CN115691509A - Interference identification method suitable for abnormal sound detection of industrial equipment - Google Patents
Interference identification method suitable for abnormal sound detection of industrial equipment Download PDFInfo
- Publication number
- CN115691509A CN115691509A CN202211704949.0A CN202211704949A CN115691509A CN 115691509 A CN115691509 A CN 115691509A CN 202211704949 A CN202211704949 A CN 202211704949A CN 115691509 A CN115691509 A CN 115691509A
- Authority
- CN
- China
- Prior art keywords
- signal
- signal frame
- voiceprint
- industrial equipment
- abnormal sound
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000002159 abnormal effect Effects 0.000 title claims abstract description 46
- 238000000034 method Methods 0.000 title claims abstract description 37
- 238000001514 detection method Methods 0.000 title claims abstract description 27
- 238000000605 extraction Methods 0.000 claims abstract description 11
- 230000011218 segmentation Effects 0.000 claims abstract description 9
- 238000004422 calculation algorithm Methods 0.000 claims description 33
- 238000005070 sampling Methods 0.000 claims description 30
- 238000004364 calculation method Methods 0.000 claims description 17
- 239000013598 vector Substances 0.000 claims description 10
- 238000007621 cluster analysis Methods 0.000 claims description 9
- 238000000354 decomposition reaction Methods 0.000 claims description 9
- 230000008569 process Effects 0.000 claims description 8
- 238000004458 analytical method Methods 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 6
- 239000006185 dispersion Substances 0.000 claims 1
- 230000007613 environmental effect Effects 0.000 abstract description 6
- 230000001788 irregular Effects 0.000 abstract description 4
- 238000007781 pre-processing Methods 0.000 abstract description 2
- 238000012544 monitoring process Methods 0.000 description 5
- 238000012545 processing Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 230000002349 favourable effect Effects 0.000 description 2
- 230000036541 health Effects 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 230000005856 abnormality Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000009432 framing Methods 0.000 description 1
- 230000005923 long-lasting effect Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000000465 moulding Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
Images
Landscapes
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The present invention relates to the field of voiceprint detection, and in particular to preprocessing of voiceprint data prior to voiceprint detection. The invention is realized by the following technical scheme: an interference identification method suitable for abnormal sound detection of industrial equipment comprises the following steps: s01, a voiceprint segmentation step, namely segmenting the voiceprint signals extracted by the microphone into a plurality of second-level signal frames; s02, a signal frame characteristic phasor M extraction step, wherein the characteristic phasor M of each signal frame is extracted, and the characteristic phasor M comprises a characteristic value of the signal frame and a frequency band distribution value of the signal frame; and S03, clustering identification. The invention aims to provide an interference identification method suitable for detecting abnormal sounds of industrial equipment, which is different from the technical scheme of only eliminating environmental noise and operating sounds of the equipment, can further identify the interference sounds of the abnormal sounds, and can effectively identify accidental and irregular interference sounds, thereby further improving the purity of abnormal sound detection.
Description
Technical Field
The present invention relates to the field of voiceprint detection, and in particular to preprocessing of voiceprint data prior to voiceprint detection.
Background
Industrial equipment, such as large-scale manufacturing equipment, molding equipment, and other commercial equipment, is widely used in factories, scientific research institutions, and other places, and is an important material for their affiliated units. During the use of these industrial equipments, the health of the equipments needs to be monitored on a daily basis so as to avoid the potential safety hazard caused by the operation stop due to the failure.
Compared with video monitoring, on-site monitoring by an administrator and other modes, voiceprint monitoring is selected as a plurality of health detection means, and the method has the advantages of controllable hardware investment cost, labor saving, wide monitoring range and the like, and is one of the important means for detecting the faults of the industrial equipment at present. Chinese patent document No. CN 112513757A discloses a system for monitoring industrial equipment, in which an audio sensor is disposed near the industrial equipment for capturing a voiceprint signal. The system is also provided with a computing device for identifying the voiceprint signals, marking abnormal sounds in the voiceprint signals as abnormal sounds and comparing the abnormal sounds with fault abnormal sounds in the database.
Further, as disclosed in chinese patent No. CN 115376552A, a joint learning method for detecting an abnormality in an industrial device by using sound includes a plurality of audio sensors, acquiring a sound wave signal from the industrial device, and a convolutional neural network, recognizing the sound wave signal acquired by the audio sensors, and determining whether the state of the industrial device is abnormal.
