CN107061996B - A kind of water supply line leakage detecting and locating method - Google Patents
A kind of water supply line leakage detecting and locating method Download PDFInfo
- Publication number
- CN107061996B CN107061996B CN201710139772.7A CN201710139772A CN107061996B CN 107061996 B CN107061996 B CN 107061996B CN 201710139772 A CN201710139772 A CN 201710139772A CN 107061996 B CN107061996 B CN 107061996B
- Authority
- CN
- China
- Prior art keywords
- signal
- leakage
- noise
- spectrum
- leak
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 78
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 title claims abstract description 70
- 238000001228 spectrum Methods 0.000 claims abstract description 69
- 230000003595 spectral effect Effects 0.000 claims abstract description 55
- 238000013528 artificial neural network Methods 0.000 claims abstract description 46
- 230000002708 enhancing effect Effects 0.000 claims abstract description 20
- 238000012549 training Methods 0.000 claims description 49
- 230000006870 function Effects 0.000 claims description 45
- 238000001514 detection method Methods 0.000 claims description 25
- 230000008569 process Effects 0.000 claims description 16
- 238000001914 filtration Methods 0.000 claims description 13
- 239000011159 matrix material Substances 0.000 claims description 8
- 238000012360 testing method Methods 0.000 claims description 7
- 238000004458 analytical method Methods 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000009432 framing Methods 0.000 claims description 4
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000000926 separation method Methods 0.000 claims 1
- 238000005314 correlation function Methods 0.000 description 8
- 230000000694 effects Effects 0.000 description 6
- 210000002569 neuron Anatomy 0.000 description 6
- 238000012546 transfer Methods 0.000 description 6
- 238000007476 Maximum Likelihood Methods 0.000 description 5
- 230000008859 change Effects 0.000 description 4
- 230000001537 neural effect Effects 0.000 description 4
- 238000002474 experimental method Methods 0.000 description 3
- 230000004807 localization Effects 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000002945 steepest descent method Methods 0.000 description 3
- 230000009466 transformation Effects 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 2
- 230000007423 decrease Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000005284 excitation Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000011946 reduction process Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 101100398248 Arabidopsis thaliana KIN10 gene Proteins 0.000 description 1
- 101000835860 Homo sapiens SWI/SNF-related matrix-associated actin-dependent regulator of chromatin subfamily B member 1 Proteins 0.000 description 1
- 102100025746 SWI/SNF-related matrix-associated actin-dependent regulator of chromatin subfamily B member 1 Human genes 0.000 description 1
- 239000000654 additive Substances 0.000 description 1
- 230000000996 additive effect Effects 0.000 description 1
- 238000005311 autocorrelation function Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 239000012141 concentrate Substances 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000002405 diagnostic procedure Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 239000004744 fabric Substances 0.000 description 1
- 239000003673 groundwater Substances 0.000 description 1
- 238000012886 linear function Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000011045 prefiltration Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 239000008399 tap water Substances 0.000 description 1
- 235000020679 tap water Nutrition 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
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
-
- 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
-
- 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
- G10L21/0216—Noise filtering characterised by the method used for estimating noise
- G10L21/0232—Processing in the frequency domain
-
- 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
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/27—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique
- G10L25/30—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique using neural networks
Landscapes
- Engineering & Computer Science (AREA)
- Human Computer Interaction (AREA)
- Acoustics & Sound (AREA)
- Signal Processing (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Computational Linguistics (AREA)
- Physics & Mathematics (AREA)
- Multimedia (AREA)
- Quality & Reliability (AREA)
- Mechanical Engineering (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Examining Or Testing Airtightness (AREA)
Abstract
The present invention provides a kind of water supply line leakage detecting and locating method, the method carries out spectrum to the signal that sensor acquires and subtracts enhancing, calculate the Spectral variance of signal after enhancing, judged using double threshold method, illustrate there is leakage if Spectral variance is in threshold range, leak point location is carried out again, filter is constituted using BP neural network, water leakage is separated from noise and carries out generalized correlation, time delay estimation is carried out according to the weight function that signal-to-noise ratio selects performance good, obtain the time delay of three sensors, it is calculated using leak location model, obtain leak dot position information.The present invention, to signal enhancing, then carries out leakage judgement using double threshold method using spectrum-subtraction, illustrates there is leakage if Spectral variance is in threshold range, and accuracy is higher;It is more obvious to the promotion of signal-to-noise ratio using " the No leakage estimation technique " estimation noise spectrum, it is filtered using BP neural network, improves the Time delay Estimation Accuracy of generalized correlation, to effectively improve leak spot placement accuracy.
Description
Technical field
The invention belongs to leak localization technical fields, and in particular to a kind of water supply line leakage detecting and locating method.
Background technique
Water supply pipe leakage detection and localization is the important prerequisite that pipe-line maintenance is repaired.According to official statistics, the country more than 600
The average leak rate of a public supply mains is more than 20%, up to 60% or more.China's most area is only in tap water
Leakage can be just judged in the case where large area leakage or even pipeline burst.Therefore, it detects in time and orients leakage hair
Raw point is of great significance for the saving of water resource.
Pipe leakage signal in practice is mixed by leakage signal and various ambient noises.Leakage signal can recognize
To be stationary signal, ambient noise include originally ground caused by pipe vibration noise, motor-driven vehicle going caused by water flowing shake
Dynamic, caused noise etc. of constructing can bring leak detection greatly tired since external interference noise not necessarily meets smooth conditions
It is difficult.
Therefore, first of all for leak detection effect when being lifted at low signal-to-noise ratio, the present invention is with leakage signal and noise
The difference of spectrum signature is starting point, proposes the leak detection algorithm based on spectrum-subtraction and Spectral variance.Spectrum-subtraction is a kind of
Effective voice enhancement algorithm, algorithm complexity is low, strong real-time, and the present invention is applied in the detection to leakage signal
On.Firstly, estimating using frequency spectrum of the algorithm to noise, then enhance water leakage by " spectrum subtraction ".Frequency spectrum
Variance is to be differed greatly using the spectral characteristic of leakage signal, and the spectral characteristic difference of ambient noise is smaller, to identify
Leakage signal.Next, being proposed a kind of based on BP mind to reduce the positioning accuracy of time delay evaluated error and then raising algorithm
Leak positioning mode through network calculates error according to output valve and desired value using error back propagation mechanism, then by error Lai
Successively modification weight, to complete network training, constructs the filtering system based on BP neural network so that error is minimum, into
And obtain accurate time delay using generalized correlation and estimate, realize the accurate positioning of leakage point.
On the basis of parser principle, emulation experiment is carried out under the conditions of different signal-to-noise ratio.The experimental results showed that should
Method can also obtain preferable detection locating effect under low signal-to-noise ratio and environmental condition complicated and changeable.
