JP2006170674A - Abnormal noise detector - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an abnormal noise detector capable of capturing efficiently the waveform of an abnormal noise caused by air separation or the like in the noisy state. <P>SOLUTION: This detector for detecting an abnormal noise in the noisy state has an acoustic sensor 3 for detecting the abnormal noise; a band-pass filter 6 for inputting a signal from the acoustic sensor 3, and allowing passage of only a frequency band set beforehand; and a differential filter 7 for dividing in time an input signal from the band-pass filter 6, and outputting a numerical value determined by subtracting a value at the (t-1) time from a value at the t time as a signal at the t time. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to an abnormal sound detector that detects an abnormal sound from a mixed signal detected under noise. More specifically, the present invention relates to an abnormal sound detector that detects leakage sound (abnormal sound) of gas or liquid such as air or gas from sounds measured under ambient noise.

  In general, control piping such as an engine valve of an automobile uses the level of air pressure (negative pressure in the pipe) to link the control of other parts. However, the control piping is driven by engine vibration or automobile running vibration. If the tube is pulled out or loosened due to vibration, or deteriorates over time and cracks occur in the vicinity of the insertion port, air will flow in from the defective portion, resulting in problems such as valve control being disturbed. Although control for engine control is performed using manifold negative pressure, the engine malfunctions due to poor control operation due to hose deterioration or the like, or contamination of secondary air. In such a state, it is necessary to find a deteriorated part or a defective part in an automobile maintenance shop.

  2. Description of the Related Art Conventionally, in an automobile maintenance factory or the like, since it is not probable that the leakage inspection such as air leakage will be digitized, a sensory inspection based on sound and vibration for a long time is forced in an idling state. However, under a large engine room noise caused by the engine operation of an automobile, the leaking jet noise (abnormal sound) of the control pipe in the audible band due to the malfunction of the control pipe is masked by the engine noise and is human. Judgment is impossible with the sense of hearing.

For this reason, for example, abnormal sounds (pneumatic leakage noise) are heard from a variety of sounds during engine rotation using a leak tester or the like. In an inspection method using a leak tester, etc., when detecting abnormal noise leaking from pipes, etc., if there is a gas leak from pipes, Since the mixed sound with the background sound generated due to the flow of the gas is generated, the background sound is deleted from the mixed sound in order to detect the leakage sound of the gas from the mixed sound.
Also, in the inspection using a leak tester etc., because it is a detection method that picks up abnormal ultrasonic waves, it is necessary to remove various noises such as ultrasonic waves generated from gear friction sounds of the engine in advance, so the engine is stopped and simulated. A method of generating a negative pressure was used.

In observing such a leak, the following techniques have been conventionally disclosed. For example, there is a technique that detects a leak from a difference between both signals detected by a leak detection acoustic sensor and a dark noise elimination acoustic sensor (see, for example, Patent Document 1).
In addition, a leak sound detection sensor and a plant noise detection sensor are installed in the pipe of the plant, and calculation means for removing the output of the plant noise detection sensor from the output of the plant leak sound detection sensor is provided to detect the leak sound of the pipe. (For example, refer to Patent Document 2).
In addition, there is one in which the first acoustic sensor and the second acoustic sensor are installed in the pipe upstream or downstream of the pipe, or arranged in a non-contact manner, and a leak signal from the pipe is obtained from the output difference between the two sensors. (For example, refer to Patent Document 3).

JP-A-3-110436 Japanese Patent Laid-Open No. 1-213539 JP-A-1-187430

  However, the above-mentioned conventional leakage observation method is proposed ignoring the fact that ultrasonic waves are generated even when the gears of the machine are rubbed or peeled, and that ultrasonic waves are also observed in the air leaking jet sound of the control pipe. It does not take into account the noise of the entire factory and the interior of automobiles and the magnitude of diffuse reflection vibration, and automatic inspection is still not progressing.

Therefore, the abnormal sound detector of the present invention is different from conventional acoustic inspection devices and ultrasonic leak inspection devices in that it aims to efficiently extract abnormal sound waveforms such as air separation under noise. is there.
In addition, the present invention detects the presence or absence of control pipe defects by detecting the presence or absence of air leaking jet noise in the control pipe resulting from the operation of the engine under noise in the engine room that occurs in a wide frequency band from audible frequencies to ultrasonic waves. An object of the present invention is to provide an abnormal sound detector capable of determining the noise.
Furthermore, an object of the present invention is to provide an abnormal noise detector that can be applied to measurement of leakage of a pressure vessel of a power plant, a piping of a plant or a control piping of an automobile engine room.