In the process, the extraction of abnormal sounds in the sound wave signals is a key step, and subsequent abnormal conditions can be identified and learned automatically only when the abnormal sounds exist. However, in the actual use environment of industrial equipment, the environment is complex, and a lot of interference exists. This causes the purity of the extracted and recognized abnormal sound to be low, which affects the subsequent recognition and detection. In the prior art, some optimized technical schemes are provided, in the abnormal sound extraction process, identification of environmental noise and identification of operation sound of industrial equipment are added, and the sound is removed from the abnormal sound, so that the sound wave purity of the abnormal sound is increased. However, even under such an optimized design scheme, the purity of extracting the abnormal sound is still not ideal, and the result of fault identification and detection of subsequent industrial equipment is not accurate, so that potential safety hazards in the use process of the industrial equipment are caused.
Disclosure of Invention
The invention aims to provide an interference identification method suitable for detecting abnormal sounds of industrial equipment, which is different from the technical scheme of only eliminating environmental noise and operating sounds of the equipment, can further identify the interference sounds of the abnormal sounds, and can effectively identify accidental and irregular interference sounds, thereby further improving the purity of abnormal sound detection.
The invention is realized by the following technical scheme: an interference identification method suitable for abnormal sound detection of industrial equipment comprises the following steps: s01, a voiceprint segmentation step, namely segmenting a voiceprint signal extracted by a microphone into a plurality of second-level signal frames; s02, a signal frame characteristic phasor M extraction step, wherein the characteristic phasor M of each signal frame is extracted, and the characteristic phasor M comprises a characteristic value of the signal frame and a frequency band distribution value of the signal frame; the characteristic value comprises a root mean square value C RMS of a signal frame and is used for reflecting the integral energy size of the signal; the mean square error of the signal frame C MES is used for reflecting the discrete degree of the signal; a kurtosis C K of the signal frame for reflecting an impulse component of the signal; the calculation formulas of the three are respectively as follows:
wherein N is the total number of audio sampling points in a signal frame, i is the number of the audio sampling points and is a natural number, and Ci is the data content of a single audio sampling point; the frequency band distribution value of the signal frame is used for reflecting the energy distribution condition of each frequency band of the signal frame; (ii) a The feature vectors M of all signal frames form a feature matrix; and S03, a cluster identification step, namely analyzing the characteristic matrix by using a cluster analysis algorithm, judging that the voiceprint signal has no short-time interference sound if all the signal frames are clustered into a cluster finally after the analysis is finished, and judging that the voiceprint signal has the short-time interference sound if the voiceprint signal cannot be clustered into a cluster finally.
In the present invention, preferably, in the step S01, in the process of dividing the voiceprint signal into a plurality of signal frames, an overlap region is provided between adjacent frames.
Preferably, each of the signal frames has the same frame length, which is T; the length of the overlapping area is T/2.
Preferably, in S02, the feature value of the signal frame further includes an skewness feature Cs for reflecting a symmetry degree of data distribution of the signal, and the formula is as follows:
(ii) a Wherein N is the number of audio sampling points in a frame, i is the number of audio sampling points and is a natural number, X is the specific data content of the sampling points,
is a mathematical expectation of the data content of the sample points.
Preferably, in S02, a square summation algorithm of amplitude values of each frequency point is adopted as a calculation method of the frequency band distribution value of the signal frame.
Preferably, in S02, the signal frame is processedThe calculation mode of the frequency band distribution value adopts a wavelet packet energy extraction algorithm, and specifically comprises the following steps: s021, a segmentation step; one signal frame is divided into small wave packets, and single frame
After i layers of wavelet packet division, the method is obtained
A sub-band of signals; s022, obtaining energy of a sub-frequency band; obtaining the energy value of each sub-frequency band, wherein the energy E of the sub-frequency band which is divided by the i-layer wavelet packet and is numbered as k is as follows:
in the formula, i is the number of decomposition layers, k represents the number of a sub-frequency band, f is a function representing the decomposition result of the wavelet packet, N is the total number of sampling points in the sub-frequency band, and N is the number of the sampling points; s023, a step of presenting a frequency band distribution value; each sub-band energy value E is represented in said characteristic phasor M of the signal frame,
n is the number of the signal frame, and i is the number of the wavelet packet decomposition layers.
