CN114234061A - Neural network-based intelligent judgment method for water leakage sound of pressurized operation water supply pipeline - Google Patents
Neural network-based intelligent judgment method for water leakage sound of pressurized operation water supply pipeline Download PDFInfo
- Publication number
- CN114234061A CN114234061A CN202111567351.7A CN202111567351A CN114234061A CN 114234061 A CN114234061 A CN 114234061A CN 202111567351 A CN202111567351 A CN 202111567351A CN 114234061 A CN114234061 A CN 114234061A
- Authority
- CN
- China
- Prior art keywords
- sound
- neural network
- model
- water supply
- adaboost
- 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
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 title claims abstract description 54
- 238000000034 method Methods 0.000 title claims abstract description 27
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 10
- 238000001228 spectrum Methods 0.000 claims abstract description 41
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 23
- 230000007246 mechanism Effects 0.000 claims abstract description 10
- 238000007781 pre-processing Methods 0.000 claims abstract description 8
- 230000007613 environmental effect Effects 0.000 claims abstract description 7
- 238000001914 filtration Methods 0.000 claims abstract description 5
- 230000009467 reduction Effects 0.000 claims abstract description 5
- 230000003595 spectral effect Effects 0.000 claims abstract description 5
- 238000012549 training Methods 0.000 claims description 18
- 238000003062 neural network model Methods 0.000 claims description 12
- 238000011176 pooling Methods 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 4
- 238000013145 classification model Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000012935 Averaging Methods 0.000 claims description 2
- 238000011410 subtraction method Methods 0.000 claims description 2
- 238000001514 detection method Methods 0.000 description 11
- 238000004364 calculation method Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000037433 frameshift Effects 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000012804 iterative process Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 239000002184 metal Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 206010063385 Intellectualisation Diseases 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 238000009412 basement excavation Methods 0.000 description 1
- 101150102092 ccdB gene Proteins 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000002203 pretreatment Methods 0.000 description 1
Images
Classifications
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F17—STORING OR DISTRIBUTING GASES OR LIQUIDS
- F17D—PIPE-LINE SYSTEMS; PIPE-LINES
- F17D5/00—Protection or supervision of installations
- F17D5/02—Preventing, monitoring, or locating loss
- F17D5/06—Preventing, monitoring, or locating loss using electric or acoustic means
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/08—Speech classification or search
- G10L15/16—Speech classification or search using artificial neural networks
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
- G10L21/0208—Noise filtering
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Acoustics & Sound (AREA)
- Computational Linguistics (AREA)
- General Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Artificial Intelligence (AREA)
- Human Computer Interaction (AREA)
- Evolutionary Computation (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Theoretical Computer Science (AREA)
- Biophysics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mechanical Engineering (AREA)
- Quality & Reliability (AREA)
- Signal Processing (AREA)
- Examining Or Testing Airtightness (AREA)
Abstract
The invention discloses an intelligent water leakage sound judging method for a water supply pipeline running under pressure based on a neural network, which comprises the following steps: collecting sound and environmental audio of a target water supply pipeline operated under pressure to obtain an audio file; preprocessing the obtained audio file, including band-pass filtering and Berouti spectral subtraction noise reduction; extracting a Mel scale frequency spectrum from the filtered signal, and making the pipeline sound into corresponding digital characteristic data; inputting the digital characteristic data into a preset mixed classification prediction model, and outputting a classification result; the network structure of the preset hybrid classification prediction model is formed by combining the network structure of a deep convolutional neural network model, the structure of an Adaboost model and a final decision rule mechanism. The preset mixed classification prediction model is formed by combining the network structure of the deep convolutional neural network model and the structure of the Adaboost model, the advantages of the deep convolutional neural network model and the Adaboost model are integrated, the robustness is enhanced, and the accuracy of water leakage sound discrimination of a water supply pipeline running under pressure can be improved.
Description
Technical Field
The application relates to the technical field of pipeline leakage detection, in particular to an intelligent water leakage sound judging method for a water supply pipeline running under pressure based on a neural network.