Summary of the invention
To solve the above-mentioned problems, the present invention provides a kind of water supply line leakage detecting and locating method, and the method is to biography
The signal of sensor acquisition is enhanced, and is then calculated the Spectral variance of signal after enhancing, is judged using double threshold method, if frequency
Spectrum variance, which is in threshold range, then illustrates there is leakage, then carries out leak point location, constitutes filter using BP neural network,
Water leakage is separated from noise, generalized correlation is carried out to water leakage, according to signal-to-noise ratio select the good weight function of performance into
Row time delay estimation, obtain three sensors when delay, calculated using leak location model, obtain leakage point position letter
Breath;
Further, which comprises
S1: water leakage is subjected to spectrum and subtracts enhancing;
S2: the Spectral variance of signal after enhancing in S1 is calculated;
S3: leakage signal is identified using one-parameter double-threshold comparison, and determines threshold value;
S4: leak detection, and output test result are carried out;
S5: to the testing result exported in S4, out-of-band noise is removed by bandpass filtering;
S6: carrying out NN filtering, removes in-band noise;
S7: generalized correlation is carried out;
S8: time delay estimation is carried out to the result of generalized correlation in S7, and is calculated by leak location model, is leaked
Water spot location information completes leak positioning;
Further, the S7 specifically: during selection of weighting function, according to the good power letter of signal-to-noise ratio selection performance
Number carry out time delay estimations, obtain three sensors when delay, calculated using the leak location model, obtain leakage point
Location information;
Further, the leak location model is three sensor location models, and the model is by equidistant point of sensor
Cloth is around water supply network, wherein the three sensor nodes number nearest apart from leak source is respectively 1,2,3, between them away from
From for L, position of the sensor 1,2 apart from leak source is set to d1 and d2, is measured between sensor 1 and 2,2 and 3 by correlation method
Time delay be respectively D12And D23, ν: ν=L/D of spread speed on pipeline where water leakage23, leak source is away from sensor 1 and 2
Distance and be respectivelyWith
Further, the calculating of spectrum-subtraction is specific as follows in the S1:
S11: pre-processing input water leakage s (n), and the pretreatment includes preemphasis and adding window framing;
S12: carrying out Short Time Fourier Analysis to the leakage signal that band is made an uproar, and the short-time energy spectrum of each frame signal is calculated
|Ym(ω)|2;
S13: before enhancing using spectrum-subtraction water leakage, noise spectrum estimation is carried out, noise spectrum estimation is obtained
ValueWith | Ym(ω)|2It subtractsTo obtain the leakage signal power spectrum after removal Noise enhancement
S14: to the spectrum amplitude value for obtaining leakage signal after result sqrt in S13Believe in conjunction with former leak
Number phase information obtain the spectrum estimation of each frame leakage signalInverse Fourier transform is carried out again, and leakage signal is carried out
Restore and reconstruct, obtains spectrum and subtract enhanced leakage signal;
Further, the calculating of signal spectrum variance is specific as follows in the S2:
S21: assuming that input signal is s (n), the length of every frame is N, and signal is transformed from the time domain to frequency domain meter by DFT
Calculate spectrum value:
Record each frequency component with a matrix | S (ω) | value;
S22: the mean value of each component is calculated:
S23: the Spectral variance value of enhanced leakage signal in previous step is calculated:And
Find out the average value of noise model Spectral variance
Further, the S3 specifically:
S31: setting two threshold values T1 and T2 find out the average value of noise model Spectral variance
S32: threshold value T1 is set asThreshold value T2 is set as
Further, the S4 specifically:
S41: being leakage signal when parameter D is higher than threshold value T2;
S42: the position T1 is higher or lower than by D, to judge the start-stop point of leakage signal;
S43: statistics is higher than the signal frame number of threshold value, if frame number is greater than a quarter of input signal totalframes, sentences
It is set to leakage, and output signal, is otherwise No leakage;
Further, during carrying out time delay estimation using neural network in the S6, process is divided into two stages:
First stage is learning process, and second stage is the course of work;
Further, the learning process are as follows:
1) select the sample data under leakage environment as training sample;
2) sample data is pre-processed, before carrying out neural network prediction, place is normalized to initial data
Reason, makes its data standard between [- 1,1];
3) training sample is constructed, the signal sample sequence after noise suppressed is as object vector under experimental conditions,
The water leakage containing different noises is measured under actual conditions, obtained sample sequence is as input signal;
Further, the course of work are as follows:
A) BP neural network for choosing three-decker establishes prediction model;The corresponding node of input layer, hidden layer, output layer
Number is respectively 1,40 and 1, and it is respectively tansig and purelin function that hidden layer, which exports layer functions,;
B) training network;Before training network, it is also necessary to which training parameter is set, wherein maximum training pace
Net.trainParam.epochs is 10000, and least mean-square error net.trainParam.goal is 0.05, and study speed
Rate is set as 0.01 etc.;
C) after neural network completion, the water leakage collected is carried out in band, out-of-band noise inhibits;
Beneficial effects of the present invention are as follows:
1) judged using double threshold method, illustrate there is leakage, accuracy if Spectral variance is in threshold range
It is higher;
2) more obvious to the promotion of signal-to-noise ratio using " the No leakage estimation technique " estimation noise spectrum;
3) noise reduction process is carried out to the leakage signal that band is made an uproar using spectrum-subtraction, effectively improves the signal-to-noise ratio meeting of water leakage
Facilitate the promotion of leak detection validity;
4) difference of noise and leakage signal spectral characteristic is utilized, the frequency spectrum of noise is evenly distributed in each frequency point
Section, and each frequency component is smaller, the value for calculating variance to its frequency spectrum is also smaller, and the spectral fluctuations of leakage signal are larger, to it
The value for calculating Spectral variance is also big.Therefore, it can effectively improve the accuracy rate of leak detection using threshold value diagnostic method;
5) in groundwater supply environment complicated and changeable, high-precision time delay estimation is not easy to realize.It proposes a kind of based on BP
The leak localization method of neural network.It locates the water leakage under varying environment by the study of neural network in advance
Then reason carries out accurate time delay estimation using generalized correlation method, improves the positioning accuracy of leakage point.
Detailed description of the invention
Fig. 1 is leak location model of the present invention;
Fig. 2 is the algorithm flow chart of leak detection of the present invention;
Fig. 3 is of the present invention based on neural network positioning mode illustraton of model;
Fig. 4 is the overall framework figure of the method for the invention;
Fig. 5 is the original and plus water leakage waveform diagram of making an uproar when the method for the invention is verified;
Fig. 6 be the method for the invention verify when under different signal-to-noise ratio short-term spectrum variogram;
Fig. 7 is that the method for the invention obtains the original and filtered waveform and spectrogram of rush hour;
Fig. 8 is neural metwork training error curve of the present invention;
Fig. 9 is the network training and time delay estimation condition under the conditions of white noise of the present invention;
Figure 10 is the network training and time delay estimation condition under the conditions of coloured noise of the present invention;
Figure 11 is neural metwork training precision under different signal-to-noise ratio of the present invention;
Each generalized related function image that Figure 12 is SNR of the present invention when being 5dB;
Figure 13 be SNR of the present invention be -5dB when each generalized related function image.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is explained in further detail.It should be appreciated that specific embodiment described herein is used only for explaining the present invention, and
It is not used in the restriction present invention.On the contrary, the present invention cover it is any be defined by the claims do on the essence and scope of the present invention
Substitution, modification, equivalent method and scheme.Further, in order to make the public have a better understanding the present invention, below to this
It is detailed to describe some specific detail sections in the datail description of invention.It is thin without these for a person skilled in the art
The present invention can also be understood completely in the description of section part.