Preferably, in the step S03 of cluster recognition, the algorithm of cluster analysis is a K-Means algorithm or a density-based clustering DBSCAN algorithm or a hierarchical clustering algorithm.
Preferably, in the step of S03 and cluster recognition, the algorithm of cluster analysis is a Mean Shift algorithm.
Preferably, in the step S01, the time length of the voiceprint signal extracted by the microphone is sampled by more than one minute.
Preferably, in the step S01, the usage rate of the voiceprint signals extracted by the microphone is 11kHz, 22kHz, 44.1kHz, 48kHz or 96kHz.
In conclusion, the invention has the following beneficial effects:
1. the voiceprint data signals of the industrial equipment can be effectively identified through the processing of segmentation, characteristic phasor extraction and cluster identification, and the defect that only interference sounds with long period and good regularity can be found in the prior art is effectively overcome. The purity of the abnormal sound for eliminating the short-time interference sound is higher, a more reasonable and real data base is provided for the detection model, and the accuracy of the later-period calculation of the detection model is improved.
2. In the selection of the characteristic phasor M of the signal frame, the characteristic value of the signal frame is included, and the frequency band distribution value of the signal frame is also included, so that the accuracy of the characteristic phasor M of the signal frame is ensured.
3. The root mean square value C RMS of the signal frame, the mean square error C MES of the signal frame and the kurtosis C K are selected from the characteristic values of the signal frame, and the selection of the three characteristic values has the advantages of less calculation amount from the calculation angle, high efficiency and good instantaneity; the method is comprehensive from the aspect of calculation effectiveness, and the effect of screening out the short-time interference by taking the combination of the three characteristic values as a clustering vector is good.
4. The clustering identification of the invention adopts a mean shift algorithm, has small calculated amount and good real-time property, and is suitable for real-time tracking occasions.
5. The division of the signal frames sets an overlapping area, firstly, the continuity between the adjacent frames can be ensured, the similarity between the adjacent frames is improved, secondly, the characteristic parameters can be smoothly changed, and the data leakage can be effectively prevented again.
6. The selection of the microphone utilization rate ensures the subsequent calculation precision and avoids influencing the transmission efficiency and overhigh microphone cost.
Drawings
Fig. 1 is a schematic flow chart of embodiment 1 of the present invention.
Detailed Description
The present invention will be described in further detail below.
The present embodiment is only for explaining the present invention, and it is not limited to the present invention, and those skilled in the art can make modifications of the present embodiment without inventive contribution as needed after reading the present specification, but all of them are protected by patent law within the scope of the claims of the present invention.
In embodiment 1, an interference identification method suitable for detecting abnormal sounds of industrial equipment is provided, in which hardware is first arranged, and a microphone is arranged in the industrial equipment and is in remote communication connection with a control center through a network module. The microphone is used for collecting voiceprint data signals in the working process of the industrial equipment, and the network module is used for communicating and transmitting the voiceprint data signals to the control center. The microphone in the prior art can be directly selected and used, and hardware does not need to be changed. The network module can be selected from GPRS network modules in the prior art, and compared with network modules such as Bluetooth and the like, the communication distance of GPRS is longer, so that the control center can be arranged in a place farther away from industrial equipment. The control center can select conventional electronic equipment such as a PC, a server and the like in the prior art, and is used for storing data, installing a software program and installing a machine learning model so as to filter, identify, machine learn and the like the voiceprint signals captured by the microphone.
In the prior art, the voiceprint signal S captured by the microphone can identify the environmental noise and the device operating sound in the voiceprint signal S, and the filtered voiceprint signal S' is obtained by subtracting the environmental noise and the device operating sound from the voiceprint signal S. However, the applicant has found that the purity of the voiceprint signal S' is still not high, because the microphone in the device may still pick up noise interference, such as accidental mechanical impact sound inside the industrial device.
Therefore, at this time, the filtered voiceprint signal S 'may still include external interference unrelated to the abnormal sound, and if the voiceprint signal S' is not further analyzed, the interference is regarded as the abnormal sound to be subjected to alarm processing, so that misjudgment is caused.
The applicant has found that these disturbances are short-term phenomena and are therefore reflected in the time and frequency domains as more pronounced changes, not long lasting and not self-similar. Based on this, the invention can focus and effectively remove the short-time interference sound, thereby improving the purity of the abnormal sound.
As shown in fig. 1, the process first proceeds to S01 and a voiceprint segmentation step.