Background
The sound listening exploration method is a conventional technical means of water supply pipeline leakage point active detection and leakage point positioning, has simple principle and long engineering application history, and is widely adopted in domestic and foreign water supply industries. At present, the leakage inspection of most urban pipe networks in China is not free from a method of manually listening leakage, and even if relevant instruments and other equipment are applied, the positioning range needs to be narrowed by manually listening before excavation. In performing leak detection surveys, workers use mechanical leak listening bars or portable measuring devices to acquire sound to detect water leaks.
The leakage detection device disclosed in the invention patent with the publication number of CN101329012A and the sound vibration method water leakage detection method disclosed in the invention patent with the publication number of CN112303502A are used for identifying water leakage sound based on pipeline operation sound, and industrial application based on the invention patent also obtains better feedback. Although the accuracy of this method is already quite good, it still has some major drawbacks, one of which, being the most prominent, is that it is a labor intensive task, and the accuracy of the detection depends to a large extent on the experience of the worker. With the popularization of the digital technology, in the field of water leakage detection which is very key and is relevant to the happy life of people, the related detection means must realize intellectualization and high efficiency.
In recent years, with the rise of artificial intelligence technology, machine learning and deep learning methods have begun to be popularized and applied in a large scale in many fields. Therefore, in order to make the acoustic-based detection method more efficient and intelligent, it is one of the best directions to introduce machine intelligence decision.
Disclosure of Invention
The application aims to provide a method for intelligently judging water leakage sound of a water supply pipeline in a pressure operation mode, so that the current leakage detection mode of the water supply pipeline is changed, and the working efficiency is improved.
In order to achieve the above object, the present application provides an intelligent method for discriminating water leakage sound of a water supply pipeline running under pressure based on a neural network, comprising:
collecting sound and environmental sound of a target water supply pipeline operated under pressure to obtain an audio file;
preprocessing the obtained audio file: noise reduction is carried out by applying a band-pass filter and a Berouti spectral subtraction method, and clutter signals are filtered;
extracting a Mel scale frequency spectrum from the filtered signal, and making the pipeline sound into corresponding digital characteristic data;
inputting the digital characteristic data into a preset mixed classification prediction model, and outputting a classification result; the network structure of the preset hybrid classification prediction model is formed by combining a network structure of a deep convolutional neural network model, a structure of an Adaboost model and a final decision rule mechanism.
Further, the audio file preprocessing unit includes:
applying a digital band-pass filter to the obtained pipeline sound and the environment sound, wherein the pass-band range is controlled to be 200 Hz-2500 Hz;
and taking the environment sound as a background sound, taking the pipeline sound as a foreground sound, extracting short-time frequency spectrums of the background sound and the foreground sound, respectively obtaining an amplitude spectrum and a phase spectrum of the foreground sound and an amplitude spectrum and a phase spectrum of the background sound, and subtracting the amplitude spectrum of the background sound from the amplitude spectrum of the foreground sound by applying a Berouti spectrum subtraction algorithm to obtain a clean pipeline sound amplitude spectrum.
Further, the inputting the pipeline sound digital characteristic data into a preset mixed classification model and outputting a classification result includes:
constructing an original deep convolutional neural network model according to the network structure of the VGG11_ bn model;
extracting Mel scale frequency spectrum digital characteristics according to historically collected pipeline water leakage sound data and a public urban environment sound data set, and training the original deep convolution neural network model by using sample digital characteristic data to obtain the deep convolution neural network model;
after an original Adaboost model is constructed, a Mel cepstrum coefficient is calculated for the Mel scale frequency spectrum digital characteristics, and the original Adaboost model is trained by utilizing the Mel cepstrum coefficient to obtain the Adaboost model;
and sending the output of the deep convolutional neural network model and the output of the Adaboost model into a final decision rule mechanism, and outputting the classification result through final decision.