The present invention will be further explained below with reference to the attached drawings and specific examples, but not as a limitation of the invention.
Most preferred embodiment is enumerated below for of the invention:
As shown in Fig. 1-Figure 13, the present invention provides a kind of water supply line leakage detecting and locating method, and the method is to sensing
The signal of device acquisition is enhanced, and is then calculated the Spectral variance of signal after enhancing, is judged using double threshold method, if frequency spectrum
Variance, which is in threshold range, then illustrates there is leakage, then carries out leak point location, constitutes filter using BP neural network, will
Water leakage is separated from noise, carries out generalized correlation to water leakage, is carried out according to the weight function that signal-to-noise ratio selects performance good
Time delay estimation, obtain three sensors when delay, calculated using leak location model, obtain leak dot position information,
The described method includes:
S1: water leakage is subjected to spectrum and subtracts enhancing;
S2: the Spectral variance of signal after enhancing in S1 is calculated;
S3: leakage signal is identified using one-parameter double-threshold comparison, and determines threshold value;
S4: leak detection, and output test result are carried out;
S5: to the testing result exported in S4, out-of-band noise is removed by bandpass filtering;
S6: carrying out NN filtering, removes in-band noise;
S7: generalized correlation for time delay estimation is carried out;
S8: leak dot position information is obtained by leak location model using the time delay estimated result in S7, completes leakage
Water positioning.
Described 8 specifically: during selection of weighting function, time delay is carried out according to the weight function that signal-to-noise ratio selects performance good
Estimation, obtain three sensors when delay, calculated using the leak location model, obtain leak dot position information.
The leak location model is three sensor location models, and the model equidistantly distributes sensor in water supplying pipe
Net week is enclosed, wherein the three sensor nodes number nearest apart from leak source is respectively 1,2,3, the distance between they are L, sensing
Position of the device 1,2 apart from leak source is set to d1 and d2, measures the difference of the time delay between sensor 1 and 2,2 and 3 by correlation method
For D12And D23, ν: ν=L/D of spread speed on pipeline where water leakage23, distance and difference of the leak source away from sensor 1 and 2
For
The calculating of spectrum-subtraction in the S1 specifically:
S11: pre-processing input water leakage s (n), and the pretreatment includes preemphasis and adding window framing;
S12: carrying out Short Time Fourier Analysis to the leakage signal that band is made an uproar, and the short-time energy spectrum of each frame signal is calculated
|Ym(ω)|2;
S13: before enhancing using spectrum-subtraction water leakage, noise spectrum estimation is carried out, noise spectrum estimation is obtained
ValueWith | Ym(ω)|2It subtractsTo obtain the leakage signal power spectrum after removal Noise enhancement
S14: to the spectrum amplitude value for obtaining leakage signal after result sqrt in S13Believe in conjunction with former leak
Number phase information obtain the spectrum estimation of each frame leakage signalInverse Fourier transform is carried out again, and leakage signal is carried out
Restore and reconstruct, obtains spectrum and subtract enhanced leakage signal;
The calculating of signal spectrum variance is specific as follows in the S2:
S21: assuming that input signal is s (n), the length of every frame is N, and signal is transformed from the time domain to frequency domain meter by DFT
Calculate spectrum value:
Record each frequency component with a matrix | S (ω) | value;
S22: the mean value of each component is calculated:
S23: the Spectral variance value of enhanced leakage signal in previous step is calculated:And
Find out the average value of noise model Spectral variance
The S3 specifically:
S31: setting two threshold values T1 and T2 find out the average value of noise model Spectral variance
S32: threshold value T1 is set asThreshold value T2 is set as
The S4 specifically:
S41: being leakage signal when parameter D is higher than threshold value T2;
S42: the position T1 is higher or lower than by D, to judge the start-stop point of leakage signal;
S43: statistics is higher than the signal frame number of threshold value, if frame number is greater than a quarter of input signal totalframes, sentences
It is set to leakage, and output signal, is otherwise No leakage.
During being filtered in the S6 using neural network, process is divided into two ranks
Section: first stage is learning process, and second stage is the course of work.
The learning process are as follows:
1) select the sample data under leakage environment as training sample;
2) sample data is pre-processed, before carrying out neural network prediction, place is normalized to initial data
Reason, makes its data standard between [- 1,1];
3) training sample is constructed, the signal sample sequence after noise suppressed is as object vector under experimental conditions,
The water leakage containing different noises is measured under actual conditions, obtained sample sequence is as input signal.
The course of work are as follows:
A) BP neural network for choosing three-decker establishes prediction model;The corresponding node of input layer, hidden layer, output layer
Number is respectively 1,40 and 1, and hidden layer, output layer functions are respectively tansig and purelin function;
B) training network;Before training network, it is also necessary to which training parameter is set, wherein maximum training pace
Net.trainParam.epochs is 10000, and least mean-square error net.trainParam.goal is 0.05, and study speed
Rate is set as 0.01 etc.;
C) after neural network completion, the water leakage collected is carried out in band, out-of-band noise inhibits.
The present invention provides a kind of water supply line leakage detecting and locating method positioning, and the method is based on system model, this is
System model is as shown in Figure 1, sensor is equidistantly distributed around water supply network, wherein three sensors nearest apart from leak source
Node serial number is respectively 1,2,3, and the distance between they are L, and position of the sensor 1,2 apart from leak source is set to d1 and d2,
It is respectively D that the time delay between sensor 1 and 2,2 and 3, which can be measured, by correlation method12And D23, it is possible thereby to first acquire leak letter
Spread speed ν where number on pipeline are as follows:
ν=L/D23 (1)
Then according to the available following relational expression of positional relationship:
Formula (1) is substituted into formula (2) available leak source to be respectively as follows: away from the distance of sensor 1 and 2
Therefore, by three sensor location models, the spread speed of pipe leakage signal can be found out and obtain leak source
The position of two end sensor of distance, to realize that leak source positions.
Leakage signal detection method principle based on spectrum-subtraction and Spectral variance is as shown in Fig. 2, spectrum-subtraction is one kind to letter
Number enhancing denoising efficient algorithm.Spectrum-subtraction is assuming that ambient noise signal is the additive noise of short-term stationarity, original signal
With propose in the incoherent situation of noise.Spectrum-subtraction is small with calculation amount, algorithm is simple, it is easy to accomplish and denoising works well
The characteristics of, application is very extensive.Its effect, which is equivalent to, has carried out certain filtering processing to signals with noise in transform domain, to obtain
More pure signal spectrum.
