 that calculates the order and frequency drop amount of each of multiple harmonics for which time changes in frequency, etc. appear substantially simultaneously. The abnormal sound detection unit 50 determines whether the frequency drop amounts for the multiple harmonics correspond to the order of each harmonic, and then makes a final misfire detection determination.

Figure 13 is a flowchart showing the processing flow for misfire testing in the fifth embodiment. As in the fourth embodiment, after extracting time-series data for multiple (all extractable) harmonics in steps 1 to 4, step 5D determines whether these multiple harmonics have experienced a frequency drop of more than a threshold and a sound pressure drop of more than a threshold. For example, it determines whether the frequency of the targeted harmonic has dropped by 10 Hz or more in one second, and whether the sound pressure has dropped by 3 dB or more in one second. Note that the threshold may be different for each individual harmonic.

If there is a frequency drop or sound pressure drop that is equal to or greater than the threshold, it is further determined whether these frequency drops and sound pressure drops are restored in a short period of time (Step 6D). For example, it is determined whether the original frequency and sound pressure are restored within 0.02 seconds.

If there is a phenomenon in which at least one harmonic has experienced a frequency drop and sound pressure drop of more than the threshold value and has recovered within a predetermined time, then in step 10, it is determined whether such a phenomenon (frequency drop and sound pressure drop and their recovery) occurred substantially simultaneously for a predetermined number or more of harmonic overtones, for example, two or more.

If the above phenomenon occurs simultaneously for two or more harmonics, proceed to step 11 and determine whether the frequency downshift for each harmonic correlates correctly with the order of each harmonic. Since the actual frequency downshift will not be a strict integer multiple, it is advisable to add an appropriate tolerance when making this determination.

If the answer in step 11 is YES, meaning that there is a frequency drop and sound pressure drop greater than the threshold and for multiple harmonics that have been restored within the specified time, the amount of each frequency drop is correctly correlated with the order of each harmonic, proceed to step 7, determine that a misfire has occurred, and display a message to that effect.

If there is a frequency drop and sound pressure drop above the threshold and no overtones are restored within the specified time, or if this is detected in only one overtone, or if the frequency drop in multiple overtones does not correspond to the order of the overtone, proceed to step 8, determine that the condition is normal (no misfire), and display a message to that effect.

In this embodiment, for multiple harmonics where a drop in frequency has occurred, the weighting condition is that the amount of frequency drop in each harmonic corresponds to the order of the harmonic, making misfire detection more stringent and reducing false detections due to noise, etc.

Next, an example of the display displayed on the display unit 60 is shown in Figure 14. In this example, to visually represent changes in fundamental frequency and harmonics, the aforementioned spectrogram display is placed on the right side of the screen (the area indicated by symbol A), and the original sound data is displayed as a two-dimensional graph below the spectrogram display. As mentioned above, the spectrogram is actually displayed in color. In this example, two misfires were detected during the approximately 15-second inspection period, and red inverted triangular arrows indicated by symbols B1 and B2 are displayed above the spectrogram display to indicate the timing of each misfire. This makes it possible to understand the operating conditions under which the misfire occurred. In addition, in the example shown, the areas on the spectrogram where changes over time in the frequency, etc., determined to be misfires appear are indicated by red ellipses (indicated by symbols H1 and H2) surrounding them.

Also, in the upper left corner of the screen, as indicated by symbol C, a relatively large label will indicate whether or not a misfire has occurred. In the illustrated example, a misfire has been detected, so "Misfire Present" is displayed in red. If no misfire has occurred, the display will read "No Misfire." Furthermore, if a misfire has occurred, the amount of sound pressure reduction (indicated by symbol D), the amount of frequency reduction (indicated by symbol E), and the number of misfires (indicated by symbol F) will be displayed below part C.

Also, a button (indicated by the symbol G) representing a speaker is located at the bottom left side. When this speaker button G is pressed via a touch screen or the like, sound data edited over a relatively short period of time, including the misfire portion, is played back for the inspector to listen to. This allows the inspector to identify the engine sound associated with the misfire.

