JP2022154180A - Inspection method and program - Google Patents

Abstract

An object of the present invention is to detect abnormal noise with high accuracy even when the noise generated from an object to be inspected is small in magnitude.
An inspection method according to an embodiment includes, for example, an acquisition step of acquiring waveform data obtained by detecting an operating sound or vibration of an inspection object, and performing a short-time Fourier transform on the waveform data. A generation step of generating a complex spectrogram, a first calculation step of calculating a predetermined phase feature amount based on the complex number spectrogram, and a second calculation step of calculating a magnitude of an abnormal component based on the phase feature amount. and including.
[Selection drawing] Fig. 2

Description

Applied for application of Article 30, Paragraph 2 of the Patent Act Proceedings of the 2021 Spring Conference of the Acoustical Society of Japan, publication date: March 12, 2021

The present invention relates to an inspection method and program.

2. Description of the Related Art Conventionally, there is a technique of acquiring waveform data of operation sound or vibration of an inspection object having an operating part such as a motor, and inspecting whether or not the inspection object has an abnormality based on the waveform data. In that case, for example, an index called a crest factor, a frequency analysis result by FFT (Fast Fourier Transform), or the like is used for inspection. According to these techniques, for example, when a large abnormal noise is generated from an object to be inspected, the abnormal noise can be detected and it can be determined that the object to be inspected has an abnormality.

JP-A-2002-71447 Japanese Patent Application Laid-Open No. 2003-214943

However, the conventional technique described above has a problem that it is difficult to detect abnormal noise when the noise generated from the object to be inspected is small in magnitude.

Therefore, one of the objects of the present invention is to realize an inspection method and program capable of detecting abnormal noise with high accuracy even when the noise generated from an object to be inspected is small in magnitude. .

The inspection method of the embodiment includes, for example, an acquisition step of acquiring waveform data obtained by detecting operating sound or vibration of an inspection object, and performing a short-time Fourier transform on the waveform data to generate a complex spectrogram. a first calculation step of calculating a predetermined phase feature amount based on the complex number spectrogram; and a second calculation step of calculating the magnitude of an abnormal component based on the phase feature amount. .
With such a configuration, for example, by using the magnitude of the abnormal component calculated from the above-described phase feature amount, even if the magnitude of the abnormal noise generated from the operating portion of the inspection object is small, the abnormal noise can be detected with high accuracy. can be detected.

Further, the inspection method further includes, for example, between the acquiring step and the generating step, a band limiting processing step of performing band limiting processing as preprocessing of the short-time Fourier transform on the waveform data. .
With such a configuration, for example, by performing band limiting processing as preprocessing for short-time Fourier transform on waveform data, the detection accuracy of abnormal noise can be further improved.

Further, in the inspection method, for example, the first calculation step calculates the phase feature amount based on the real part and the imaginary part in the complex spectrogram, and the second calculation step calculates the phase feature amount After the group delay is calculated by differentiating in the frequency direction, the group delay is subjected to smoothing filter processing, and the magnitude of the abnormal component is calculated based on the group delay subjected to the smoothing filter processing.
With such a configuration, for example, specifically, after calculating the group delay by differentiating the phase feature amount in the frequency direction, smoothing filter processing is performed on the group delay, and the group delay that has undergone the smoothing filter processing is High-precision abnormal sound detection can be realized by a method of calculating the magnitude of the abnormal component based on the above.

Further, the inspection method further includes a determination step of determining that the inspection object has an abnormality when the magnitude of the abnormal component is larger than a first threshold value.
With such a configuration, for example, it is possible to determine the abnormality of the product, which is the inspection object, at a predetermined timing such as at the time of shipment from the factory.

Further, in the inspection method, for example, the first calculation step calculates the phase feature amount by converting the phase of the complex spectrogram based on the instantaneous frequency, and the second calculation step calculates the phase The magnitude of the abnormal component is calculated by differentiating the feature amount in the time direction.
With such a configuration, for example, specifically, the phase feature amount is calculated by converting the phase of the complex number spectrogram based on the instantaneous frequency, and the phase feature amount is differentiated in the time direction to calculate the magnitude of the abnormal component. By the method of calculating , highly accurate abnormal noise detection can be realized.