The object of voiceprint segmentation is the voiceprint signal S extracted by the microphone in the prior art, and whether the voiceprint signal S is filtered by environmental noise and equipment working sound or not can be realized without limitation in the technical scheme. In the present embodiment, the ambient noise and the device operation sound have been filtered.
The sampling time of the voiceprint signal S is not short enough, and if the sampling time is short enough, the sample data in the later period is less, which is not favorable for the calculation in the subsequent steps. Preferably on a timescale of minutes, in this example the sampling time is 1 minute. The sampling frequency of the digital signal of the microphone can be 11kHz, 22kHz, 44.1kHz, 48kHz, 96kHz and the like. The higher the sampling rate, the shorter the sampling time interval, the more sample data can be obtained in unit time, and the more accurate the representation of sound waveform and the like. In the embodiment, the utilization rate is 48kHz, which not only ensures the subsequent calculation accuracy, but also avoids affecting the transmission efficiency and the excessive microphone cost.
In this step, the voiceprint signal S is divided into a number of signal frames. The voiceprint signal S is often in a time-domain format, and the divided signal frames are naturally also in a time-domain format. The frame is a signal segment on the order of seconds, and in the present embodiment, the time length of each frame may be set to 1 second.
And then, the step S02 of extracting the feature vector of the signal frame is carried out.
In this step, a feature vector of each signal frame is calculated and extracted.
The category of the feature vector can be specified individually by designers according to actual conditions. In the present embodiment, however, the designation "3+1" is used. 3 means 3 eigenvalues and 1 means a frequency band distribution value.
Specifically, the 3 characteristic values are root mean square values of the signal frames, and are recorded as C RMS (root mean square) and used for reflecting the energy of the signals; the mean square error of the signal frame is marked as C MES and is used for reflecting the discrete degree of the signal; the kurtosis of the signal frame, denoted as C K, is used to reflect the impulse component of the signal.
The calculation formulas of the three are respectively as follows:
in the formula, N is the number of audio sampling points in one frame, and i is the number of the audio sampling points and is a natural number. Ci is the data content of a single sample point.
The calculation of the frequency band distribution value of the signal frame has various modes, and can be realized by selecting a square summation algorithm or a similar algorithm of amplitude values of various frequency points in the prior art. The specific algorithm is the content of the prior art, and is not described herein in detail.
And then, the step S03 of cluster identification is carried out.
Through the above two steps, a plurality of signal frames have appeared, each signal frame containing a complete feature vector M. The feature vectors M of all the signal frames form a feature matrix, and the feature matrix is subjected to cluster analysis.
The algorithm of the cluster analysis can adopt a K-Means algorithm or a DBSCAN algorithm (a density-based clustering algorithm) or a hierarchical clustering algorithm and the like in the prior art.
Through a cluster analysis algorithm, if the final results of all the signal frames are clustered into a cluster, it is judged that the abnormal sound does not contain the short-time interference sound, and if the abnormal sound cannot be clustered into a cluster at last but forms different clusters, it is indicated that the short-time interference sound is contained.
The technical scheme is completed, and through the steps, the method can further effectively identify accidental and irregular short-time interference sound of the industrial equipment in the using process on the basis of the prior art, so that the accidental and irregular short-time interference sound is prevented from becoming abnormal sound, and the misjudgment of the system is further influenced; the purity of abnormal sound detection is improved.
The subsequent processing may be selected by the technician, for example, the sampled sound with the short-time interference sound may be re-identified, or the removal processing may be performed. For example, the technician performs voiceprint recognition on the cluster, and then removes the interference cluster, thereby obtaining the discontinuous abnormal sound. The voiceprint recognition technology of the interference cluster and the removal of the interference cluster are applications of conventional technical means of those skilled in the art, are not technical innovation points of the invention, and are not described herein again.
Embodiment 2 differs from embodiment 1 in the detail processing in S01. In the present embodiment, an overlap region is provided between adjacent frames. For example, the frame length of the framing frame is set to T, and the overlapping area of T/2 between adjacent frames is set, so that the displacement between adjacent frames is also T/2. For example, the first frame may be 1-2 seconds, the second frame may be 1.5-2.5 seconds, and the third frame may be 2-3 seconds.
In the present embodiment, the reason for increasing the overlap region is to make the adjacent frames have continuity, so as to improve the similarity between the adjacent frames. The short-time interference sound has no similarity, and if dissimilar frames can still be found in adjacent frames with continuity, the reliability of a late-stage clustering result can be further improved. Secondly, the phenomenon that the change of the characteristic parameters is possibly large when two separated frames are exactly between two syllables is avoided. Adding overlap between adjacent frames therefore smoothes the change in the characteristic parameters. Thirdly, in order to further improve the data integrity, when there is an overlap between adjacent frames, the possibility of data leakage loss of the adjacent frames in the subsequent analysis is further reduced.
Embodiment 3 is different from embodiment 1 in that in step S02, a wavelet packet energy extraction algorithm is used for the calculation of the energy value of the signal frame.
Wavelet packet analysis is an extension of wavelet analysis, and its basic idea is to centralize information energy, divide frequency bands into multiple layers, further decompose high-frequency parts, and adaptively select corresponding frequency bands according to the characteristics of the analyzed signals so as to match the frequency spectrums of the signals. This has the effect that the time-frequency resolution can be increased, thereby providing a more refined analysis method for the voiceprint signal.
Therefore, the characteristic phasor of the signal frame includes an energy value, and in the present embodiment, the energy value is the sum of the energies of the frequency bands divided by the wavelet packet.
The method specifically comprises the following steps:
s021, and a segmentation step.
One signal frame is divided by wavelet packet, namely, two-step division is carried out on low-frequency and high-frequency bands at the same time, and the division levels can be set by designers according to actual situations. The final entire signal frame is divided into a plurality of uniform frequency bands. Thus a single frame
After i layers of wavelet packet division, the method is obtained
The sub-bands of the frequency of the individual signals,
the single frame at this time can be represented as:
wherein f is a function representing the wavelet packet decomposition result, and t is a set, and the set includes all sampling points in the sub-band.
S022, obtaining the energy of the sub-frequency band.
If i =2, i.e. two layers are divided, a signal frame is divided into 4 sub-bands, i.e. sub-band 0, sub-band 1, sub-band 2 and sub-band 3, so the energy E of sub-band k is:
in this formula, i is the number of decomposition levels, k represents the subband number,
f is a characteristic wavelet packetAnd (4) decomposing the function of the result, wherein N is the total number of the sampling points in the sub-frequency band, and N is the sampling point number.
S023, a step of presenting the frequency band distribution values.
In the feature phasor M of this signal frame, 3 feature values and the band distribution value described in embodiment 1 need to be represented, and the band distribution value lists all of the sub-bands.
For example, the nth signal frame, after i-layer wavelet packet decomposition, has the final characteristic phasor expressed as:
it should be noted that, in other links of the present technical solution, various data signals are often in a time domain format. In the wavelet packet energy extraction algorithm, the data signal is in a frequency domain format. The conversion between the time domain format and the frequency domain format can be realized by means of discrete fast fourier transform FFT, and the specific conversion process is the content of the prior art and is not described herein again.
Embodiment 4 differs from embodiment 1 in S02 and the signal frame feature vector extraction step.
In embodiment 1, the pattern in which the characteristic phasor is "3+1" is employed, and the 3 characteristic values are the root mean square value cerms of the signal frame, the mean square error C MES of the signal frame, and the kurtosis C K of the signal frame, respectively. In this embodiment, a characteristic value, i.e. skewness characteristic Cs, is added on the basis to reflect the data distribution symmetry of the signal, and the formula is as follows:
n is the number of audio sampling points in a frame, and i is the number of the audio sampling points and is a natural number.
X is the specific data content of the sample point,
is the data content of a sample pointThe mathematical expectation of (2).
Embodiment 5 differs from embodiment 1 in that in S03, a cluster identification step, a Mean Shift algorithm, i.e., a Mean Shift algorithm, is used.
The most obvious advantages of the mean shift algorithm are that the calculation amount is small, the method is simple and easy to implement, the real-time performance is good, and the method is suitable for real-time tracking occasions. The application scenario of the invention is the detection of abnormal sound of equipment, and when the equipment generates the abnormal sound, the equipment is indicated to have a fault, and the production and life safety is possibly influenced. Therefore, the mean shift method has the characteristic of good real-time performance, and is more favorable for quickly judging abnormal sounds and giving an alarm. The selection of the feature vector '3+1' is also to select a feature value with a small calculation amount while giving consideration to effects as much as possible, and is also based on real-time and efficient consideration.