Further, training the original deep convolutional neural network model by using sample digital feature data to obtain the deep convolutional neural network model, including:
after the sample digital characteristic data is input into the original deep convolutional neural network model, processing a convolutional layer, a pooling layer and a full-link layer of the original deep convolutional neural network model in sequence to obtain a corresponding output value;
calculating an error value between the output value and a preset target value, and judging whether the error value is greater than a preset expected value or not;
if the error value is larger than the preset expected value, respectively adjusting the weight values of the convolutional layer, the pooling layer and the full-connection layer according to the error value;
inputting the sample digital feature data into the original depth convolution neural network model after the weight adjustment to obtain a corresponding output value:
after calculating an error value corresponding to the output value, judging whether the error value is greater than the preset expected value:
and if the error value is not greater than the preset expected value, obtaining the deep convolutional neural network model.
Further, training the original Adaboost model by using the mel-frequency cepstrum coefficient to obtain the Adaboost model, including:
after the original Adaboost model is constructed, extracting a Mel cepstrum coefficient from the Mel scale frequency spectrum digital characteristics;
initializing the weight of a training sample, wherein for m samples, the weight of each training sample is initialized to 1/m;
and iteratively training the weak classifiers. In the iterative process, the sample weights need to be updated. If a sample point has been accurately classified, its weight is reduced when the next weak classifier is trained. Conversely, if the sample point is not classified accurately, its weight is increased.
Further, the final decision rule mechanism has a decision rule description as: if the Adaboost output is greater than the decision threshold or the deep convolutional neural network output is greater than the decision threshold, judging that the final result is the water leakage class, and otherwise, judging that the final result is the normal working class.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below.
Fig. 1 is a flowchart of an intelligent method for judging water leakage sound of a water supply pipeline running under pressure based on a neural network according to an embodiment of the present application.
Fig. 2 is a flowchart of a pipeline tone preprocessing procedure provided in an embodiment of the present application.
Fig. 3 is a flowchart illustrating a neural network model operation according to an embodiment of the present disclosure.
Detailed Description
The core of the application is to provide an intelligent water leakage sound distinguishing method for the water supply pipeline running under pressure, which can effectively replace the traditional mode of recognizing water leakage events of the water supply pipeline by artificial listening, and improve the detection efficiency.
In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some embodiments of the present application, but not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
The embodiment of the application provides a water leakage sound intelligent judgment method for a water supply pipeline running under pressure, which comprises the following steps:
s1, obtaining an audio file of the sound and the environmental sound of the target water supply pipeline with pressure operation.
The embodiment of the application firstly requires to acquire the audio file of the sound and the environmental sound of the target water supply pipeline with pressure operation. The target pipeline is required to be a metal pipeline; the pipeline sound required to be obtained, namely the listening position, needs to be on the pipeline or metal equipment such as a valve, a fire hydrant and the like connected with the pipeline; the audio file format is required to be wav format, PCM-16 coding is carried out, the sampling rate is not lower than 22.05kHz, and single channel or double channels can be adopted; the environmental sound is required to be sound near the water leakage sound collection point, and the recording time of the two audio files is required to be not less than 4s and not more than 1 min. The audio acquisition device is not specifically limited, and the environment is not specifically limited, for example, the audio acquisition device may be a specially-made embedded USB recorder manufactured based on an MEMS microphone, and the environment may be a road, a park, a community, or the like.
And S2 audio file preprocessing, which is to perform band-pass filtering on the two audios obtained in S1 and then perform noise reduction on the pipeline sound, wherein the preprocessing flow is shown in FIG. 2. The pretreatment method is as follows:
according to the embodiment of the application, time length alignment is carried out after the audio file is obtained. Based on the time length t1 of the pipe sound s1, if the time length t2 of the environment sound s2 is greater than the time length t1 of the pipe sound, only the environment sound with the time length t1 is read, and if the time length t2 of the environment sound s2 is less than the time length t1 of the pipe sound, the content of the environment sound s2 is repeatedly filled until the time length t1 is met.