Further, in the inspection method, for example, the second calculation step performs a correction process of dividing the power spectrum at a predetermined time by the power spectrum maximum value after performing the differential operation.
With such a configuration, for example, by performing the above-described correction processing, the magnitude of the abnormal component corresponding to the abnormal noise becomes conspicuous, and the detection accuracy of the abnormal noise can be further improved.

Further, the inspection method further includes, for example, a determination step of determining that the inspection object has an abnormality when the magnitude of the abnormal component is greater than a second threshold.
With such a configuration, for example, it is possible to determine the abnormality of the product, which is the inspection object, at a predetermined timing such as at the time of shipment from the factory.

Further, the program of the embodiment provides a computer with an acquisition step of acquiring waveform data obtained by detecting an operating sound or vibration of an inspection object, and performing a short-time Fourier transform on the waveform data to obtain a complex spectrogram , a first calculation step of calculating a predetermined phase feature based on the complex spectrogram, a second calculation step of calculating the magnitude of an abnormal component based on the phase feature, It is a program for executing
With such a configuration, for example, a computer can be caused to execute each of the steps described above.

FIG. 1 is an overall configuration diagram of an inspection system according to the first embodiment. FIG. 2 is a flow chart showing processing by the information processing apparatus of the first embodiment. FIG. 3 is a flow chart showing details of the process of step S7 in FIG. FIG. 4 is a flow chart showing details of the process of step S8 in FIG. FIG. 5 is a graph showing how the intensity level of the abnormal noise component changes over time in the first embodiment. FIG. 6 is a flowchart showing processing by the information processing apparatus of the second embodiment. FIG. 7 is a graph showing temporal changes in (a) the phase feature amount, (b) the ipctv feature amount, and (c) the ipctv feature amount after power spectrum correction in the second embodiment.

Exemplary embodiments (first embodiment, second embodiment) of the present invention are disclosed below. The configurations of the embodiments shown below and the actions and effects brought about by the configurations are examples. The present invention can be realized by configurations other than those disclosed in the following embodiments. Moreover, according to the present invention, at least one of various effects (including derivative effects) obtained by the configuration can be obtained.

(First embodiment)
First, an example of an object to be inspected and a conventional inspection method will be described. An inspection object is, for example, an automobile part having an operating part such as a motor. Some automobile parts generate a breakthrough noise with a very short duration (short-pulse noise) when actuated. Since this abnormal noise is difficult to quantify in the production line, it has been detected, for example, based on auditory sense of a sensory inspector.

In addition, it is also possible to inspect using an index called a crest factor, a frequency analysis result obtained by FFT (Fast Fourier Transform), or the like. However, these prior arts have a problem that it is difficult to detect abnormal noise when the noise generated from the object to be inspected is small in magnitude.

In the following, therefore, a technique will be described that can detect abnormal noise with high accuracy without relying on a sensory inspector, even if the abnormal noise generated from an object to be inspected is small in magnitude.

FIG. 1 is an overall configuration diagram of an inspection system S according to the first embodiment. The inspection system S includes an inspection device 1 , a sound pickup device 2 and a control device 3 .

The inspection apparatus 1 is a soundproof facility (such as a soundproof room) that shields an object to be inspected 4 and a microphone 5 housed therein from external sounds. The microphone 5 detects sound generated from the inspection object 4 placed nearby and outputs waveform data.

The control device 3 is connected to the inspection object 4 and controls the operation of the actuated portion of the inspection object 4 . The control device 3 is, for example, a PLC (Programmable Logic Controller).

The sound collection device 2 includes a data logger 6 and an information processing device 7 . The data logger 6 temporarily stores waveform data input from the microphone 5 .

The information processing device 7 is, for example, a computer device, and includes a storage section 71 , an input section 72 , an output section 73 , a communication section 74 and a processing section 75 .

The storage unit 71 includes various storage devices such as RAM (Random Access Memory), ROM (Read Only Memory), HDD (Hard Disk Drive), and SSD (Solid State Drive), and stores various programs and data. do.