Claims (10)
1. An interference identification method suitable for abnormal sound detection of industrial equipment is characterized by comprising the following steps: s01, a voiceprint segmentation step, namely segmenting the voiceprint signals extracted by the microphone into a plurality of second-level signal frames; s02, a signal frame characteristic phasor M extraction step, wherein the characteristic phasor M of each signal frame is extracted, and the characteristic phasor M comprises a characteristic value of the signal frame and a frequency band distribution value of signal energy of the signal frame; the characteristic value comprises the root mean square value of the signal frame
Reflecting the overall energy of the signal; mean square error of signal frame
Reflecting the degree of dispersion of the signal; kurtosis of signal frames
For reflecting the impulse component of the signal; the calculation formulas of the three are respectively as follows:
wherein N is the total number of audio sampling points in a signal frame, i is the number of the audio sampling points and is a natural number, and Ci is the data content of a single audio sampling point; the frequency band distribution value of the signal energy of the signal frame is used for reflecting the energy distribution condition of each frequency band of the signal frame; the feature vectors M of all the signal frames form a feature matrix; and S03, a cluster identification step, namely analyzing the characteristic matrix by using a cluster analysis algorithm, judging that the voiceprint signal has no short-time interference sound if all the signal frames are clustered into a cluster finally after the analysis is finished, and judging that the voiceprint signal has the short-time interference sound if the voiceprint signal cannot be clustered into a cluster finally.
2. The interference recognition method for abnormal sound detection of industrial equipment according to claim 1, wherein: in the step S01, in the process of dividing the voiceprint signal into a plurality of signal frames, an overlapping area is set between adjacent frames.
3. The interference recognition method for abnormal sound detection of industrial equipment according to claim 2, wherein: each signal frame has the same length which is T; the length of the overlapping area is T/2.
4. The interference recognition method for abnormal sound detection of industrial equipment according to claim 1, wherein: in S02, the feature value of the signal frame further includes an skewness feature Cs, which is used to reflect the symmetry of the data distribution of the signal, and the formula is as follows: 
(ii) a Wherein N is the number of audio sampling points in a frame, i is the number of audio sampling points and is a natural number, X is the specific data content of the sampling points,
is a mathematical expectation of the data content of the sample points.
5. The interference recognition method for abnormal sound detection of industrial equipment according to claim 1, wherein: in S02, a square summation algorithm of amplitude values of each frequency point is adopted for a calculation manner of the frequency band distribution value of the signal frame.
6. The interference recognition method suitable for abnormal sound detection of industrial equipment according to claim 1, wherein: in S02, a wavelet packet energy extraction algorithm is adopted for calculating the frequency band distribution value of the signal frame, and the method specifically includes the following steps: s021, a segmentation step; one signal frame is divided into small wave packets, and single frame
After i layers of wavelet packet division, the method is obtained
A sub-band of signals; s022, obtaining energy of a sub-frequency band; obtaining the energy value of each sub-frequency band, wherein the energy E of the sub-frequency band which is divided by the i-layer wavelet packet and is numbered as k is as follows:
in the formula, i is the number of decomposition layers, k represents the number of a sub-frequency band, f is a function representing the decomposition result of the wavelet packet, N is the total number of sampling points in the sub-frequency band, and N is the number of the sampling points; s023, a step of presenting a frequency band distribution value; each will beThe sub-band energy values E are present in said characteristic phasors M of the signal frame,
n is the number of the signal frame, and i is the number of the wavelet packet decomposition layers.
7. The interference recognition method for abnormal sound detection of industrial equipment according to claim 1, wherein: in the step S03 of cluster identification, the algorithm of cluster analysis is a K-Means algorithm or a clustering DBSCAN algorithm based on density or a hierarchical clustering algorithm.
8. The interference recognition method for abnormal sound detection of industrial equipment according to claim 1, wherein: in the step of S03 and cluster identification, the algorithm of cluster analysis is a Mean Shift algorithm.
9. The interference recognition method for abnormal sound detection of industrial equipment according to claim 1, wherein: in the step S01, the time length of the voiceprint signal extracted by the microphone is sampled for more than one minute.
10. The interference recognition method suitable for abnormal sound detection of industrial equipment according to claim 9, wherein: in the step S01, the sampling rate of the voiceprint signal extracted by the microphone is 11kHz, 22kHz, 44.1kHz, 48kHz, or 96kHz.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211704949.0A CN115691509A (en) | 2022-12-29 | 2022-12-29 | Interference identification method suitable for abnormal sound detection of industrial equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211704949.0A CN115691509A (en) | 2022-12-29 | 2022-12-29 | Interference identification method suitable for abnormal sound detection of industrial equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115691509A true CN115691509A (en) | 2023-02-03 |
Family
ID=85056623
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211704949.0A Pending CN115691509A (en) | 2022-12-29 | 2022-12-29 | Interference identification method suitable for abnormal sound detection of industrial equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115691509A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111814872A (en) * | 2020-07-07 | 2020-10-23 | 长沙理工大学 | Power equipment environmental noise identification method based on time domain and frequency domain self-similarity |
| CN112164390A (en) * | 2020-09-16 | 2021-01-01 | 珠海格力电器股份有限公司 | Equipment fault identification method and system |
| CN114842870A (en) * | 2022-03-15 | 2022-08-02 | 国网安徽省电力有限公司 | Voiceprint anomaly detection method based on multi-band self-supervision |
| CN115064183A (en) * | 2022-05-11 | 2022-09-16 | 浙江运达风电股份有限公司 | Wind generating set running state monitoring method and system based on artificial intelligence algorithm |
-
2022
- 2022-12-29 CN CN202211704949.0A patent/CN115691509A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111814872A (en) * | 2020-07-07 | 2020-10-23 | 长沙理工大学 | Power equipment environmental noise identification method based on time domain and frequency domain self-similarity |
| CN112164390A (en) * | 2020-09-16 | 2021-01-01 | 珠海格力电器股份有限公司 | Equipment fault identification method and system |
| CN114842870A (en) * | 2022-03-15 | 2022-08-02 | 国网安徽省电力有限公司 | Voiceprint anomaly detection method based on multi-band self-supervision |
| CN115064183A (en) * | 2022-05-11 | 2022-09-16 | 浙江运达风电股份有限公司 | Wind generating set running state monitoring method and system based on artificial intelligence algorithm |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP7199608B2 (en) | Methods and apparatus for inspecting wind turbine blades, and equipment and storage media therefor | |
| CN109034046B (en) | Method for automatically identifying foreign matters in electric energy meter based on acoustic detection | |
| CN111814872B (en) | Power equipment environmental noise identification method based on time domain and frequency domain self-similarity | |
| CN107993648A (en) | A kind of unmanned plane recognition methods, device and electronic equipment | |
| CN115909675A (en) | Distributed edge computing power equipment sound monitoring method | |
| CN110988137A (en) | Abnormal sound detection system and method based on time-frequency domain characteristics | |
| CN115467787A (en) | Motor state detection system and method based on audio analysis | |
| CN115035913A (en) | Sound abnormity detection method | |
| CN116778964A (en) | Power transformation equipment fault monitoring system and method based on voiceprint recognition | |
| CN116776075A (en) | Fan blade health monitoring system and monitoring method thereof | |
| CN114004996A (en) | Abnormal sound detection method, abnormal sound detection device, electronic equipment and medium | |
| Xie et al. | Detecting frog calling activity based on acoustic event detection and multi-label learning | |
| CN111933186A (en) | Method, device and system for fault identification of on-load tap-changer | |
| CN117116293A (en) | Machine equipment fault diagnosis system in complex sound field environment | |
| CN115691509A (en) | Interference identification method suitable for abnormal sound detection of industrial equipment | |
| Marchi et al. | Audio onset detection: A wavelet packet based approach with recurrent neural networks | |
| CN116364108A (en) | Transformer voiceprint detection method and device, electronic equipment and storage medium | |
| CN116907029A (en) | Method for detecting abnormality of fan in outdoor unit, control device and air conditioner outdoor unit | |
| CN115293205A (en) | Anomaly detection method, self-encoder model training method and electronic equipment | |
| CN112237433B (en) | Electroencephalogram signal abnormity monitoring system and method | |
| CN114121025A (en) | Voiceprint fault intelligent detection method and device for substation equipment | |
| CN112466322A (en) | Electromechanical device noise signal feature extraction method | |
| CN112233693A (en) | Sound quality evaluation method, device and equipment | |
| Umaeswari et al. | Development of Innovative Algorithm for Sound Detection Based on FFT and Goertzel Algorithms | |
| CN116386872B (en) | Training method, method and device for identifying sleep quality, medium and electronic equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20230203 |