Signal y with noisem(n) it can indicate are as follows:
ym(n)=sm(n)+dm(n), m=1,2 ...;N=0,1 ..., N-1 (4)
Wherein dmIt (n) is noise signal, smIt (n) is pure leakage signal.M indicates that signal analyzes frame number, and N indicates letter
Number analysis frame length.Fourier transformation is done simultaneously to above formula both ends to obtain:
Ym(ω)=Sm(ω)+Dm(ω) (5)
Wherein, Ym(ω)、Sm(ω) and Dm(ω) respectively corresponds ym(n),sm(n) and dm(n) after carrying out Fourier transformation
Spectral density.It is squared, available YmThe short-time energy of (ω) is composed:
|Ym(ω)|2=| Sm(ω)|2+|Dm(ω)|2+Sm(ω)·Dm(ω)*+Sm(ω)*·Dm(ω)
(6)
It is obtained by above formula:
As it is assumed that sm(n) and dm(n) independent, so Sm(ω) and Dm(ω) is also independent, and assumes Dm(ω) is zero equal
The Gaussian Profile of value, so:
It can obtain:
|Ym(ω)|2=| Sm(ω)|2+|Dm(ω)|2 (9)
Due to stationary noise frequency spectrum before and after leakage it is considered that do not change substantially, in this way can be by leakage before
So-called " quiet section " estimates the energy spectrum of noise | Dm(ω)|2.The purpose of Signal Enhanced Technology based on short-time spectrum amplitude Estimation
Exactly try to obtain | Sm(ω) | estimationAnd thus obtain sm(n) estimationI.e. enhanced leakage signal.
It can obtain:
It can be obtained by enhanced leakage signal in this way:
Define the gain function of m-th of frequency component are as follows:
It can obtain:
Sm(ω)=Gm(ω)·Ym(ω) (13)
That is, to each spectrum component of signals with noise multiplied by a coefficient Gm(w), leakage signal after being denoised
Frequency spectrum.Using original tape make an uproar leakage signal phase spectrum come replace estimation after signal phase spectrum.It is carried out in Fu again
Leaf inverse transformation can be obtained by the leakage signal of denoising.
Actual signal and the spectral characteristic for being mingled in noise signal therein differ greatly.The live part collection of leakage signal
In in this frequency range of 500~3000Hz, and the range value fluctuation of each frequency range is acutely, and signal is steady in a short time
, it and is approximately zero in the spectrum of other frequency ranges that being reflected in frequency domain, which is exactly to concentrate on extremely narrow frequency range,.Therefore to leakage signal meter
Calculate Spectral variance, it will obtain a biggish value.White noise is the noise being most widely present in practical applications, its power
Then relatively flat, the bandwidth of distribution is composed, and spectrum is smaller, variance is calculated to its frequency spectrum, obtained value is smaller, and obvious small
In the Spectral variance value of leakage signal.Therefore, leakage signal and noise signal can be distinguished according to Spectral variance.
Detailed process is as follows for the calculating of signal spectrum variance:
(1) assume that input signal is s (n), the length of every frame is N.Signal is transformed from the time domain into frequency-domain calculations by DFT
Spectrum value:
Record each frequency component with a matrix | S (ω) | value.
(2) mean value of each component is calculated:
(3) Spectral variance value is calculated:
Can be seen that Spectral variance reflection from formula (16) is the fluctuating quantity of each component of signal frequency domain, and frequency spectrum is with frequency
The variation of rate is more violent, then the value of D is bigger.And for white noise signal, spectral ripple is more gentle and the frequency range of distribution is wide.Cause
This Spectral variance is smaller.As a result, echo signal can be detected according to Spectral variance.
BP neural network is one kind with the forward direction based on error back propagation (Error Back Propagation, BP)
Network, Fig. 3 are its topological structure, it is made of input layer, hidden layer, output layer three parts, wherein every layer can include more
A neuron node realizes full connection as individually input between layers, and the neuron between same layer is connectionless.
Neural network structure figure as shown in Figure 3 can then calculate the reality output for obtaining network and desired output difference
Are as follows:
D (n)=[d1,d2,…,dJ] (18)
The then error of nth iteration are as follows:
ej(n)=dj(n)-Yj(n) (19)
So obtaining error energy is defined as:
During error back propagation, according to steepest descent method, error is calculated to the gradient of weight, further along gradient
Opposite direction carry out weighed value adjusting,
According to the chain type of differential rule, the calculation method of gradient is obtained:
Introduce the concept of partial gradient:
For output layer, transmission function is generally linear function, therefore its derivative is denoted as constant k, learning rate η,
Substitute into the correction amount that above formula obtains neuron weight:
The key technology of leak positioning is exactly that the time delay estimation of signal is received between sensor, in general, the two paths of signals time
Delay can carry out peak detection by the cross-correlation function to two paths of signals and estimate.As shown in figure 1, it is assumed that leakage point issued
Signal is S (n), then
Wherein, signal received by sensor 1 and 2 is respectively S1(n) and S2(n), n1(n) and n2(n) believe for noise
Number, α is decay factor, TSFor the sampling period.S1(n) and S2(n) cross-correlation function is defined as:
According to the property of stationary random process auto-correlation function, have
It then obtains, i.e. time delay between sensor 1 and 2 are as follows:
The present invention relates to the leakage signal detection algorithm based on spectrum-subtraction and Spectral variance, algorithm principle is as shown in Figure 2.
Spectrum is subtracted enhancing algorithm to combine with Spectral variance.Noise reduction process is carried out to the leakage signal that band is made an uproar using spectrum-subtraction first,
The signal-to-noise ratio for effectively improving water leakage may consequently contribute to the promotion of leak detection validity.Noise and leakage signal are utilized later
The difference of spectral characteristic, the frequency spectrum of noise is evenly distributed in each frequency segmentation, and each frequency component is smaller, to its spectrometer
The value for calculating variance is also smaller.The Spectral variance of noise signal has notable difference relative to the Spectral variance value of leakage signal.Finally
Leakage signal is accurately identified by the way that suitable threshold value is arranged, and judges leakage.
The realization step of specific algorithm are as follows:
1) input water leakage s (n) is pre-processed, including preemphasis, adding window framing.
2) Short Time Fourier Analysis is carried out to the leakage signal that band is made an uproar.The short-time energy spectrum of each frame signal is calculated |
Ym(ω)|2。
3) before enhancing using spectrum-subtraction water leakage, noise spectrum estimation is carried out first, noise spectrum is obtained and estimates
EvaluationWith | Ym(ω)|2It subtractsTo obtain the leakage signal power spectrum after removal Noise enhancement
4) the spectrum amplitude value of leakage signal is obtained after sqrtIt is obtained in conjunction with the phase information of former water leakage
To the spectrum estimation of each frame leakage signalInverse Fourier transform is carried out again, leakage signal is restored and is reconstructed, and is obtained
Subtract enhanced leakage signal to spectrum.