An abnormal sound detector according to claim 1 of the present invention is a detector that detects abnormal sound under noise, and is configured to input an acoustic sensor that detects the abnormal sound and a signal from the acoustic sensor in advance. A band-pass filter that passes only the set frequency band and an input signal from the band-pass filter are divided in time, and a value obtained by subtracting the value of (t-1) time from the value of t time is a signal at t time. And a difference filter that outputs as a difference filter.
According to a second aspect of the present invention, in the noise detector of the first aspect, the signal processing in the differential filter is performed by sampling the output signal from the band pass filter by a switching element, and between adjacent two points on the time axis. The difference is taken.
According to a third aspect of the present invention, in the abnormal sound detector according to the first or second aspect, the signal processing in the differential filter is a process of multiplying and adding the coefficient sequence of the filter to the band-pass filter output data signal-processed by the AD converter. It is characterized by being.
According to a fourth aspect of the present invention, in the noise detector according to any one of the first to third aspects, the band-pass filter selects a high-pass filter and a low-pass filter by switching switch elements. And
The abnormal noise detector according to claim 5 is the abnormal noise detector according to any one of claims 1 to 4, wherein the abnormal noise is caused by a pinhole or a crack in a piping of an automobile engine room, a power plant pressure vessel, a plant, etc. It originates from a gas or liquid jet.

The abnormal sound detector of the present invention can efficiently extract a waveform of air separation unlike a conventional acoustic inspection device or ultrasonic leak inspection device.
In addition, the abnormal noise detector of the present invention detects the presence or absence of air leaking jet noise in the control piping resulting from the operation of the engine under the noise of the mechanism in the engine room that occurs in a wide frequency band from audible frequencies to ultrasonic waves. The presence or absence of defects in the control piping can be determined.
Furthermore, the abnormal noise detector of the present invention can be applied to measure leakage of a control pipe of a car engine room, a power plant pressure vessel, a plant or the like.
When applying to an automobile engine room, keep the engine running without stopping and hold the tester in the engine room and use sound algorithm analysis to find leaked and defective parts accurately and safely in a short time. Can do.
In addition, because of non-contact work, the occurrence of injuries and accidents can be reduced, and areas other than the audible area (area that cannot be heard by human ears) can be detected.
Furthermore, it is possible to cope with detection of air leaks from pipes and the like in various facilities operating in the factory other than automobile engines. For example, a jet of air, gas, or liquid can be detected from pinholes (cracks) or cracks (cracks) such as power plant pressure vessels and plant piping.

  Hereinafter, the best embodiment of the present invention will be described. In the present invention, examples of measurement objects using an abnormal sound detector include control piping in an automobile engine room, a pressure vessel in a power plant, piping in a plant, and the like. In addition, when there is liquid in the pipe, it may not be a vigorous jet due to its viscosity, and jet noise may not be generated, so it is desirable to inspect for leakage in a gas-filled state. .

  In general, a measurement object is mixed with noise that is inherent vibration generated during operation. For example, in an engine room of an automobile, noise generated during engine operation is always generated, and these are sounds generated by friction between the gas and the gas when the gas flows in the piping. In the present invention, the abnormal sound detector of the present invention detects leakage from the measurement object under the noise in which these noises exist.

The sound is generally vibrational and is thought to be a waveform that vibrates cleanly as in the sin-cos trigonometric world. When a rod or film vibrates, such a waveform is obtained, but a waveform generated by collision or separation is a non-linear waveform called a pulse. Non-linear waveforms (pulses) include both audible and ultrasonic sounds.
From the experience so far, noise generated by machines in the engine room includes acoustic noise of low frequency (100 Hz) to 5 kHz and ultrasonic noise of 30 kHz or more. In particular, a sharp peak exists at 40 kHz, 80 kHz of its overtone, and 120 kHz of its overtone. The noise in the band during this period is attenuated and is hardly detected as noise.

Conventionally, it has been common knowledge of an air leak detector to detect ultrasonic vibration of 40 kHz caused by a jet due to air leakage, but there are many sound sources of the same 40 kHz in the engine room, and the conventional ultrasonic detector It was found that can not be used.
Here, there is an opening of 8 times (3 octaves) between the upper limit of the audible range of 5 kHz and the ultrasonic wave of 40 kHz, and a 1 octave region from 10 kHz to 20 kHz becomes a window without noise in consideration of dispersion of each noise. Therefore, we tried to detect whether this area could not be detected.
Then, a relatively large air leak in the piping on the upper surface of the engine room could be detected well. In other words, the energy of the shock wave generated by the air layer separation when ejected from the air jet or pores is not limited to the 40 kHz ultrasonic wave in which the air easily vibrates, but relatively high frequency audible sound around 10 kHz that feels “shoe shoe” by hearing. Although it is often produced and cannot be discerned by hearing with other noise sources, it can be discriminated by selecting a frequency.

On the other hand, there is a frequency window with a low noise level in the region from 5 kHz to 40 kHz based on the measurement of the noise distribution of the mechanical part in the engine room. From this measurement result, the sound due to the air leakage jet in the control pipe is found in the region from 5 kHz to 40 kHz. It was also found that vibration can be detected even in the environment of the mechanical noise in the engine room.
On the other hand, using differential filters, starting with control piping in automobile engine rooms, pressure holes in power plants, plant piping, etc., pinholes (fine holes) and cracks (cracks), air, gas or liquid An object of the present invention is to provide an abnormal noise detector for detecting a jet.
For example, when there is gas in the pressure vessel or piping and positive or negative pressure is applied, and there are holes or cracks in the piping, etc., and a jet of air or gas is generated from that location, The jet of air radiates sound distributed from an ultrasonic wave of 40 kHz to an audible frequency of several kHz, and the leaked portion becomes a sound source for generating abnormal sound.

  Hereinafter, the abnormal sound detector of the present invention will be described in detail with reference to the drawings. FIG. 1 is a schematic explanatory diagram of the abnormal sound detector (A) according to the embodiment. In FIG. 1, reference numeral (1) denotes piping, and examples thereof include control piping in an automobile engine room, a pressure vessel in a power plant, or piping in a plant. In the state where the abnormal sound detector of the present invention is used, it is necessary that gas or liquid is applied at a positive pressure or a negative pressure in the pipe (1). Reference numeral (2) is a leakage point (abnormal sound source). For example, a control pipe or a pressure vessel in an automobile engine room includes a pinhole (a minute hole) or a crack (a crack), and air, When a jet of gas or liquid is generated, the presence or absence is detected using the abnormal sound detector of the present invention.

Reference numeral (3) denotes an acoustic sensor, and examples of the acoustic sensor include an acceleration pickup that directly detects vibration of a pipe, an ultrasonic sensor that captures airborne sound in an ultrasonic band, and an acoustic microphone that captures in an audible region. Since the acoustic microphone can be applied at low cost, it can be preferably applied in the present invention. The acoustic sensor 3 may be either a contact type or a non-contact type. Reference numeral (4) denotes a sensor power supply, which is provided to drive the acoustic sensor (3), and preferably includes an amplifier (5) for increasing the signal strength.
The amplifier (5) can amplify the output from the acoustic sensor (3) so that the jet sound signal generated from the leaked part (2) becomes a vibration voltage sufficient to be used in the subsequent electronic circuit.

Reference numeral (6) denotes a band-pass filter, and the input signal is a signal having a frequency band lower than the set frequency and a frequency band higher than another set frequency blocked by the band-pass filter (6) and output. For example, a sound up to 5 kHz in a low frequency band such as mechanical noise is blocked.
The band pass filter (6) may be configured as a single filter, or may be configured as a combination filter by connecting one low pass filter and another high pass filter in series. .

  Reference numeral (7) denotes a difference filter, which divides the input signal in terms of time, and outputs a numerical value obtained by subtracting the value of (t-1) time from the value of t time as a signal at t time. Sampling by a chopper is made into a discrete (digital) signal in time, and can be considered separately for the previous data and the next data for each time interval. Taking the difference means differentiating the time series data. It is also possible to average between discrete times by passing through the hold circuit.

Reference numeral (8) denotes an absolute value processing unit that rectifies a signal having polarity that has passed through the difference filter (7) and converts it into a nonpolar signal.
Reference numeral (9) denotes a smoothing processing unit that performs processing to smooth the nonpolar signal and to reduce the intensity fluctuation pitch (time interval) of the nonpolar signal.
Reference numeral (10) denotes a display, which includes a display element using an LED, an LCD, or the like. In the display device (10), for example, the signal smoothed by the smoothing processing unit (9) is divided into stepwise threshold values, each signal is output, and the number of light emission of the LED display elements is changed for each division. You can display the level.

In the present invention, it is preferable to use an acoustic microphone as a sensor, because it does not sense sound with a frequency higher than 15 kHz, so that ultrasonic noise of 30 kHz or higher generated by the machine can be cut. An ultrasonic sensor is not appropriate because it malfunctions in an engine room with many ultrasonic sound sources.
Further, when the acoustic microphone can cut ultrasonic noise of 30 kHz or higher, the band pass filter (6) may be a high pass filter. For example, a 5 kHz high-pass filter is the same as a 5 kHz to 15 kHz band-pass filter.
It is thought that it is sufficient to extract the sound of the frequency in this region as it is, but the sound of the air leak is basically a high frequency that is a pulse waveform, so the higher the frequency in the band (band), the more the original sound Have close information.
Therefore, in order to improve the sensitivity as the frequency increases, the difference, that is, differentiation is performed. Detection is performed by applying a function for extracting a shock wave to a filter correlation coefficient to a signal in a frequency band with low noise near an engine room or the like.

  Next, details of the difference filter (7) will be described with reference to FIG. In FIG. 2, reference numeral (7-1) denotes a switch element 1, which has a function of opening and closing signal transmission and sampling data by a sampling clock signal. Reference numeral (7-2) denotes a hold circuit 1, which has a function of holding the output level of the switch element 1 (7-1). Reference numeral (7-3) denotes an amplifier 1, which transmits the level of the hold circuit 1 (7-2) as it is. Reference numeral (7-4) denotes a switch element 2, which has a function of sampling data by opening and closing the transmission of the output of the hold circuit 1 (7-2) by a sampling clock signal. Reference numeral (7-5) denotes a hold circuit 2, which holds the output level of the switch element 2 (7-4) and transmits an inverted output. Reference numeral (7-6) denotes an amplifier 2, which outputs the sum of the output of the amplifier 1 (7-3) and the hold circuit 2 (7-5). Reference numeral (7-7) denotes a sampling clock oscillator, which transmits a clock signal for turning on / off the switch element 1 (7-1) and the switch element 2 (7-4).

  Here, the air jet is distributed from an ultrasonic wave of 40 kHz to an audible region of several kHz. Among them, vibration that can be felt is also generated from the contact portion of the operating device, a part of which is radiated as noise in the audible region outside the mechanism portion, and the frequency is distributed from 100 Hz to 5 kHz. At this time, even in the jet sound of air, the sound in the band of 20 kHz or higher and the sound in the band of 5 kHz or lower are excluded, and the signal in the band of 5 kHz or higher and 20 kHz or lower can pass.

Therefore, the output of the band pass filter (6) is input to the switch element 1 (7-1). The switch element 1 (7-1) transmits an input as an output only while the sampling clock signal is 1.
The hold circuit 1 (7-2) accumulates the voltage level of the signal transmitted by the switch element 1 (7-1), and maintains the voltage level while the switch element stops transmitting.
The amplifier 1 (7-3) is provided to transmit the output of the hold circuit 1 (7-2), but if the output impedance of the hold circuit 1 (7-2) can be sufficiently lowered, the amplifier 1 (7-2) is provided. 7-3) is not essential.
The switch element 2 (7-4) stores the signal of the hold circuit 1 (7-2) so that the hold circuit 2 (7-5) can store the signal by delaying one period of the division time. ) Operates with a sampling clock such that the switch element 2 is turned on immediately before it is turned on.
The hold circuit 2 (7-5) accumulates the voltage level of the signal transmitted by the switch element 2 (7-4), and maintains the voltage level while the switch element 2 (7-4) stops transmitting.
The output of the hold circuit 2 (7-5) is an inverted output so that a negative value of the value of the switch element 2 (7-4) is output.
A value obtained by subtracting the previous sampling information from the latest sampling information is input to the amplifier 2 (7-6), and the difference value is output.
As for the input of the amplifier 2 (7-6), two inputs may be input to the positive and negative two terminals, or an amplifier having two inputs by providing a voltage dividing circuit in front of one input terminal. Alternatively, a differential amplifier may be used. When the differential amplifier has two positive and negative input terminals, the hold circuit 2 (7-5) may be a positive output instead of an inverted output.

  As described above, the function of the difference filter (7) divides the input signal in terms of time, and uses the value obtained by subtracting the value of (t-1) time from the value of t time as the output signal at t time. is there. Obtaining a difference between two adjacent points on the signal waveform in terms of time means extracting the slope of the signal section and performing differentiation. Since the slope of the amplitude change is larger as the signal frequency is higher, if this is differentiated, a higher value is output as the frequency is higher.

The main component of the jet sound is a signal in a band around 40 kHz, and in the band up to about 1 kHz, the sound pressure gradually decreases as the frequency decreases, and the amplitude intensity of the signal decreases. When the jet flow sound is input to the difference filter (7), it passes through a high region where the signal of the jet flow is strong so that a large value is obtained, and passes weakly in a region where the signal is low.
The ratio of the input signal amplitude to the output signal amplitude of the band pass filter (6) is the same ratio regardless of the frequency of the signal in the section of the pass band, and the frequency characteristic in the pass band of the filter is flat with respect to the amplitude ( The same weight).
The signal that has passed through the band pass filter (6) is also mixed with engine noise to some extent, but by passing a large band of signals, the ratio of signal to noise can be changed so that the signal becomes stronger. .

  When the differential filter (7) is configured as described above, the corner frequency at which the frequency characteristic of the amplitude input / output ratio changes becomes half the frequency of the sampling clock signal for turning on / off the switch element. The slope of the characteristic with respect to the frequency axis was +6 dB / oct (octave) in the entire low frequency range below the corner frequency. That is, when the frequency is doubled, the output amplitude ratio is increased by 6 dB with respect to the same input amplitude.

As described above, in the frequency band where the noise is strong, the jet sound signal is passed through the band pass filter (6), and the noise mixed in the passed band is reduced by the difference filter (7). To reduce noise.
The time interval of the time division needs to be at least above the upper limit frequency of the band pass filter (6). For example, when the upper limit is 20 kHz, the sampling period for correctly transmitting the signal needs to be 2 times or more, preferably about 3 times, so the sampling clock is 60 kH or more.
In the present invention, if the on-time of the sampling clock of the divided time is long, the fluctuation output increases and does not become a correct difference. Therefore, the on-time is shortened as much as possible, or the switch element and the hold circuit are provided after the amplifier (2). It can be improved by adding more.

  Next, as shown in FIG. 3, another different configuration will be described as the difference filter (7). That is, the fact that the data is digitized and can be processed by a computer will be described.

In FIG. 3, reference numeral (7-10) denotes a microcomputer having an AD converter, a memory, a program processing unit, and a display driver circuit. The AD converter (7-11) converts an analog signal into a byte unit with a polarity sign. The memory (7-12 (a)) records the data converted by the AD converter (7-11). The program processing unit (7-13) is sequentially referred to by the internal clock by the program written in the non-volatile memory (7-12 (a)) from the outside, and is processed according to the contents.
And a display drive circuit (7-14) displays a leak level with several LED by a display.