And performing digital band-pass filtering on the two aligned audios. An 8-order Butterworth band-pass filter is used, and the filtering range is controlled to be 200 Hz-2500 Hz.
And carrying out short-time Fourier transform on the filtered signal. Let the filtered pipe sound data be r1 and the ambient sound data be r 2. And performing short-time Fourier transform on r1 and r2, wherein the transform parameters are as follows: the sampling rate is 22.05kHz, the window selection signal is a Hanning window, the window length is 2048 data points, the frame shift is 512 data points, the fft point number is 2048 points, and the length is equal to the window length.
After short-time Fourier transform, r1 has amplitude p1 and phase angle a1, and r2 has amplitude p2 and phase angle a 2.
The Berouti spectral subtraction algorithm was applied to p1, a1, p2, a 2. The algorithm may be expressed in the following pseudo-code:
letD(w)＝P_s(w)-alpha*P_n(w)
P'_s＝D(w)ifD(w)>beta*P_n(w)else beta*P_n(w)
alpha>＝1,0<beta<<1
wherein the square of the absolute value of P1 can be substituted into P _ s (w), and the square of the absolute value of P2 can be substituted into P _ n (w).
And the obtained P' _ s is squared and then subjected to short-time Fourier inversion with the phase angle a1, and clean pipeline sound data r3 can be obtained.
S3, calculating frequency spectrum of the clean pipeline sound data, and extracting digital features.
After the clean pipeline sound data r3 is obtained, the extraction method for extracting the Mel frequency spectrogram from r3 is as follows:
in the first step, r3 is cut for a length of 4s, with adjacent segments having a 3s repeat region and the last segment being discarded less than 4 s.
In the second step, for the 4s audio, the signal is first framed into 128 frames, each covering 31.25 ms. These frames are subjected to a short-time fourier transform and the spectrum on the log-mel scale is extracted.
The calculation formula is expressed as:
the short-time Fourier transform formula:
where x (n) is the original signal amplitude, w (n) is the window selection signal, in the present invention, the window selection signal selects the hamming window, and R represents the frame shift length, in the present invention, the frame shift selects 7.8125 ms. Then 128 short-time spectrum X can be obtained by the above formula128(ej ω)。
Mel scale calculation formula:
the short-time spectrum is converted to a frequency scale on the mel scale, and these frequencies are passed through a 128-channel triangular filter. Taking the weighted sum of all signal amplitudes in the frequency bandwidth of each triangular filter as the output of a certain band-pass filter, and then carrying out logarithmic operation on the output of all the filters, wherein the calculation formula is as follows:
wherein, o (l), c (l), and h (l) are the lower limit, center frequency, and upper limit frequency of the l-th triangular filter, respectively, and the triangular filters are made to follow the following relations:
c(l)＝h(l-1)＝o(l+1)
then, the calculated log-mel spectrum is:
mel(l)＝lgm(l)
third, the 128 × 128 two-dimensional mel-frequency spectrum arrays are stacked to form four-dimensional tensors, and the shape of each tensor is n × 1 × 128 × 128.
S4, inputting the digital characteristic data into a preset mixed classification model, and outputting a classification result; the network structure of the preset hybrid classification prediction model is integrated by the network structure of the deep convolutional neural network model and the structure of the Adaboost model.
Further, when a deep convolutional neural network model is constructed, a structure of VGG11_ bn is adopted, that is, a main part of the VGG11 network with convolutional-pooling-batch normalization operation is selected, a certain simplification is performed in a full connection layer part, and the output sizes of a first full connection layer and a second full connection layer are set to be 128.