5) leak detection is carried out using Spectral variance, calculates enhanced each frame signal in previous stepWith noise
The Spectral variance value D of model.And find out the average value of noise model Spectral variance
6) present invention uses one-parameter double-threshold comparison method, only one parameter " frequency during detecting leakage signal
Variance D " is composed, then identifies leakage signal with double threshold.Two threshold values T1 and T2 are set, when parameter D is higher than threshold value T2
When be judged as leakage signal, then when be higher or lower than T1 from D and judge the start-stop point of leakage signal.Wherein, threshold value
T1 is set asThreshold value T2 is set as
7) in leak detection algorithm operational process, statistics is higher than the signal frame number of threshold value, if frame number is greater than input letter
The a quarter of number totalframes, then be judged to leaking, and output signal.It otherwise is No leakage.
It is as shown in Figure 3 to leak location algorithm principle.The general thought of the method is: generalized correlation method carries out time delay estimation
As a result more accurate, however a large amount of priori knowledges and statistical property are needed to determine the use of weight function, this is for underground leak
It is not easy to realize for detection system, therefore combines generalized correlation method to be constituted neural network positioning mode based on BP neural network.
And common generalized correlation function include: Roth processor, SCOT (smooth coherence transfer), PHAT (phse conversion),
This 5 kinds of Eckart, ML (maximum likelihood) are optimized along with 6 kinds of prefilter models can be obtained in traditional correlation method altogether, are being returned
It receives out after common 6 kinds of weight functions, acquires water supply line leakage when noiseless or high s/n ratio in laboratory conditions first
Then water signal acquires the water leakage under different noise situations.Discrete sampling, structure are carried out to the water leakage under different situations
At one-dimensional row vector matrix so as to neural network use.In order to overcome generalized correlation method to need a large amount of this drawback of priori conditions,
Need the learning functionality and adaptivity using neural network.After carrying out many experiments, the BP mind for building three-decker is determined
Through network.
Wherein the number of nodes of input layer is 1, the one-dimensional row vector matrix that input content is made of water leakage sampled point.
Hidden layer selects tansig function as excitation function, and the number of nodes of hidden layer is 40, and the neuron number of output layer is 1.Choosing
Use purline as output function, network establishes sentence are as follows:
Bpnet=newff (minmax (P), [41], ' tansig', ' purelin'}, ' traingdx', '
learngdm') (29)
Exporting content is the discrete signal after NN filtering, and it is anti-that algorithm carries out error using steepest descent method
The available BP neural network with filter function for mixed noise of successive ignition to propagation, by neural network.
Next generalized correlation method is used, suitable weight function is chosen and carries out peak detection, the time delay estimation obtained in this way is more smart
It is quasi-.Finally by formula 3, that is, it can determine the position of leakage point, Fig. 3 show the structural model of neural network positioning mode.Right
5000HZ is low by the 500HZ that the generation high frequency out-of-band noise such as bubble explosion, water impact medium and sensor electric signal are brought into
After out-of-band noise processing, it is also necessary to prevent the noise jamming with water leakage with frequency range.It is just filtered at this time using through band logical
Wave treated data as neural network input and establish BP neural network to carry out leak source positioning.This system uses three layers
The BP neural network of structure, wherein the neuron node number of input layer is 1, and object vector is to believe after earlier data is handled
Number sample sequence, be to obtain under experimental conditions;Its input vector is in actual conditions containing in the case of various unknown noises
Signal sample sequence.The common excitation function of the hidden layer of network includes tansig and logisg, by network training, for
Reach same anticipation error, the training pace of tansig is less than logsig, therefore selects tansig function as sharp here
Function is encouraged, it is 40 that the multiple training test by neural network, which obtains node in hidden layer, and the neuron number of output layer is 1,
Error back propagation is carried out using steepest descent method, trained neural network is obtained by successive ignition, passes through actual conditions
Carry out the feasibility of verification method.
During carrying out time delay estimation using neural network, process is divided into two stages: first stage is to learn
Habit process, second stage are the courses of work, the specific steps are as follows:
1) it chooses sample data and constructs training sample.Since water supply line is buried, in underground, local environment is complicated and changeable,
It is a unstable nonlinear system, it is therefore necessary to select the sample data under normal environment, the unusual sample being otherwise drawn into
This can reduce the predictive ability of network.Sample data is pre-processed.It, be to original number before carrying out neural network prediction
According to being normalized, make its data standard between [- 1,1].
2) training sample is constructed.The signal sample sequence after noise suppressed is as object vector under experimental conditions, so
The water leakage containing different noises is measured in practical situations afterwards, obtained sample sequence is as input signal.
3) BP neural network for choosing three-decker establishes prediction model.The corresponding node of input layer, hidden layer, output layer
Number is respectively 1,40 and 1, and it is respectively tansig and purelin function that hidden layer, which exports layer functions,.
4) training network.Before training network, it is also necessary to which training parameter is set, wherein maximum training pace
Net.trainParam.epochs is 10000, and least mean-square error net.trainParam.goal is 0.05, and study speed
Rate is set as 0.01 etc..
After neural network completion, the water leakage measured is carried out in band, out-of-band noise inhibits.The present invention mentions
Water supply line leak detection location model out is totally divided into two parts: progress leak detection first leaks hair when detecting
After life, followed by the positioning of leakage point.Which kind of noise no matter its overall framework as shown in figure 4, the results showed that use
Estimation mode, spectrum-subtraction can promote the signal-to-noise ratio of water leakage;And use " the No leakage estimation technique " estimation noise spectrum to noise
The promotion of ratio is the most obvious.Next when verifying to Spectral variance algorithm simulating, the Spectral variance of noise model is calculated first,
Double threshold threshold value T1 and T2 are determined, then calculates the Spectral variance that spectrum subtracts each signal frame after enhancing.As shown in figure 5,0.3 second it
It is preceding and 3.6 seconds after be the quiet section for not leaking generation.Three width image as shown in Figure 5 is non-superimposed noise, signal-to-noise ratio respectively
It is -5dB, the short-term spectrum variance image being calculated under three kinds of background noise conditions for 5dB and signal-to-noise ratio.Cross is parallel in figure
The dotted line of axis is the image of threshold value T1, and solid line is the image of threshold value T2.The solid marks for being parallel to the longitudinal axis are opening for leakage signal
Beginning, dashed lines labeled are leakage signal end.By image as it can be seen that the signal detecting method based on spectrum-subtraction and Spectral variance not
Only there is good separating capacity in high s/n ratio, still there is powerful performance in low signal-to-noise ratio.Even if believing
When making an uproar than for -5dB, only occurs primary erroneous judgement at 2.7 seconds, but do not influence the validity of overall algorithm, leakage signal section and back
Still difference is obvious for the waveform of scape noise segment.
Fig. 6 show the signal amplitude after normalization.Calculate the Spectral variance of signal shown in Fig. 6, the results showed that, it is inputting
10 gradients that Signal-to-Noise changes from -5dB to 5dB, the Spectral variance value maximum of noise model is no more than 30, and enhances
The Spectral variance value of water leakage is 680 or more afterwards.The two difference is obvious, can effectively be detected by the way that suitable threshold value is arranged
Leakage signal out.