The input unit 72 is means for a user to input information, and is, for example, a keyboard, mouse, touch panel, or the like.

The output unit 73 is means for outputting information, and is, for example, a display device or an audio output device. The display device is, for example, an LCD (Liquid Crystal Display) or the like. The audio output device is, for example, a speaker.

The communication unit 74 is a communication interface for communicating with external devices (control device 3, data logger 6, etc.).

The processing unit 75 is configured by, for example, a CPU (Central Processing Unit), and implements each functional unit by reading and executing a program stored in the storage unit 71 . The processing unit 75 includes an acquisition unit 751, a preprocessing unit 752, a generation unit 753, a calculation unit 754, a filtering unit 755, a determination unit 756, and a control unit 757 as functional units. Part or all of each functional unit may be configured by hardware such as a circuit including an ASIC (Application Specific Integrated Circuit).

Acquisition unit 751 acquires various types of information from an external device (control device 3, data logger 6, etc.). The acquisition unit 751 acquires waveform data related to the operation sound of the inspection object 4 from the data logger 6, for example. In this embodiment, the "operating sound" may be a normal operating sound alone or a combination of a normal operating sound and an abnormal sound.

The preprocessing unit 752 performs band limiting processing and the like on the waveform data as preprocessing for short-time Fourier transform.

The generation unit 753 performs a short-time Fourier transform on the preprocessed waveform data to generate a complex spectrogram (details will be described later).

The calculation unit 754 calculates a first calculation step of calculating a predetermined phase feature amount based on the complex number spectrogram, and calculates the intensity level of the abnormal noise component (an example of the magnitude of the abnormal component) based on the phase feature amount. A second calculation step is performed.

As a first calculation step, the calculation unit 754 calculates the phase feature amount based on, for example, the real part and the imaginary part in the complex spectrogram (details will be described later).

Further, after the first calculation step, the calculation unit 754 performs a third calculation step of calculating group delay by differentiating the phase feature amount in the frequency direction (details will be described later). In this case, the filter processing unit 755 performs a filter processing step of performing smoothing filter processing on the group delay in order to make the abnormal sound component stand out (details will be described later). After that, as a second calculation step, the calculator 754 calculates the intensity level of the abnormal sound component based on the group delay after the smoothing filter processing (details will be described later).

The determination unit 756 determines that there is an abnormality in the inspection object 4 when the intensity level of the abnormal sound component is greater than the first threshold (details will be described later).

The control unit 757 executes processing other than the processing executed by each unit 751-756. For example, the control unit 757 controls the control device 3 to operate the inspection object 4 when inspecting the inspection object 4 .

FIG. 2 is a flow chart showing processing by the information processing device 7 of the first embodiment. First, in step S<b>1 , the acquisition unit 751 acquires waveform data related to the operation sound of the inspection object 4 from the data logger 6 .

Next, in step S2, the preprocessing unit 752 performs band limiting processing and the like on the waveform data as preprocessing for short-time Fourier transform.

Next, in step S3, the generation unit 753 performs a short-time Fourier transform on the preprocessed waveform data to generate a complex spectrogram.

Next, in step S4, the calculator 754 calculates a predetermined phase feature amount based on the real part and the imaginary part of the complex spectrogram. A phase feature amount can be expressed as Φ(k, l), for example. k is the frequency index number. l is the time index number. More specifically, the phase feature amount is, for example, the angle between the positive direction of the real axis and the line connecting the origin and the point corresponding to the real part and the imaginary part in the complex number spectrogram on the complex number plane.

Next, in step S5, the calculation unit 754 calculates a group delay by differentiating the phase feature quantity in the frequency direction. The group delay τ(k,l) is, for example, given by the following equation (1). Note that ∇ _k (•) represents a differential operator in the frequency direction.
τ(k,l)=cos(-∇kΦ( _k ,l)) Equation (1)

Next, in step S6, the filter processor 755 performs smoothing filter processing on the group delay. The group delay τ' after smoothing filter processing is given by the following equation (2).
τ′＝Gaussian(τ) Equation (2)

Next, in step S7, the calculator 754 calculates the intensity level of the abnormal noise component based on the group delay τ' after smoothing filter processing.