  FIG. 4 shows a program processing example of the difference filter (7) by the microcomputer (7-10) shown in FIG. In FIG. 4, the program processing unit (7-13) by the microcomputer (7-10) is stored in the memory (7-12 (a)) and starts processing when the power is turned on.

  The AD converter (7-11) converts the input analog signal into digital information in units of polarity code bytes, and stores A memory addresses N at the head. After the storage, the conversion process is performed again, and overwriting is performed from the post-processing memory address N. This process is indicated by an arrow from the program processing unit (7-13) to the AD converter (7-11). When overwriting is completed, the contents of the status display memory (7-12 (b)) are rewritten from 0 to 1. Since the analog signal is continuous, the sampling cycle i that the AD converter (7-11) can take and the number A of data to be converted are first designated in the program processing.

In the program processing unit (7-13), when the data conversion process of the AD converter (7-11) is completed and the set number of digital data is stored in the memory (7-12 (a)), the state is changed. It is checked that the display signal changes from 0 to 1, and AD converter processing completion confirmation (7-131) is performed.
As for data stored in the AD converter (7-11), data D1 is stored at the memory start address N, and A data is stored up to the final data DA. In order for the calculation program to perform sequential calculation processing, there is a variable called a pointer that moves the address one by one. In the initial state, the pointer n is n = N, indicating that the data at address N is used for the first calculation. Yes.

  A filter processed by software is called a digital filter, and an inner product of a coefficient sequence of the filter and a data sequence is taken. The coefficient sequence is stored from the start address M. It is assumed that there are a pieces of data (a is an even number) in the coefficient sequence (K1, K2, K3... Ka) of the filter. Usually, the length of the filter is represented by K, and K = a.

  When the status display becomes 1, the pointers of the memory addresses N, M, n, m used in the program processing are returned to the initial values, and the calculation for clearing the contents of the L address is initialized (7-132). Here, the first pointer is n = N and m = M.

  The initialization is completed and calculation (7-133) is entered. First, the data D1 stored in the memory address N indicated by the pointer n and the coefficient K1 stored in the memory address M indicated by the pointer m are extracted. Next, D1 is multiplied by K1, and the solution is added to the data L at the address of the pointer L (initially 0). The contents of the pointer n are updated to N + 1, and the contents of m are updated to M + 1 (7-134). By this processing, the next reference data becomes D2 at address N + 1, and the coefficient becomes K2 at address M + 1. It is checked whether the pointer n has reached n = a and the end of one cycle is confirmed (7-135).

The numerical value a is the length of the digital filter K sequence. If a = 2, the coefficients are K1 (+1) and K2 (−1). If a = 4, for example, the coefficients are K1 (+0.3), K2 (+0.7), K3 (−0. 7) A number sequence of K4 (−0.3) is stored.
When the frequency characteristic when a = 2, that is, when the coefficients are K1 (+1) and K2 (−1) is observed as a change in the amplitude ratio, the corner frequency becomes half the sampling frequency, and the slope of the frequency characteristic is the corner. It becomes +6 dB / oct in a frequency region lower than the frequency.
The reference book “Basics and Applications of Operational Amplifiers, Hideo Tsunoda, Tokyo Denki University Press” on page 62, Fig. 3-7 Basic board diagrams (a) of various CR circuits show examples of differentiating circuits, with a slope of 20 dB / decade. It is written. One decade means a power of 10 that is 10 times one decimal digit, while the above oct (octave) of / oct means twice. The power of 2 is almost equal to 10, and the difference of 20 dB when the slope of the frequency characteristic is 10 times on the log (logarithm) axis is converted to 2 times, and 20 dB is divided by 3.3. This is expressed as 6 dB / oct. Even in a differentiation circuit using an ideal C (capacitor) R (resistance) in an electronic circuit or a differentiation circuit based on the difference as described above, the effect of one differentiation appears as 6 dB / oct in terms of the slope of the frequency characteristic.
When the coefficient of a = 4 is K1 (+0.33), K2 (+0.67), K3 (−0.67), K4 (−0.33), the corner frequency is 1/4 of the sampling frequency, The slope of the frequency characteristic is +6 dB / oct below the corner frequency, and −6 dB / oct above the corner frequency.

If the pointer is less than m = M + a in the check of the end confirmation of one cycle (7-135), the result is returned to (7-133), and the product of the data indicated by the updated pointer and the coefficient is obtained to obtain the pointer L It is added to the data and updated. If the pointer m is m = M + a and the result is Yes, the product-sum processing of the data starting from D1 and the coefficient K has completed one cycle for the digital filter K, and the processing proceeds to the next.
After the sum of products of all the filter sequences and the data starting with the data D1 is executed and the calculation of one cycle is completed, the address is initialized (7-136) and the process proceeds to update (7-137). After the initialization of calculation (7-132), the calculation of one new period starting from the data D2 following the data D1 is performed next.
The pointer of the coefficient K of the digital filter returns to the initial value and is used repeatedly. Each time the calculation for one period is repeated, the pointer N is incremented by one, and it is determined whether N = A is completed in the confirmation of the end of one period (7-138).
Here, A is the number of data converted and stored by the AD converter. That is, the sum of products with the coefficient K from D1 to DA and the total addition is stored in the pointer L as data L.
In a general digital filter process, the pointer L is updated to be +1 at the same time as the pointer N, and finally A data from the pointer L1 to the pointer LA is stored. However, if only the level display is performed as in the object of the present invention, the average value of the data from the pointer L1 to the pointer LA is often taken. Result.