In conclusion either still calculating data from emulating image reflects letting out based on spectrum-subtraction and Spectral variance
Leakage signal detection method detection effect is obvious, and validity is strong.After completing leak detection, it is accurately fixed to need to carry out leak
Position.As shown in Figure 7 is respectively the waveform of the water leakage measured rush hour and frequency spectrum and the feelings after bandpass filtering
Condition.In the water leakage for obtaining rush hour, out-of-band noise can reduce by bandpass filtering, it is right after handling in this way
Obtained signal is sampled, and the matrix that sampled point is constituted is as input vector, when night noise is smaller to the letter measured
It number is sampled, the matrix of composition is trained neural network, training is as shown in Figure 8 as object vector.Obviously,
0.64% error precision is reached by 9991 training, next using another section of water leakage of rush hour as survey
Examination input, output obtain the signal filtered by BP neural network.By identical step process sensor 1,2 in rush hour
Water leakage, time delay next acquired using generalized correlation method.
First have to selection weight function during carrying out pre-filtering, common weight function include: it is substantially related, at Roth
Reason, the smooth coherence transfer of SCOT, PHAT phse conversion, Eckart, this 6 kinds of ML maximum likelihood weight function.When selecting weight function,
It is ensured that have a spike in cross-correlation function rather than a broad peak, and have between high-resolution and stability one it is simultaneous
It cares for.It is next just different respectively after having carried out feasibility analysis to BP neural network and generalized correlation in terms of leak positioning
Noise type, signal-to-noise ratio, the weighting function of generalized correlation method and neural network parameter these four variables to arithmetic accuracy
Influence is analyzed, and then obtains full experiment conclusion.Influence of the different noise types to algorithm.Guaranteeing that its dependent variable is identical
In the case where, successively mixed signal noise be white Gaussian noise, coloured noise both carry out arithmetic results comparison.First
It is mixed into white noise, training error such as Fig. 9 (a) is shown, then the signal of the sensor 1 and 2 after NN filtering is carried out
Generalized correlation, shown in obtained cross-correlation function image such as Fig. 9 (b).Next frequency is mixed into make an uproar in the coloured of 500-2000HZ
It is equivalent to increase in-band noise after sound, the broad sense cross-correlation function of neural metwork training result and two sensors is respectively such as Figure 10
In (a) and 10 (b) shown in.
In the case where being mixed into white noise, neural metwork training precision reaches anticipation error 0.001 through 1599 steps, and is mixed into
The training precision of coloured noise only reaches 0.0024 through 10000 steps, illustrates that the network training precision containing white noise is better than containing coloured
The case where noise.
Influence of the different signal-to-noise ratio to arithmetic accuracy, for being mixed into white Gaussian noise, in the range of 10dB to -12dB
Change signal-to-noise ratio, let the signal go through BP neural network, the error curve wherein obtained under 10dB and -12dB is as shown in figure 11, will
Error result under multiple signal-to-noise ratio is arranged into following table, it can be found that signal-to-noise ratio is higher, the training result of neural network is unreasonable
Think, it is more accurate using the positioning of generalized correlation.
Following table is the network training result under different signal-to-noise ratio
Algorithm parameter setting itself can also impact positioning accuracy.The setting of BP algorithm parameter mainly includes hidden layer
These aspects of number of nodes, transfer function, training method.Research object is set to the leak that signal-to-noise ratio containing white Gaussian noise is 10dB and believes
Number, it is trained in the network containing different node in hidden layer first, if maximum step-length is 10000 steps, changes hidden layer
Number of nodes and carry out repeatedly training and obtain the mean value of training precision, when training sample is 5000, training result is as shown in the table,
The optimal number of nodes of available hidden layer is 17, and training error precision is minimum at this time and is 0.00234, the results showed that hidden layer
Number of nodes has an impact on error precision but influences smaller, and the increase of number of nodes influences operation time and operand bigger.
Following table is the training precision table of comparisons of hidden layer node and neural network
And for transfer function, the common transfer function of hidden layer is logsig and tansig, for the same letter
Number, in identical maximum step-length, the training precision of tansig is higher than logsig, so selecting tansig as hidden layer
Transfer function.The weight function of generalized correlation method also has an impact to arithmetic accuracy, and common weight function has 5 kinds, along with traditional phase
Pass method (CC), altogether 6 kinds of processing methods.Different methods is respectively adopted, generalized correlation processing is carried out to same water leakage, when
The Signal to Noise Ratio (SNR) 1 of sensor 1 is 10dB, when the Signal to Noise Ratio (SNR) 2 of sensor 2 is 5dB, different generalized related function images
As shown in figure 12.
When the Signal to Noise Ratio (SNR) 1 of sensor 1 is 10dB, and the Signal to Noise Ratio (SNR) 2 of sensor 2 is -5dB, different broad sense
Correlation function image is as shown in figure 13, it can be seen that as SNR1=10dB and SNR2=-5dB, with the decline of signal-to-noise ratio, before
Correlation function peak value more sharp SCOT and ROTH has been submerged in noise, and the peak value of CC, PHAT also become flat therewith
And the influence vulnerable to secondary peak.The peak value sharpness for reviewing ML and Eckart has significant change there is no the decline with signal-to-noise ratio
Change.In conclusion when selecting generalized related function to carry out leak positioning, when the noise that sensor receives signal is relatively high
When, select the locating effect of PHAT and SCOT weight function preferable;When its noise is relatively low, ML and Eckart weight function is selected
Locating effect is preferable.
Embodiment described above, only one kind of the present invention more preferably specific embodiment, those skilled in the art
The usual variations and alternatives that member carries out within the scope of technical solution of the present invention should be all included within the scope of the present invention.
Claims (11)
1. a kind of water supply line leakage detecting and locating method, which is characterized in that the method carries out the signal that sensor acquires
Then enhancing is calculated the Spectral variance of signal after enhancing, is judged using double threshold method, if Spectral variance is in threshold range
Inside then illustrate there is leakage, then carry out leak point location, filter is constituted using BP neural network, by water leakage from noise
Separation carries out generalized correlation to water leakage, according to the weight function that signal-to-noise ratio selects performance good, carries out time delay estimation, obtains three
A sensor when delay, calculated using leak location model, obtain leak dot position information.
2. the method according to claim 1, wherein the described method includes:
S1: water leakage is subjected to spectrum and subtracts enhancing;
S2: the Spectral variance of signal after enhancing in S1 is calculated;
S3: leakage signal is identified using one-parameter double-threshold comparison, and determines threshold value;
S4: leak detection, and output test result are carried out;
S5: to the testing result exported in S4, out-of-band noise is removed by bandpass filtering;
S6: carrying out NN filtering, removes in-band noise;
S7: generalized correlation for time delay estimation is carried out;
S8: leak dot position information is obtained by leak location model using the time delay estimated result in S7, it is fixed to complete leak
Position.