After the deep convolution neural network is constructed, the historical collected pipeline sound data including water leakage sound and normal sound of pipeline operation are used, and the Mel scale frequency spectrum digital features are extracted by combining the public urban environment sound data set. After the Mel scale frequency spectrum digital characteristic data of the sample is input into the original deep convolution neural network model, processing the convolution layer, the pooling layer, the batch normalization layer and the full-connection layer of the original deep convolution neural network model in sequence to obtain a corresponding output value; calculating an error value between the output value and a preset target value, and judging whether the error value is greater than a preset expected value or not; if the error value is larger than the preset expected value, respectively adjusting the weight values of the convolutional layer, the pooling layer and the full-connection layer according to the error value; inputting the sample digital feature data into the original depth convolution neural network model after the weight adjustment to obtain a corresponding output value: after calculating an error value corresponding to the output value, judging whether the error value is greater than the preset expected value: and if the error value is not greater than the preset expected value, obtaining the deep convolutional neural network model.
Further, after the Adaboost model is constructed, clean pipeline sound data collected historically comprise water leakage sound and normal sound of pipeline operation, Mel scale frequency spectrum digital features are extracted by combining a public urban environment sound data set, and then 40-order Mel cepstrum coefficients are extracted based on the digital features to form a sample set. Initializing the weight of a training sample, wherein for m samples, the weight of each training sample is initialized to 1/m; and iteratively training the weak classifiers. In the iterative process, the sample weights need to be updated. If a sample point has been accurately classified, its weight is reduced when the next weak classifier is trained. Conversely, if the sample point is not classified accurately, its weight is increased.
And weighting and averaging each weak classifier to obtain a strong classifier which is used as the Adaboost model.
When two models obtained by training are applied in practice, as shown in fig. 2, preprocessed audio is firstly subjected to mel spectrum feature extraction and stacked into a tensor, and then the tensor is sent into a deep convolution neural network after training is completed to calculate a result; and extracting a Mel cepstrum coefficient vector from the tensor, sending the Mel cepstrum coefficient vector into an Adaboost model after training, and calculating a result. The results of the two models are jointly sent to a decision mechanism, and the final result is determined by the decision mechanism.
The operating logic of the decision mechanism can be expressed as: if the Adaboost output is greater than the decision threshold or the deep convolutional neural network output is greater than the decision threshold, judging that the final result is the water leakage class, and otherwise, judging that the final result is the normal working class.
This application organically combines classic signal processing technique and emerging artificial intelligence technique to introduce this kind of technique into this kind of field of detecting that leaks, can effectively improve the rate of accuracy that the sound detected that leaks. In addition, when the model is selected, the deep learning technology and the classic machine learning model are organically combined, so that the advantages of the two models can be exerted to the maximum extent, and the robustness of the system can be improved.
The above description of the embodiments is only intended to help understand the method of the present application and its core ideas. It should be noted that, for those skilled in the art, it is possible to make several improvements and modifications to the present application without departing from the principle of the present application, and such improvements and modifications also fall within the scope of the claims of the present application.
Claims (6)
1. An intelligent water leakage sound judging method for a water supply pipeline running under pressure based on a neural network is characterized by comprising the following steps: collecting sound and environmental sound of a target water supply pipeline operated under pressure to obtain an audio file; preprocessing the obtained audio file: noise reduction is carried out by applying a band-pass filter and a Berouti spectral subtraction method, and clutter signals are filtered; extracting a Mel scale frequency spectrum from the filtered signal, and making the pipeline sound into corresponding digital characteristic data; inputting the digital characteristic data into a preset mixed classification prediction model, and outputting a classification result; the network structure of the preset hybrid classification prediction model is formed by combining a network structure of a deep convolutional neural network model, a structure of an Adaboost model and a final decision rule mechanism.
2. The intelligent water leakage sound distinguishing method based on the neural network for the water supply pipeline running under pressure as claimed in claim 1, wherein the audio data preprocessing unit comprises two steps: band-pass filtering and spectral subtraction noise reduction;
applying a digital band-pass filter to the pipeline sound and the environmental sound, wherein the pass-band range is controlled to be 200 Hz-2500 Hz;
and taking the environment sound as a background sound, taking the pipeline sound as a foreground sound, extracting short-time frequency spectrums of the background sound and the foreground sound, respectively obtaining an amplitude spectrum and a phase spectrum of the foreground sound and an amplitude spectrum and a phase spectrum of the background sound, and subtracting the amplitude spectrum of the background sound from the amplitude spectrum of the foreground sound by applying a Berouti spectrum subtraction algorithm to obtain a clean pipeline sound amplitude spectrum.
3. The method for intelligently distinguishing the water leakage sound of the water supply pipeline running under pressure according to claim 1, wherein the step of inputting the digital characteristic data of the pipeline sound into a preset mixed classification model and outputting a classification result comprises the following steps:
constructing an original deep convolutional neural network model according to the network structure of the VGG11_ bn model;
extracting Mel scale frequency spectrum digital characteristics by combining a public urban environment sound data set according to historically collected pipeline sound data including water leakage sound and normal pipeline operation sound, and training the original deep convolution neural network model by utilizing Mel scale frequency spectrum digital characteristic data of a sample to obtain the deep convolution neural network model;
after an original Adaboost model is constructed, extracting a Mel cepstrum coefficient from the Mel scale frequency spectrum digital characteristics, and training the original Adaboost model by using the Mel cepstrum coefficient to obtain the Adaboost model;
and sending the output of the deep convolutional neural network model and the output of the Adaboost model into a final decision rule mechanism, and outputting the classification result through final decision.
4. The method for intelligently distinguishing the water leakage sound of the water supply pipeline running under pressure according to claim 3, wherein the training of the original deep convolutional neural network model by using the Mel scale frequency spectrum digital feature data of the sample to obtain the deep convolutional neural network model comprises the following steps:
after the Mel scale frequency spectrum digital characteristic data of the sample is input into the original deep convolution neural network model, processing a convolution layer, a pooling layer, a batch normalization layer and a full connection layer of the original deep convolution neural network model in sequence to obtain a corresponding output value;
calculating an error value between the output value and a preset target value, and judging whether the error value is greater than a preset expected value or not;
if the error value is larger than the preset expected value, respectively adjusting the weight values of the convolutional layer, the pooling layer and the full-connection layer according to the error value;
inputting the sample digital feature data into the original depth convolution neural network model after the weight adjustment to obtain a corresponding output value:
after calculating an error value corresponding to the output value, judging whether the error value is greater than the preset expected value:
and if the error value is not greater than the preset expected value, obtaining the deep convolutional neural network model.
5. The method for intelligently distinguishing the water leakage sound of the water supply pipeline running under pressure according to claim 3, wherein after an original Adaboost model is constructed, a 40-order Mel cepstrum coefficient is extracted from the Mel scale frequency spectrum digital features, and the original Adaboost model is trained by using the Mel cepstrum coefficient to obtain the Adaboost model, and the method comprises the following steps:
after the original Adaboost model is constructed, extracting a 40-order Mel cepstrum coefficient from the Mel scale frequency spectrum digital characteristics;
initializing the weight of a training sample, wherein for m samples, the weight of each training sample is initialized to 1/m;
iteratively training a weak classifier; in the iteration process, the sample weight needs to be updated; if a sample point has been accurately classified, its weight is reduced when the next weak classifier is trained; conversely, if the sample point is not classified accurately, its weight is increased;
and weighting and averaging each weak classifier to obtain a strong classifier which is used as the Adaboost model.
6. The method for intelligently judging the water leakage sound of the water supply pipeline running under pressure according to claim 3, wherein the decision rule of the final decision rule mechanism is described as follows: if the Adaboost output is greater than the decision threshold or the deep convolutional neural network output is greater than the decision threshold, judging the class of water leakage, and otherwise, judging the class of normal operation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111567351.7A CN114234061A (en) | 2021-12-20 | 2021-12-20 | Neural network-based intelligent judgment method for water leakage sound of pressurized operation water supply pipeline |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111567351.7A CN114234061A (en) | 2021-12-20 | 2021-12-20 | Neural network-based intelligent judgment method for water leakage sound of pressurized operation water supply pipeline |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114234061A true CN114234061A (en) | 2022-03-25 |
Family
ID=80759850
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111567351.7A Pending CN114234061A (en) | 2021-12-20 | 2021-12-20 | Neural network-based intelligent judgment method for water leakage sound of pressurized operation water supply pipeline |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114234061A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115655594A (en) * | 2022-12-12 | 2023-01-31 | 杭州兆华电子股份有限公司 | Method and device for distinguishing and identifying throttling and leakage of pipeline based on acoustic cloud chart |
| CN115950590A (en) * | 2023-03-15 | 2023-04-11 | 凯晟动力技术（嘉兴）有限公司 | Gas engine leakage early warning system |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101839795A (en) * | 2010-05-10 | 2010-09-22 | 任振伟 | The leak diagnostic systems of pressure-bearing pipe of boiler and method |
| CN107061996A (en) * | 2017-03-10 | 2017-08-18 | 内蒙古大学 | A kind of water supply line leakage detecting and locating method |
| CN107461611A (en) * | 2017-08-24 | 2017-12-12 | 南京邮电大学 | The leakage detection method and leak detecting device being combined based on small echo and EMD reconstruct |
| CN109034046A (en) * | 2018-07-20 | 2018-12-18 | 国网重庆市电力公司电力科学研究院 | Foreign matter automatic identifying method in a kind of electric energy meter based on Acoustic detection |
| CN109084186A (en) * | 2018-08-17 | 2018-12-25 | 常州大学 | Pipeline leakage signal recognition methods based on improved ELMD multi-scale entropy |
| CN110047512A (en) * | 2019-04-25 | 2019-07-23 | 广东工业大学 | A kind of ambient sound classification method, system and relevant apparatus |
| CN111044285A (en) * | 2019-11-22 | 2020-04-21 | 军事科学院***工程研究院军用标准研究中心 | Method for diagnosing faults of mechanical equipment under complex conditions |
| CN111625763A (en) * | 2020-05-27 | 2020-09-04 | 郑州航空工业管理学院 | Operation risk prediction method and prediction system based on mathematical model |
| CN113009447A (en) * | 2021-03-05 | 2021-06-22 | 长安大学 | Road underground cavity detection early warning method based on deep learning and ground penetrating radar |
| CN113514544A (en) * | 2020-12-29 | 2021-10-19 | 大连理工大学 | Mobile robot pavement material identification method based on sound characteristics |
-
2021
- 2021-12-20 CN CN202111567351.7A patent/CN114234061A/en active Pending
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101839795A (en) * | 2010-05-10 | 2010-09-22 | 任振伟 | The leak diagnostic systems of pressure-bearing pipe of boiler and method |
| CN107061996A (en) * | 2017-03-10 | 2017-08-18 | 内蒙古大学 | A kind of water supply line leakage detecting and locating method |
| CN107461611A (en) * | 2017-08-24 | 2017-12-12 | 南京邮电大学 | The leakage detection method and leak detecting device being combined based on small echo and EMD reconstruct |
| CN109034046A (en) * | 2018-07-20 | 2018-12-18 | 国网重庆市电力公司电力科学研究院 | Foreign matter automatic identifying method in a kind of electric energy meter based on Acoustic detection |
| CN109084186A (en) * | 2018-08-17 | 2018-12-25 | 常州大学 | Pipeline leakage signal recognition methods based on improved ELMD multi-scale entropy |
| CN110047512A (en) * | 2019-04-25 | 2019-07-23 | 广东工业大学 | A kind of ambient sound classification method, system and relevant apparatus |