Here, FIG. 3 is a flow chart showing the details of the process of step S7 in FIG. First, in step S71, the calculator 754 inputs the group delay τ′.

Next, in step S72, the calculator 754 initializes the frequency index i and the time index index j to zero. Also, the calculation unit 754 prepares (declares) F, which is a one-dimensional array having a length of the time index number l.

Next, in step S73, the calculation unit 754 initializes i and the number count parameter count to 0 as processing after No in step S79.

Next, in step S74, the calculation unit 754 determines whether or not τ'(i, j) is greater than the threshold value α. If Yes, the process proceeds to step S75, and if No, the process proceeds to step S76. The threshold α is set in advance.

In step S75, the calculator 754 increments count (by 1).

In step S76, the calculator 754 increments i (by 1).

Next, in step S77, the calculation unit 754 determines whether or not i has reached k. If Yes, the process proceeds to step S78, and if No, the process returns to step S74.

In step S78, the calculation unit 754 substitutes the value of count for F[j] and increments j (by 1).

Next, in step S79, the calculation unit 754 determines whether or not j has reached l. Thus, a one-dimensional array F can be created.

Returning to FIG. 2, in step S8 after step S7, the determination unit 756 determines that the inspection object 4 has an abnormality when the intensity level of the abnormal noise component is greater than the first threshold. Here, FIG. 4 is a flow chart showing the details of the process of step S8 in FIG.

In step S81, the determination unit 756 receives a one-dimensional array F as input.

Next, in step S82, the determination unit 756 initializes j to 0.

Next, in step S83, the determination unit 756 determines whether or not F[j] is greater than β (first threshold value). If Yes, proceed to step S87. move on.

In step S87, the determination unit 756 determines that there is an abnormality, and terminates the process of step S8. In other words, if even one value of F[j] exceeds the threshold value β, it is determined to be abnormal.

In step S84, the determination unit 756 increments j (by 1).

Next, in step S85, the determination unit 756 determines whether or not j has reached l. If Yes, the process proceeds to step S86, and if No, the process returns to step S83.

In step S86, the determination unit 756 determines that the process is normal, and terminates the process of step S8. That is, if all the values of F[j] are equal to or less than the threshold value β, it is determined to be normal.

Next, FIG. 5 is a graph showing how the intensity level of the abnormal noise component changes over time in the first embodiment. When the inspection object 4 generates only normal operation sounds, the intensity level changes within the range of 0.1 or less. Then, when the inspection object 4 generates an abnormal noise in addition to the normal operating noise (when the time is about 3.6), the intensity level rises to about 0.4. Therefore, if the threshold value is set to, for example, about 0.2, it is possible to detect only abnormal noise without judging normal operating noise as abnormal. Further, since the intensity level of the abnormal noise component is calculated based on the above-described phase feature amount instead of the amplitude of the waveform data, even if the magnitude of the abnormal noise is small, it can be detected with high accuracy.

As described above, according to the inspection system S of the first embodiment, by using the intensity level of the abnormal noise component calculated from the phase feature amount described above, when the magnitude of the abnormal noise generated from the inspection object 4 is small, However, the abnormal noise can be detected with high accuracy.

Details are as follows. A normal operating sound generated from the inspection object 4 is considered to have a stable rhythm. Further, the waveform data relating to the operation sound is continuous, but when the short-time Fourier transform is performed by the information processing device 7, the waveform data is handled discretely in the time direction using a discrete time signal. Then, when the above-described processing is performed on the waveform data, the above-described intensity level of the normal operation noise is lowered, and the above-described intensity level of the abnormal noise is raised. Therefore, even if the noise generated from the inspection object 4 is small, the noise can be detected with high accuracy.

In addition, by performing band limiting processing as preprocessing for the short-time Fourier transform on the waveform data, it is possible to further improve the detection accuracy of abnormal sounds.

Specifically, after calculating the group delay by differentiating the phase feature amount in the frequency direction, smoothing filter processing is performed on the group delay, and the magnitude of the abnormal component is calculated based on the group delay that has undergone the smoothing filter processing. High-precision abnormal noise detection can be realized by the method of calculating the noise.

Then, for example, at a predetermined timing such as when the product is shipped from the factory, it is possible to determine whether the product, which is the inspection object 4, is abnormal.

In addition to small abnormal noises, abnormal noises buried in variations in the amplitude distribution of normal operating sounds of the inspection object 4, and frequency analysis of normal operating sounds of the inspection object 4 (of each frequency of FFT Power component) can be detected with high accuracy even if it is an abnormal noise buried in the distribution variation.

In particular, in recent years, the electrification of vehicles has progressed, and the interior of the vehicle has become quieter. useful for

(Second embodiment)
Next, a second embodiment will be described. Duplicate descriptions of items similar to those of the first embodiment will be omitted as appropriate. Compared to the first embodiment, in the processing unit 75 of the information processing device 7 in the configuration of FIG. 1, the filter processing unit 755 is not used, and the processing contents of the calculation unit 754 and the determination unit 756 are different.

The calculation unit 754 performs a first calculation step of calculating a predetermined phase feature amount based on the complex spectrogram generated by the generation unit 753, and an abnormal noise component intensity level (abnormal component magnitude) based on the phase feature amount. and a second calculation step of calculating (an example of).

As a first calculation step, the calculator 754 calculates the phase feature amount by transforming the phase of the complex spectrogram based on the instantaneous frequency. In addition, as a second calculation step, the calculation unit 754 calculates the intensity level of the abnormal sound component by differentiating the phase feature quantity in the time direction. Further, as a second calculation step, the calculation unit 754 may perform a correction process of dividing the power spectrum at a predetermined time by the maximum value of the power spectrum after performing the differential operation described above.

The determination unit 756 executes a determination step of determining that there is an abnormality in the inspection object when the intensity level of the abnormal noise component is greater than the second threshold.

FIG. 6 is a flowchart showing processing by the information processing device 7 of the second embodiment. Steps S1-S3 are the same as steps S1-S3 in FIG.

After step S3, in step S11, the calculator 754 calculates the phase feature amount by transforming the phase of the complex spectrogram based on the instantaneous frequency. Here, the conversion is, for example, a process equivalent to a process of rotating coordinates on a complex number plane.

Next, in step S12, the calculation unit 754 differentiates the phase feature amount in the time direction, performs a correction process of dividing the power spectrum at a predetermined time by the power spectrum maximum value, and calculates the intensity level of the abnormal sound component. Calculate

If this correction process is not performed, for example, if the complex spectrogram is F(k, l) and the instantaneous frequency is E _ipc , the phase feature quantity v(k, l) after differential operation (hereinafter referred to as “ipctv feature quantity”) Also called.) is expressed by the following formula (3).
v(k, l)=|∇ _l E _ipc F(k, l)| Expression (3)

Further, when the above-described correction processing is performed, the phase feature amount v(k, l) after the differentiation operation (hereinafter also referred to as "ipctv feature amount after power spectrum correction") is as shown in the following equation (4): is.
v(k,l)={|F(k,l)|/(max(F(k,l))}×
|∇ _l E _ipc F(k, l)| Expression (4)

Next, in step S13, the determination unit 756 determines that the inspection object 4 has an abnormality when the intensity level of the abnormal sound component is greater than the second threshold.

Next, FIG. 7 shows how each of (a) the phase feature amount, (b) the ipctv feature amount, and (c) the ipctv feature amount after power spectrum correction in the second embodiment changes over time. graph. Here, abnormal noise occurs at the time of about 1.2 and the time of about 1.8.

In (a) the phase feature amount and (b) the ipctv feature amount, the values at the time of abnormal noise are slightly larger than the values at other times. In addition, in the (c) ipctv feature amount after power spectrum correction, the value at the time of abnormal noise is significantly larger than the value at other times. Therefore, by using the ipctv feature amount after power spectrum correction, it can be seen that the detection accuracy of abnormal noise can be particularly improved.

As described above, according to the inspection system S of the second embodiment, specifically, the phase feature amount is calculated by converting the phase of the complex number spectrogram based on the instantaneous frequency, and the phase feature amount is differentiated in the time direction. High-precision abnormal sound detection can be realized by a method of calculating the magnitude of the abnormal component by performing calculations.

Further, for example, by performing the above-described correction processing, the intensity level of the abnormal sound component corresponding to the abnormal sound becomes conspicuous, and the detection accuracy of the abnormal sound can be further improved.

Then, for example, at a predetermined timing such as when the product is shipped from the factory, it is possible to determine whether the product, which is the inspection object 4, is abnormal.

The program for executing the above process executed by the information processing device 7 is a file in an installable format or an executable format, and can be downloaded from a CD-ROM, a CD-R, a memory card, a DVD (Digital Versatile Disk), It may be stored in a computer-readable storage medium such as a flexible disk (FD) and provided as a computer program product. Alternatively, the program may be stored on a computer connected to a network such as the Internet, and provided by being downloaded via the network. Also, the program may be provided or distributed via a network such as the Internet.

Although the embodiment of the present invention has been described above, the above embodiment is presented as an example and is not intended to limit the scope of the invention. This novel embodiment can be embodied in various other forms, and various omissions, replacements, and modifications can be made without departing from the scope of the invention. This embodiment and its modifications are included in the scope and gist of the invention, and are included in the scope of the invention described in the claims and its equivalents.

For example, the inspection object 4 is not limited to automobile parts, and may be other objects as long as they have moving parts.

The data used for the inspection is not limited to the waveform data obtained by detecting the operating sound of the inspection object 4, but may be waveform data obtained by detecting the vibration of the inspection object 4. may

DESCRIPTION OF SYMBOLS 1... Inspection apparatus, 2... Sound collection apparatus, 3... Control apparatus, 4... Inspection object, 5... Microphone, 6... Data logger, 7... Information processing apparatus, 71... Storage part, 72... Input part, 73... Output Part 74... Communication unit 75... Processing unit 751... Acquisition unit 752... Preprocessing unit 753... Generation unit 754... Calculation unit 755... Filter processing unit 756... Judgment unit 757... Control unit, S …inspection system

Claims (8)

an acquiring step of acquiring waveform data obtained by detecting operating sound or vibration of the inspection object;
a generation step of performing a short-time Fourier transform on the waveform data to generate a complex spectrogram;
a first calculation step of calculating a predetermined phase feature amount based on the complex spectrogram;
and a second calculation step of calculating the magnitude of the abnormal component based on the phase feature amount. 2. The inspection method according to claim 1, further comprising, between said acquisition step and said generation step, a band limitation processing step of performing band limitation processing as preprocessing of said short-time Fourier transform on said waveform data. . The first calculating step calculates the phase feature amount based on the real part and the imaginary part in the complex spectrogram,
In the inspection method, after the first calculation step,
a third calculation step of calculating a group delay by differentiating the phase feature amount in the frequency direction;
a filtering step of smoothing filtering the group delay;
2. The inspection method according to claim 1, wherein after said filtering step, said second calculating step calculates the magnitude of said abnormal component based on said smoothed filtered group delay. 4. The inspection method according to claim 2, further comprising a determination step of determining that the inspection object has an abnormality when the magnitude of the abnormal component is greater than a first threshold. The first calculating step calculates the phase feature amount by transforming the phase of the complex spectrogram based on the instantaneous frequency,
2. The inspection method according to claim 1, wherein said second calculating step calculates the magnitude of said abnormal component by differentiating said phase feature quantity in a time direction. 6. The inspection method according to claim 5, wherein said second calculation step performs a correction process of dividing a power spectrum at a predetermined time by a power spectrum maximum value after said differential operation. 7. The inspection method according to claim 6, further comprising a determination step of determining that the inspection object has an abnormality when the magnitude of the abnormal component is greater than a second threshold. to the computer,
an acquiring step of acquiring waveform data obtained by detecting operating sound or vibration of the inspection object;
a generation step of performing a short-time Fourier transform on the waveform data to generate a complex spectrogram;
a first calculation step of calculating a predetermined phase feature amount based on the complex spectrogram;
and a second calculation step of calculating the magnitude of the abnormal component based on the phase feature amount.

Images