3. according to the method described in claim 2, it is characterized in that, the S7 specifically: during selection of weighting function, need
To select the good weight function of performance to carry out time delay estimation according to signal-to-noise ratio, obtain three sensors when delay, utilize the leakage
Water location model is calculated, and leak dot position information is obtained.
4. according to the method described in claim 3, it is characterized in that, the leak location model be three sensor location models,
The model equidistantly distributes sensor around water supply network, wherein the three sensor nodes number nearest apart from leak source
Respectively 1,2,3, the distance between they are L, and position of the sensor 1,2 apart from leak source is set to d1 and d2, passes through correlation
It is respectively D that method, which measures the time delay between sensor 1 and 2,2 and 3,12And D23, spread speed ν: ν on pipeline where water leakage
=L/D23, distance of the leak source away from sensor 1 and 2 and it is respectivelyWith
5. according to the method described in claim 2, it is characterized in that, the S1 specifically:
S11: pre-processing input water leakage s (n), and the pretreatment includes preemphasis and adding window framing;
S12: carrying out Short Time Fourier Analysis to the leakage signal that band is made an uproar, and the short-time energy spectrum of each frame signal is calculated | Ym
(ω)|2;
S13: before enhancing using spectrum-subtraction water leakage, noise spectrum estimation is carried out, noise spectrum estimation value is obtainedWith | Ym(ω)|2It subtractsTo obtain the leakage signal power spectrum after removal Noise enhancement
S14: to the spectrum amplitude value for obtaining leakage signal after result sqrt in S13In conjunction with the phase of former water leakage
Position information obtains the spectrum estimation of each frame leakage signalCarry out inverse Fourier transform again, to leakage signal carry out restore and
Reconstruct obtains spectrum and subtracts enhanced leakage signal.
6. according to the method described in claim 2, it is characterized in that, in the S2 calculating of signal spectrum variance it is specific as follows:
S21: assuming that input signal is s (n), the length of every frame is N, and signal is transformed from the time domain to frequency-domain calculations frequency by DFT
Spectrum:
Record each frequency component with a matrix | S (ω) | value;
S22: the mean value of each component is calculated:
S23: the Spectral variance value of enhanced leakage signal in previous step is calculated:And it finds out
The average value of noise model Spectral variance
7. according to the method described in claim 6, it is characterized in that, the S3 specifically:
S31: setting two threshold values T1 and T2 find out the average value of noise model Spectral variance
S32: threshold value T1 is set asThreshold value T2 is set as
8. the method according to the description of claim 7 is characterized in that the S4 specifically:
S41: when parameter D is higher than threshold value T2, i.e., Spectral variance is higher than higher threshold value, and Spectral variance is in threshold range
Outside, illustrate that pipeline has occurred that leakage;
S42: the position T1 is higher or lower than by D, to judge the start-stop point of leakage signal;
S43: statistics is higher than the signal frame number of threshold value T1, if frame number is greater than a quarter of input signal totalframes, determines
This signal is leaks, and output signal, is otherwise No leakage.
9. according to the method described in claim 2, it is characterized in that, the process being filtered in the S6 using neural network
In, process is divided into two stages: first stage is learning process, and second stage is the course of work.
10. according to the method described in claim 9, it is characterized in that, the learning process are as follows:
1) select the sample data under leakage environment as training sample;
2) sample data is pre-processed, before carrying out neural network prediction, initial data is normalized, is made
Its data standard is between [- 1,1];
3) training sample is constructed, the signal sample sequence after noise suppressed is as object vector under experimental conditions, in reality
In the case of measure the water leakage containing different noises, obtained sample sequence is as input signal.
11. according to the method described in claim 9, it is characterized in that, the course of work are as follows:
A) BP neural network for choosing three-decker establishes prediction model;The corresponding interstitial content of input layer, hidden layer, output layer
Respectively 1,40 and 1, hidden layer, output layer functions are respectively tansig and purelin function;
B) training network;Before training network, it is also necessary to which training parameter is set, wherein maximum training pace
Net.trainParam.epochs is 10000, and least mean-square error net.trainParam.goal is 0.05, and study speed
Rate is set as 0.01;
C) after neural network completion, the water leakage collected is carried out in band, out-of-band noise inhibits.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710139772.7A CN107061996B (en) | 2017-03-10 | 2017-03-10 | A kind of water supply line leakage detecting and locating method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710139772.7A CN107061996B (en) | 2017-03-10 | 2017-03-10 | A kind of water supply line leakage detecting and locating method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107061996A CN107061996A (en) | 2017-08-18 |
CN107061996B true CN107061996B (en) | 2019-04-16 |
Family
ID=59621679
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710139772.7A Active CN107061996B (en) | 2017-03-10 | 2017-03-10 | A kind of water supply line leakage detecting and locating method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107061996B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110159868A (en) * | 2019-05-27 | 2019-08-23 | 北京奥蓝仕技术有限公司 | Data processing method, device and the SMART SLEEVE of SMART SLEEVE |
Families Citing this family (27)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107869653B (en) * | 2017-12-14 | 2019-12-06 | 北京化工大学 | Pipeline flow sensitivity matrix leakage detection method |
CN108286655B (en) * | 2017-12-28 | 2024-03-15 | 天津工业大学 | Intelligent water flow photoelectric detection and water leakage judging method and device |
CN110136210B (en) * | 2018-02-07 | 2023-06-09 | 中国石油化工股份有限公司 | Intelligent early warning system for leakage of monitoring device |
CN108443714B (en) * | 2018-04-09 | 2019-11-15 | 山东省科学院激光研究所 | Determine the method and device of pipe leakage |
CN108758354B (en) * | 2018-05-03 | 2023-09-12 | 太原理工大学 | Heat supply pipeline leakage detection system and method based on infrasonic wave and reference point |
CN108758356B (en) * | 2018-05-20 | 2019-08-23 | 东北石油大学 | A kind of information processing method and system |
CN109443669B (en) * | 2018-09-30 | 2021-03-26 | 深圳市英威腾电源有限公司 | Water leakage detection method and device |
CN109555977A (en) * | 2018-11-23 | 2019-04-02 | 水联网技术服务中心(北京)有限公司 | The equipment and recognition methods of leak noise measuring |
CN109555979B (en) * | 2018-12-10 | 2020-03-17 | 清华大学 | Water supply pipe network leakage monitoring method |