| CN111044285A (en) * | 2019-11-22 | 2020-04-21 | 军事科学院***工程研究院军用标准研究中心 | Method for diagnosing faults of mechanical equipment under complex conditions |
| CN111625763A (en) * | 2020-05-27 | 2020-09-04 | 郑州航空工业管理学院 | Operation risk prediction method and prediction system based on mathematical model |
| CN113514544A (en) * | 2020-12-29 | 2021-10-19 | 大连理工大学 | Mobile robot pavement material identification method based on sound characteristics |
| CN113009447A (en) * | 2021-03-05 | 2021-06-22 | 长安大学 | Road underground cavity detection early warning method based on deep learning and ground penetrating radar |
Non-Patent Citations (3)
| Title |
|---|
| 吕鑑等: "基于BP 神经网络的城市给水管网爆损预测模型研究", 《给水排水》, vol. 35, 31 December 2009 (2009-12-31) * |
| 焦敬品等: "基于BP 神经网络的管道泄漏声信号识别方法研究", 《仪器仪表学报》, vol. 37, no. 11, 30 November 2016 (2016-11-30) * |
| 燕宗伟等: "基于遗传算法优化BP神经网络的 管道泄漏检测方法研究", 《当代化工》, vol. 49, no. 1, 31 January 2020 (2020-01-31) * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115655594A (en) * | 2022-12-12 | 2023-01-31 | 杭州兆华电子股份有限公司 | Method and device for distinguishing and identifying throttling and leakage of pipeline based on acoustic cloud chart |
| CN115655594B (en) * | 2022-12-12 | 2023-03-10 | 杭州兆华电子股份有限公司 | Method and device for distinguishing and identifying throttling and leakage of pipeline based on acoustic cloud chart |
| CN115950590A (en) * | 2023-03-15 | 2023-04-11 | 凯晟动力技术（嘉兴）有限公司 | Gas engine leakage early warning system |
| CN115950590B (en) * | 2023-03-15 | 2023-05-30 | 凯晟动力技术（嘉兴）有限公司 | Gas engine leakage early warning system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108447495B (en) | Deep learning voice enhancement method based on comprehensive feature set | |
| CN110246510B (en) | End-to-end voice enhancement method based on RefineNet | |
| CN114234061A (en) | Neural network-based intelligent judgment method for water leakage sound of pressurized operation water supply pipeline | |
| CN113405825B (en) | Belt conveyor fault diagnosis method based on sound signals | |
| CN110176250B (en) | Robust acoustic scene recognition method based on local learning | |
| KR20060082465A (en) | Method and apparatus for classifying voice and non-voice using sound model | |
| CN110890087A (en) | Voice recognition method and device based on cosine similarity | |
| CN108198561A (en) | A kind of pirate recordings speech detection method based on convolutional neural networks | |
| WO2019232833A1 (en) | Speech differentiating method and device, computer device and storage medium | |
| CN111696580A (en) | Voice detection method and device, electronic equipment and storage medium | |
| CN109300470B (en) | Mixing separation method and mixing separation device | |
| CN111081223A (en) | Voice recognition method, device, equipment and storage medium | |
| Hao et al. | Time-domain neural network approach for speech bandwidth extension | |
| Li et al. | Multisensory speech enhancement in noisy environments using bone-conducted and air-conducted microphones | |
| CN109102818B (en) | Denoising audio sampling algorithm based on signal frequency probability density function distribution | |
| CN117542373A (en) | Non-air conduction voice recovery system and method | |
| CN111785262A (en) | Speaker age and gender classification method based on residual error network and fusion characteristics | |
| Kaminski et al. | Automatic speaker recognition using a unique personal feature vector and Gaussian Mixture Models | |
| Dov et al. | Voice activity detection in presence of transients using the scattering transform | |
| CN113327589B (en) | Voice activity detection method based on attitude sensor | |
| CN115331678A (en) | Generalized regression neural network acoustic signal identification method using Mel frequency cepstrum coefficient | |
| KR101327572B1 (en) | A codebook-based speech enhancement method using speech absence probability and apparatus thereof | |
| CN113744725A (en) | Training method of voice endpoint detection model and voice noise reduction method | |
| CN113488069A (en) | Method and device for quickly extracting high-dimensional voice features based on generative countermeasure network | |
| CN112908344A (en) | Intelligent recognition method, device, equipment and medium for bird song |
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 |