  In the case of No, the process returns to the initialization of calculation (7-132) and continues. When Yes, the process proceeds to the display unit (7-139) and the processing procedure is repeated (7-140). Here, a coefficient b is prepared so that the maximum value that can be taken by the data L of the pointer L is matched with the maximum rate at which the bit of the display unit is displayed, and a solution obtained by multiplying the data L and the coefficient b is displayed on the display unit drive circuit. Write to the memory in (7-14).

  When the solution is written in the display driver circuit (7-14), the microcomputer (7-10) drives the external display (10) to light up the number of LEDs corresponding to the data level. In addition, when the data L is converted into a logarithm (LogL) and normalized by a normalization coefficient matched to the logarithm, the LED is turned on when the data L is at a background noise level near 0. Not easy to see.

  Further, when detecting an abnormal sound of an automobile engine, it is necessary to effectively remove noise (for example, mechanical operating sound) generated from the engine room of the automobile. In the engine room of an automobile, mechanical vibrations that can be experienced are generated in rotating parts such as camshaft rotation or mission system rotation, timing belts and fan belts driven by shaft gears, etc. Vibration and mechanical noise are generated. However, since vibrations with a frequency lower than several tens of Hz are difficult to propagate in the air, the noise level is not particularly problematic because of low noise level. Therefore, in the present invention, the frequency sound is narrowed down to a frequency band of 5 kHz to 20 kHz.

Further, when the band pass is a low-pass type, the change in dB linearly decreases at a low frequency with the corner frequency of the characteristic filter as a boundary. This inclination is expressed as GdB / oct. Oct (octave) is the ratio that doubles the frequency used in the musical scale. The sign of dB in the case of representing the characteristic is represented by a gradient toward the direction of increasing the frequency, and becomes a sign of + in the high-pass type and-in the low-pass type.
In general, although the frequency characteristic of a filter having a band pass characteristic is represented, the input / output ratio of the amplitude is reduced by 3 dB (decibel: −3 dB corresponds to a 30% decrease in comparison with a linear scale). A frequency at a point is called a corner frequency (also called a pole frequency) as a characteristic switching frequency (see page 60 of the aforementioned reference book).
If this corner frequency is set to 7 kHz and the slope is set to +12 dB / oct or more, frequencies lower than this can be blocked. As described above, since the gradient of one derivative is +6 dB, a gradient of +12 dB / oct can be realized by adding a differential circuit or a digital filter process that is differentiated twice. For example, the noise of 3.5 kHz from the corner frequency is attenuated by -12 dB, that is, 75%, and the noise of 1 kHz, which is lower, can be removed by 98% at -33 dB.

Next, noise removal in the ultrasonic region will be described. In contact portions that transmit kinetic energy, such as gears and belts, there is friction due to small collisions and sliding, and ultrasonic waves are emitted in a frequency band of about 40 kHz or more due to vibration of such portions. In order to remove ultrasonic noise in this band, the corner frequency of the low-pass characteristic is set to 20 kHz.
If there is a slope of 12 dB as in the above case, the 40 kHz noise on 1 oct will be -12 dB, and if m is the attenuation, then -12 = 20 * log ₁₀ (m) and m will be -0.6 (10 = −12 / 20), which is 0.25, which is 75% attenuated relative to the origin of 10 to the 0th power.

  As described above, if the corner frequency of the band pass characteristic is set to the upper limit of 20 kHz and the lower limit of 7 kHz, noise can be considerably removed. At the same time, in the air jet sound, the sound in the band of 20 kHz or more and the sound in the band of 7 kHz or less are removed, and the signal in the band of 5 kHz or more and 20 kHz or less passes. The ratio between the input signal amplitude and the output signal amplitude of the bandpass filter (6) is said to be flat (same weight) in amplitude in the passband section.

  The signal enhancement by the difference filter (7) will be described below. The difference filter (7) is divided in time, and uses a numerical value obtained by subtracting the value of (t-1) time from the value of t time as an output signal at t time. Obtaining the difference between the two points before and after on the signal waveform time axis means that the slope of the signal section is extracted and differentiation is performed. As the signal frequency is higher, the slope of the amplitude change is larger, and if this is differentiated, the higher the signal, the higher the output.

  For example, the leakage of the control pipe in the engine room is not a jet that occurs when the inside is at a positive pressure, but a suction phenomenon that occurs at a negative pressure inside. Even in the suction phenomenon, the main component of the jet sound is a signal in the 40 kHz band, similar to the jet, and the sound pressure gradually decreases and the signal amplitude also decreases as the frequency decreases in the band up to about 1 kHz.

  When the jet sound is input to the difference filter (7), it passes through a high region where the jet sound is high so that a large value appears, and weakly passes through a low region where the signal is low. The time division time interval needs to be at least above the upper limit frequency of the bandpass filter (6).

  Here, if the gradient of the difference filter (7) to be used is too large, the high-frequency attenuation characteristic of the low-pass filter may be offset. The ultrasonic noise of 40 kHz from which 20 kHz or more has been removed restores the original sound pressure intensity. Therefore, it is preferable that the difference filter (7) has an inclination of at most +6 dB / oct or less so that the 40 kHz ultrasonic wave is not recovered. The sampling frequency is 60 kHz, for example.

If 40 kHz noise is extremely large from the beginning, the slope of the low-pass band filter may be set to −12 dB / oct or more, for example, to −18 dB / oct by a three-time integration circuit or the like.
As another method, if the sampling frequency is raised to 80 kHz and the filter coefficient is set to 4, the corner frequency becomes 20 kHz, which is a quarter of 80 kHz, as shown in FIG. In the band, it becomes +6 dB as before. At the same time, in the region higher than the corner frequency, the slope is −6 dB / oct and is added to the characteristics of other filters. Therefore, the slope of the low-pass band filter ((b) in FIG. 7) is also combined with −12 dB / oct. As shown in Fig. 7 (c), it becomes -18 dB / oct, and a synergistic effect can be obtained.

  Although the engine noise is mixed to some extent with the signal that has passed through the band pass filter (6), the ratio of the signal and noise can be changed so that the signal becomes stronger by passing a large band of the signal. That is, in the frequency band where the noise is strong, the band-pass filter (6) passes the jet sound signal to reduce the noise, and the noise entering the passed band passes the signal by the difference filter (7) to reduce the noise. Let

  A signal that vibrates positively or negatively is converted into a unipolar (eg, positive) signal by a rectifier circuit. The rectified signal is smoothed at intervals of about 0.1 seconds at a time interval, and the change is reduced to a speed at which human eyes can confirm the change. The level of the smoothed signal is divided into stages, and the number of light emission of the LED is changed according to the classification.

Next, an abnormal sound detector according to another embodiment will be described with reference to FIGS. The abnormal sound detector according to the present embodiment has a band pass filter switching unit (11), and is different in that the configuration of the band pass filter that receives a signal from the band pass filter switching unit (11) is different.
5 and 6, the band pass filter switching unit (11) blocks the pass signal of the band pass filter (6) from a frequency band lower than the set frequency and a frequency band higher than another set frequency. To play a role.

  FIG. 6 shows the operation of the band pass filter switching unit (11), and controls the signal to pass through the band pass filter (6) by the signal from the band pass filter switching unit (11). In FIG. 6, the high-pass filter (61) passes a signal in a region higher than the corner frequency, and the low-pass filter (62) passes a signal in a region lower than the corner frequency. The switch element 1 (63) shuts off the low-pass filter (62) when the on / off signal of the amplifier 1 (112) is 0, and the low-pass filter (62) when the on / off signal is 1. Pass through. The switch element 2 (64) cuts off the output of the high pass filter (61) when the on / off signal of the amplifier 2 (113) is 0, and connects the output when the on / off signal is 1. To serve as a bandpass filter (6).

That is, in FIG.5 and FIG.6, the abnormal sound detector of this invention acts as follows. The band pass filter (6) is composed of a high-pass filter (61) and a low-pass filter (62), and the switch element 1 (63) is separately input with the output of the low-pass filter (62). The switch element 2 (64) can be blocked by an off signal, and the switch element 2 (64) can block the output of the high-pass filter (61) by a separately input on / off signal. The switch element 1 (63) and the switch element 2 The output of (64) is combined and one output of the switch (111) for selecting on / off of ultrasonic detection is connected to the amplifier 1 (112), and the amplifier 1 sends an on / off signal to the switch element 1 (63). The 0 output of the switch (111) for selecting the on / off of ultrasonic detection is connected to the amplifier 2 (113), and the amplifier 2 is turned on / off to the switch element 2 (64). The switch element 1 (63) and the switch element 2 (64) are turned on and off so that the on / off signals have opposite polarities, and only one of the switch element 1 (63) and the switch element 2 (64) is sent. The low-pass filter (62) is turned on and off.
When the switch (111) is operated to select the ultrasonic detection, the output of the switch element 1 (63) is cut off and the switch element 2 (64) is turned on to turn on the high-pass filter (61). The output is passed through to the final stage, the function of blocking the high frequency band signal by the low pass filter (62) is disabled, and the ultrasonic signal suitable for the high frequency range of the signal can be used. ing.

  The abnormal sound detector of the present embodiment can be effectively used in the following cases. For example, an acoustic signal (abnormal noise) related to leakage from the leaking point (2) of the pipe (1) is detected by the acoustic sensor (3), and the acoustic signal amplified by the amplifier (5) is passed through the bandpass filter (6). In doing so, it is switched by the band pass filter switching (11), and the low-pass filter (62) and the high-pass filter (61) are independently passed through the low-pass acoustic signal and the high-pass acoustic signal, This can be mixed to obtain the same acoustic signal as when the acoustic signal is passed through the bandpass filter (6).

A case where noise in the ultrasonic region in the engine room is removed using the abnormal sound detector of the present embodiment will be described below.
There is a small collision or friction at the contact part where gears, belts, etc. transmit kinetic energy, and ultrasonic waves of 40 kHz or higher are emitted based on the vibration of such part. In order to eliminate this, the corner frequency of the low-frequency over-character is set to 20 kHz. Similar to the above, if there is a 12 dB slope, the 40 kHz noise over 1 oct is attenuated by -12 dB, or 75%.

  As described above, if the corner frequency of the band pass characteristic is set to the upper limit of 20 kHz and the lower limit of 7 kHz, noise can be considerably removed. At the same time, in the jet sound of air, the sound in the band of 20 kHz or more and the sound in the band of 7 kHz or less are excluded, and the signal in the band of 5 kHz to 20 kHz can pass.

  The abnormal sound detector according to the present embodiment removes a large amount of ultrasonic noise mixed by the low-pass filter (62) when the microphone searches the entire engine room by selecting and switching the switch, and the microphone becomes a leak point. When the microphone is close to the position where the sound source of the ultrasonic noise is not visible and the noise is reduced, the low-pass filter (62) is disabled and the ultrasonic signal is also adjusted. Can be used for detection.

  In addition, a separate high-pass filter is provided so that the corner frequency can be set to 20 kHz so that only ultrasonic waves higher than that can pass, and the switch element 2 (64) is connected to this output. The output of the pass filter can be used, and in another case, switching can be performed so that an ultrasonic band signal can be used.

The abnormal noise detector of the present invention detects the presence or absence of air leaking jet noise in a control pipe resulting from the operation of the engine under the noise of a mechanical part in the engine room that occurs in a wide frequency band from an audible frequency to an ultrasonic wave. The presence or absence of defects can be determined.
Furthermore, it can be applied to measurement of leaks such as power plant pressure vessels, plant piping, or automobile engine room control piping.
In addition, because of non-contact work, the occurrence of injuries and accidents can be reduced, and areas other than the audible area (area that cannot be heard by human ears) can be detected.
Furthermore, it is possible to detect a jet of air, gas or liquid from pinholes (fine holes) or cracks (cracks) such as a pressure vessel of a power plant and piping of a plant.
Therefore, the abnormal sound detector of the present invention has advantages such as shortening the maintenance time at the maintenance shop and reducing the cost of inspection and maintenance.

It is a schematic explanatory drawing about one Embodiment of an abnormal sound detector. It is explanatory drawing about the detail of a difference filter. It is explanatory drawing about the difference filter of another different structure. It is explanatory drawing which shows the example of a program process of the difference filter by the microcomputer shown in FIG. It is a schematic explanatory drawing about other embodiment of an abnormal sound detector. It is explanatory drawing which shows the effect | action of a band pass filter switching part. It is explanatory drawing explaining the effect | action of a difference filter.

Explanation of symbols

A: Abnormal sound detector 1: Piping 2: Leakage location (abnormal sound source)
3: Acoustic sensor 4: Sensor power supply 5: Amplifier 6: Band pass filter 7: Difference filter 7-1: Switch element 1
7-2: Hold circuit 1
7-3: Amplifier 1
7-4: Switch element 2
7-5: Hold circuit 2
7-6: Amplifier 2
7-7: Sampling clock transmitter 7-10: Microcomputer 7-11: AD converter 7-12 (a): Memory 7-12 (b): Status display memory 7-13: Program processing unit 7-14: Display Unit drive circuit 7-131: Confirms the end of AD converter processing.
7-132: Initialization 7-133: Calculation 7-134: Update 7-135: Period end confirmation 7-136: Address initialization 7-137: Update 7-138: Period end confirmation 7-139: Display unit 7-140: Repeat processing procedure 8: Absolute value processing unit 9: Smoothing processing unit 10: Display unit 11: Band pass filter switching unit 61: High pass filter 62: Low pass filter 63: Switch element 1
64: Switch element 2
111: Switch for selecting on / off of ultrasonic detection 112: Amplifier 1
113: Amplifier 2

Claims (5)

A detector for detecting abnormal noise under noise,
An acoustic sensor for detecting the abnormal sound;
A band-pass filter that inputs a signal from the acoustic sensor and passes only a preset frequency band; and
A differential filter that divides the input signal from the band-pass filter in terms of time and outputs a numerical value obtained by subtracting a value of (t−1) time from a value of t time as a signal of t time. An abnormal sound detector. 2. The abnormal sound according to claim 1, wherein the signal processing in the differential filter is performed by sampling an output signal from the band pass filter by a switching element and taking a difference between two adjacent points on the time axis. Detector. 3. The abnormal sound detector according to claim 1, wherein the signal processing in the differential filter is a process of multiplying and adding the filter coefficient sequence to the band-pass filter output data signal-processed by the AD converter. The abnormal sound detector according to claim 1, wherein the band-pass filter selects a high-pass filter and a low-pass filter by switching a switch element. 5. The noise according to claim 1, wherein the abnormal noise is generated from a jet of gas or liquid due to pinholes or cracks in piping of an automobile engine room, a power plant pressure vessel, a plant or the like. The abnormal sound detector according to Crab.

Images