The above display is one example; for example, the result of an abnormal noise determination could be simply displayed as OK/NG letters, or as a circle or cross symbol. It is desirable to avoid displaying too much information on the screen so that it can be understood in a short amount of time.

In addition, in the above display example, a spectrogram is used to visually show the time changes in fundamental frequency and harmonics, but other displays that visualize sound, such as a mel spectrogram or keptogram, may also be used.

Next, as a different embodiment of the abnormal sound detection of the present invention, the spectrogram display may be treated as image data and processed using a pre-created abnormal sound detection model. That is, as described above, a spectrogram display corresponding to the sound data can be generated using frequency data obtained by frequency analysis of the sound data. However, an abnormal sound detection model can be generated in advance using machine learning (or deep learning) of a large number of spectrogram display image data, including inspection examples with and without misfires. By processing the spectrogram display of the sound of the engine being inspected as image data using the abnormal sound detection model, it is possible to determine the presence or absence of abnormal sounds from characteristic changes in the harmonic data on the image.

The above describes one embodiment in which this invention is applied to detecting engine misfires during the finished vehicle inspection process, but the invention is not limited to this and can be applied in a variety of other ways. For example, misfires can be easily detected during inspections and maintenance at commercial vehicle dealerships or repair shops. Furthermore, the invention can be applied to detecting not only misfires but also abnormal combustion in engines, abnormalities in mechanical components, and other abnormalities that are reflected in the engine sound.

Furthermore, in the above embodiment, misfire detection is performed while the engine rotation speed is increased, but it is also possible to detect misfire while maintaining a constant rotation speed.

Note that the information presentation unit is not limited to the above-mentioned display unit that displays images, but may also present information using audio only.

10: Sound measurement and acquisition unit 20: Frequency conversion unit 30: Fundamental frequency calculation unit 40: Harmonic overtone identification unit 50: Abnormal sound determination unit 60: Display unit

Claims (14)

Acquire the sound produced by the engine to generate sound data,
This sound data is converted according to frequency to generate frequency data,
determining a fundamental frequency associated with the rotational speed of the engine based on the frequency data;
Using this fundamental frequency, data on harmonics of the fundamental frequency is extracted from the frequency data.
Based on the harmonic data, it is determined whether or not there is an abnormal sound contained in the sound data.
The results of this assessment will be presented.
A method for detecting abnormal noise in an engine, comprising:
determining whether or not an abnormal sound is present based on a time change in the frequency of at least one harmonic;
How to detect abnormal engine noise. Acquire the sound produced by the engine to generate sound data,
This sound data is converted according to frequency to generate frequency data,
determining a fundamental frequency associated with the rotational speed of the engine based on the frequency data;
Using this fundamental frequency, data on harmonics of the fundamental frequency is extracted from the frequency data.
Based on the harmonic data, it is determined whether or not there is an abnormal sound contained in the sound data.
The results of this assessment will be presented.
A method for detecting abnormal noise in an engine, comprising:
A method for detecting abnormal noise in an engine , which determines whether or not an abnormal noise is present based on the time variation of the frequency and sound pressure of at least one harmonic. The engine abnormal noise detection method described in claim 1 or 2, in which a change in harmonics associated with an engine misfire is determined to be the abnormal noise. 2. The method for detecting abnormal noise in an engine according to claim 1 , wherein the occurrence of the abnormal noise is determined to be due to a misfire when the frequency of the overtones drops and returns to normal within a predetermined time range corresponding to a misfire. Acquire the sound produced by the engine to generate sound data,
This sound data is converted according to frequency to generate frequency data,
determining a fundamental frequency associated with the rotational speed of the engine based on the frequency data;
Using this fundamental frequency, data on harmonics of the fundamental frequency is extracted from the frequency data.
Based on the harmonic data, it is determined whether or not there is an abnormal sound contained in the sound data.
The results of this assessment will be presented.
A method for detecting abnormal noise in an engine, comprising:
When the sound pressure of at least one harmonic has decreased and returned to normal within a predetermined time range corresponding to a misfire, it is determined that the abnormal noise is due to a misfire.
How to detect abnormal engine noise. 3. The engine abnormal noise detection method according to claim 2 , wherein when the frequency and sound pressure of the overtones decrease and return to normal within a predetermined time range corresponding to a misfire, it is determined that the abnormal noise is caused by a misfire. The method for detecting abnormal engine noise according to any one of claims 1 to 6 , wherein the sound of the engine mounted on the vehicle is collected by a microphone outside the vehicle. The engine abnormal noise detection method according to any one of claims 1 to 7 , wherein the fundamental frequency is specified by identifying a frequency that indicates a peak in sound pressure within a possible fundamental frequency range based on the number of cylinders of the engine and the rotational speed range of the engine. The engine abnormal noise detection method according to any one of claims 1 to 8 , wherein frequencies showing sound pressure peaks in the vicinity of integer multiples of the fundamental frequency are identified as harmonics. It determines whether multiple harmonics each experience a drop in frequency at the same time,
Determine whether each frequency downshift corresponds to each order;
2. The method for detecting abnormal noise in an engine according to claim 1 , wherein an abnormal noise is determined to exist when the noise corresponds to each of the orders. generating a spectrogram representation corresponding to the sound data using the frequency data;
A model for detecting abnormal noise is generated by machine learning using a large amount of spectrogram image data in advance.
The method for detecting abnormal noise from an engine according to any one of claims 1 to 10 , further comprising processing a spectrogram display of the sound of the engine to be inspected with the abnormal noise determination model to determine whether or not an abnormal noise is present. a sound acquisition unit that acquires sounds generated from the engine and generates sound data;
a frequency data generating unit that generates frequency data by converting the sound data according to the frequency;
a fundamental frequency calculation unit that determines a fundamental frequency associated with the rotation speed of the engine based on the frequency data;
a harmonic identification unit that uses the fundamental frequency to extract data on harmonics of the fundamental frequency from the frequency data;
an abnormal sound determination unit that determines whether or not an abnormal sound is included in the sound data based on the harmonic data;
an information presentation unit that presents the result of the determination;
An engine abnormal noise detection device comprising:
the abnormal sound determination unit determines the presence or absence of an abnormal sound from a time change in frequency of at least one harmonic.
Engine noise detection device. a sound acquisition unit that acquires sounds generated from the engine and generates sound data;
a frequency data generating unit that generates frequency data by converting the sound data according to the frequency;
a fundamental frequency calculation unit that determines a fundamental frequency associated with the rotation speed of the engine based on the frequency data;
a harmonic identification unit that uses the fundamental frequency to extract data on harmonics of the fundamental frequency from the frequency data;
an abnormal sound determination unit that determines whether or not an abnormal sound is included in the sound data based on the harmonic data;
an information presentation unit that presents the result of the determination;
An engine abnormal noise detection device comprising:
the abnormal sound determination unit determines the presence or absence of an abnormal sound from time changes in frequency and sound pressure of at least one harmonic.
Engine noise detection device. a sound acquisition unit that acquires sounds generated from the engine and generates sound data;
a frequency data generating unit that generates frequency data by converting the sound data according to the frequency;
a fundamental frequency calculation unit that determines a fundamental frequency associated with the rotation speed of the engine based on the frequency data;
a harmonic identification unit that uses the fundamental frequency to extract data on harmonics of the fundamental frequency from the frequency data;
an abnormal sound determination unit that determines whether or not an abnormal sound is included in the sound data based on the harmonic data;
an information presentation unit that presents the result of the determination;
An engine abnormal noise detection device comprising:
the abnormal noise determination unit determines that the abnormal noise is caused by a misfire when the sound pressure of at least one harmonic has decreased and returned to normal within a predetermined time range corresponding to a misfire.
Engine noise detection device.