CN109451529A (en) * | 2019-01-02 | 2019-03-08 | 成都华日通讯技术有限公司 | A kind of miniature spectral sensor net system and its synergistic data processing method |
CN109681789B (en) * | 2019-03-01 | 2020-07-03 | 重庆邮电大学 | Fluid pipeline leakage acoustic vibration coherent detection positioning method based on variational modal decomposition |
CN110195823A (en) * | 2019-07-16 | 2019-09-03 | 重庆科技学院 | A kind of micro- leakage detection method of pipeline based on deep neural network |
CN110261821B (en) * | 2019-07-18 | 2021-12-10 | 中电科思仪科技股份有限公司 | Time difference estimation method and device based on information demodulation correlation |
CN110529745B (en) * | 2019-08-05 | 2021-03-12 | 上海第二工业大学 | Algorithm for detecting pipeline leakage and position by optical fiber in frequency domain |
CN110953485B (en) * | 2019-12-06 | 2021-03-19 | 北京无线电计量测试研究所 | Gas pipeline leakage point positioning method and system |
CN111425768B (en) * | 2020-03-31 | 2022-04-26 | 长安大学 | Oil leakage point and oil leakage rate detection method for oil pipeline |
CN111365625A (en) * | 2020-04-13 | 2020-07-03 | 安徽理工大学 | Pipeline leakage sound wave positioning method based on delay summation |
CN111350948A (en) * | 2020-04-13 | 2020-06-30 | 安徽理工大学 | Pipeline leakage position calculation method based on beam forming |
CN112013285B (en) * | 2020-08-26 | 2022-04-19 | 辽宁石油化工大学 | Method and device for detecting pipeline leakage point, storage medium and terminal |
CN111895278A (en) * | 2020-09-03 | 2020-11-06 | 武汉轻工大学 | Urban water supply pipe network leakage point positioning method, device, equipment and storage medium |
CN112992178A (en) * | 2021-02-05 | 2021-06-18 | 安徽绿舟科技有限公司 | Method for analyzing pipeline leakage through chaotic voiceprint |
CN113780381B (en) * | 2021-08-28 | 2022-07-01 | 特斯联科技集团有限公司 | Artificial intelligence water leakage detection method and device |
CN114001880B (en) * | 2021-10-29 | 2024-05-10 | 清华大学合肥公共安全研究院 | Water supply pipeline leakage amount assessment method based on in-pipe leakage acoustic wave analysis |
CN114738679B (en) * | 2022-04-11 | 2023-09-22 | 长江大学 | Pipeline missing detection device and pipeline missing detection method |
CN116293486A (en) * | 2023-03-15 | 2023-06-23 | 东莞先知大数据有限公司 | Water pipe leakage detection method and device, electronic equipment and storage medium |
CN116464918B (en) * | 2023-05-06 | 2023-10-10 | 江苏省特种设备安全监督检验研究院 | Pipeline leakage detection method, system and storage medium |
CN116577037B (en) * | 2023-07-12 | 2023-09-12 | 上海电机学院 | Air duct leakage signal detection method based on non-uniform frequency spectrogram |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0714019B1 (en) * | 1994-11-22 | 1999-09-22 | Gaz De France (Service National) | Procedure and device for the detection and localisation of leaks of fluid flowing in a pipe |
DE202008017243U1 (en) * | 2008-05-02 | 2009-09-17 | Rbs Wave Gmbh | System for monitoring and / or locating water losses in a water pipeline network |
CN101551064A (en) * | 2009-05-22 | 2009-10-07 | 重庆大学 | Water supply pipe leakage detection locating signal processing method |
CN102193082A (en) * | 2011-03-18 | 2011-09-21 | 哈尔滨工业大学 | Device for positioning leak source of three-sensor multi-scale constrained pipe network |
CN104061445A (en) * | 2014-07-09 | 2014-09-24 | 中国石油大学(华东) | Pipeline leakage detection method based on neural network |
-
2017
- 2017-03-10 CN CN201710139772.7A patent/CN107061996B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0714019B1 (en) * | 1994-11-22 | 1999-09-22 | Gaz De France (Service National) | Procedure and device for the detection and localisation of leaks of fluid flowing in a pipe |
DE202008017243U1 (en) * | 2008-05-02 | 2009-09-17 | Rbs Wave Gmbh | System for monitoring and / or locating water losses in a water pipeline network |
CN101551064A (en) * | 2009-05-22 | 2009-10-07 | 重庆大学 | Water supply pipe leakage detection locating signal processing method |
CN102193082A (en) * | 2011-03-18 | 2011-09-21 | 哈尔滨工业大学 | Device for positioning leak source of three-sensor multi-scale constrained pipe network |
CN104061445A (en) * | 2014-07-09 | 2014-09-24 | 中国石油大学(华东) | Pipeline leakage detection method based on neural network |
Non-Patent Citations (2)
Title |
---|
三传感器约束多尺度相关管道漏点定位方法;张建利等;《哈尔滨工业大学学报》;20120630;第44卷(第6期);第37-40页 |
基于频谱方差和谱减法的语音端点检测新算法;王路露等;《计算机工程与应用》;20131210;第194-197页 |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110159868A (en) * | 2019-05-27 | 2019-08-23 | 北京奥蓝仕技术有限公司 | Data processing method, device and the SMART SLEEVE of SMART SLEEVE |
Also Published As
Publication number | Publication date |
---|---|
CN107061996A (en) | 2017-08-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107061996B (en) | A kind of water supply line leakage detecting and locating method | |
CN103440871B (en) | A kind of method that in voice, transient noise suppresses | |
CN110456332B (en) | Underwater acoustic signal enhancement method based on automatic encoder | |
CN106328150A (en) | Bowel sound detection method, device and system under noisy environment | |
CN103854662A (en) | Self-adaptation voice detection method based on multi-domain joint estimation | |
CN104836759B (en) | Sane underwater acoustic channel method of estimation based on joint norm optimization | |
CN113378661A (en) | Direct current electric energy signal denoising method based on improved wavelet threshold and related detection | |
Yao et al. | An adaptive seismic signal denoising method based on variational mode decomposition | |
CN113472390B (en) | Frequency hopping signal parameter estimation method based on deep learning | |
CN105510968A (en) | Seismic oceanography-based seawater physical property measuring method | |
CN112086100B (en) | Quantization error entropy based urban noise identification method of multilayer random neural network | |
CN113642484A (en) | Magnetotelluric signal noise suppression method and system based on BP neural network | |
CN115758082A (en) | Fault diagnosis method for rail transit transformer | |
CN117609702A (en) | Pipeline leakage acoustic emission signal denoising method, system, equipment and medium | |
CN115410589A (en) | Attention generation confrontation voice enhancement method based on joint perception loss | |
Xu et al. | U-former: Improving monaural speech enhancement with multi-head self and cross attention | |
Tian et al. | Underwater multi-target passive detection based on transient signals using adaptive empirical mode decomposition | |
Guo et al. | LFM signal optimization time-fractional-frequency analysis: Principles, method and application | |
CN110146923A (en) | A kind of efficient high accuracy depth domain methods of seismic wavelet extraction | |
CN117574062A (en) | Small loop transient electromagnetic signal denoising method based on VMD-DNN model | |
CN107635181A (en) | A kind of multiple access based on channel study perceives the feedback optimized method in source | |
Unoki et al. | A method based on the MTF concept for dereverberating the power envelope from the reverberant signal | |
CN113822363B (en) | Convolutional neural network time delay estimation method based on quadratic rational kernel function | |
Zhao et al. | Pipeline leak fault feature extraction based on wavelet packet analysis and application | |
Yang et al. | Seismic random noise attenuation via a two-stage U-net with supervised attention |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |