CN109931506A - Pipeline leakage detection method and device - Google Patents
Pipeline leakage detection method and device Download PDFInfo
- Publication number
- CN109931506A CN109931506A CN201910192757.8A CN201910192757A CN109931506A CN 109931506 A CN109931506 A CN 109931506A CN 201910192757 A CN201910192757 A CN 201910192757A CN 109931506 A CN109931506 A CN 109931506A
- Authority
- CN
- China
- Prior art keywords
- time
- leakage
- mixed signal
- frequency
- convolutional neural
- 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
Abstract
The embodiment of the present invention provides a kind of pipeline leakage detection method and device, this method comprises: carrying out time frequency analysis to pipeline mixed signal based on Short Time Fourier Transform, obtains the corresponding time-frequency spectrum of the mixed signal；The time-frequency spectrum is converted into two-dimensional time-frequency image；The time-frequency image is input to preset convolutional neural networks model, according to the output of the convolutional neural networks model as a result, obtaining leakage type corresponding with the mixed signal.Time frequency analysis is carried out to pipeline mixed signal by Short Time Fourier Transform, the corresponding time-frequency spectrum of mixed signal is obtained, without extracting leakage signal from mixed signal, to be not easily susceptible to the interference of mixed noise.In addition, the time-frequency spectrum is converted into two-dimentional time-frequency image, is input to preset convolutional neural networks, preset neural network model after the mixed signal sample training with leakage type according to obtaining, the recognition result that leakage type can be exported, so that detection process is quick and accurate.
Description
Technical field
The present embodiments relate to signal processing technology field more particularly to a kind of pipeline leakage detection method and device.
Background technique
Water supply line plays extremely crucial effect in the water resource transport of long range, but since most of pipeline is laid with
In the inferior extreme natural environment in field or ground, it is easy to be damaged.Once water supply line is damaged and leaks, need
The maintenance of pipeline is carried out in time, in order to avoid bring additional economic loss.Therefore, the quick detection of pipe leakage, which has, greatly answers
With value, such as scheme guidance can be provided for pipeline maintenance.
Currently, for the signal detection of the pipe leakage in water supply network, it is most of to use time frequency analysis and signal processing
Means extract the information in relation to leakage signal from the mixed signal that pipeline obtains.However, due in grid noise it is relatively low,
Mixed signal is often mixed with spuious interference signal and noise, causes leakage signal extremely faint, while leakage signal may be only
There are in Mr. Yu's section time range, lead to not completely remove very noisy and interference signal from mixed signal and obtain to be detected
Leakage signal.In addition, there are huge offices in nonstationary random response for traditional time-frequency distributions when signal-to-noise ratio changes greatly
Sex-limited, the signal after reconstructing, which remains unchanged, not can be removed the influence and interference of mixed noise, cause the leakage signal extracted corresponding
Detection classification results it is unreliable.To sum up, current pipeline leakage detection method, due to by noise and interference effect, it is difficult to mention
Accurate leakage signal is taken, it is unreliable so as to cause testing result.
Summary of the invention
To solve the above-mentioned problems, the embodiment of the present invention provides a kind of pipeline leakage detection method and device.
In a first aspect, the embodiment of the present invention provides a kind of pipeline leakage detection method, comprising: be based on Short Time Fourier Transform
(short-time Fourier transform, abbreviation STFT) carries out time frequency analysis to pipeline mixed signal, obtains described mixed
Close the corresponding time-frequency spectrum of signal；The time-frequency spectrum is converted into two-dimensional time-frequency image；The time-frequency image is input to default
Convolutional neural networks model (Convolutional Neural Networks, abbreviation CNN), according to the convolutional Neural net
The output of network model is as a result, obtain leakage type corresponding with the mixed signal；It wherein, include leakage in the mixed signal
Signal, the convolutional neural networks model according to leakage type label sample mixed signal sample training after obtain, institute
Stating leakage type includes without leakage, osculum leakage, the leakage of middle mouth and big mouth leakage.
Second aspect, the embodiment of the present invention provide a kind of pipeline leakage detection device, comprising: time frequency analysis module is used for
Time frequency analysis is carried out to pipeline mixed signal based on Short Time Fourier Transform, obtains the corresponding time-frequency spectrum of the mixed signal；Figure
As generation module, for the time-frequency spectrum to be converted into two-dimensional time-frequency image；Type detection module is used for the time-frequency figure
As being input to preset convolutional neural networks model, according to the output of the convolutional neural networks model as a result, obtain with it is described
The corresponding leakage type of mixed signal；It wherein, include leakage signal, the convolutional neural networks model root in the mixed signal
According to leakage type label mixed signal sample training after obtain, the leakage type include without leakage, osculum leak, in
Mouth leakage and the leakage of big mouth.
The third aspect, the embodiment of the present invention provides a kind of electronic equipment, including memory, processor and is stored in memory
Computer program that is upper and can running on a processor, processor realize the inspection of first aspect present invention pipeline leakage when executing program
The step of survey method.
Fourth aspect, the embodiment of the present invention provide a kind of non-transient computer readable storage medium, are stored thereon with calculating
Machine program, when which is executed by processor the step of realization first aspect present invention pipeline leakage detection method.
Pipeline leakage detection method and device provided in an embodiment of the present invention, due to passing through Short Time Fourier Transform to pipeline
Mixed signal carries out time frequency analysis, obtains the corresponding time-frequency spectrum of mixed signal, without extracting leakage signal from mixed signal,
To be not easily susceptible to the interference of mixed noise.In addition, the time-frequency spectrum is converted into two-dimentional time-frequency image, it is input to preset convolution
Neural network model, preset neural network model according to leakage type mixed signal sample training after obtain, can
The recognition result of output leakage type, so that detection process is quick and accurate.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is this hair
Bright some embodiments for those of ordinary skill in the art without creative efforts, can be with root
Other attached drawings are obtained according to these attached drawings.
Fig. 1 is pipeline leakage detection method flow chart provided in an embodiment of the present invention；
Fig. 2 is pipeline leakage structure diagram of detection device provided in an embodiment of the present invention；
Fig. 3 is the entity structure schematic diagram of a kind of electronic equipment provided in an embodiment of the present invention.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art
Every other embodiment obtained without creative efforts, shall fall within the protection scope of the present invention.
The accurate detection of water supply line leakage has great application value, can provide scheme for water supply line maintenance and refer to
It leads.But current pipeline leakage detection method, due to by noise and interference effect, it is difficult to accurate leakage signal is extracted, from
And cause testing result unreliable.
To solve this problem, the embodiment of the present invention provides a kind of pipeline leakage detection method.This method can be applied to
The scene for stating water supply line leak detection can be used for the leak detection scene of other pipelines, such as petroleum, natural gas
The leak detection of conveying, the present invention is not especially limit this, carries out by taking the application scenarios of water supply line as an example below
Explanation.In addition, this method can be realized by the leak detection apparatus of dedicated setting as executing subject, it can also be by existing
Detection device and computer in conjunction with realizing, the embodiment of the present invention is also not especially limited this.For ease of description, this hair
Bright embodiment carries out pipeline leakage detection method provided in an embodiment of the present invention so that executing subject is leak detection apparatus as an example
It illustrates.
Fig. 1 is pipeline leakage detection method flow chart provided in an embodiment of the present invention, as shown in Figure 1, the embodiment of the present invention
A kind of pipeline leakage detection method is provided, comprising:
101, time frequency analysis is carried out to pipeline mixed signal based on Short Time Fourier Transform, when acquisition mixed signal is corresponding
Frequency spectrum.
In 101, in practical leak detection, since pipeline environment is more complicated, leak detection apparatus actual acquisition is arrived
Signal be the mixed signal for including leakage signal and random disturbances noise signal, the additivity group both being generally regarded as
It closes, it may be assumed that
X (t)=s (t)+n (t) (1)
Wherein, n (t) is random disturbances noise, and s (t) is leakage signal.Mixed signal is leakage signal and noise signal
Combination, belongs to nonstationary random signal.
The leakage signal of water supply line is one kind of acoustic emission signal.Since there are pressure differences inside and outside pipeline, when somewhere is sent out
When raw leak, managing interior high pressure water can spray outward from perforation or crack, the water and leakage hole and surrounding soil of ejection passage
Equal media occur to hit friction, can generate the vibration signal of different frequency.In combination with leak point to signal acquisition point, (acquisition is let out
Reveal signal location point) distance, power leakage signal model can be built.For example, the leakage signal model established is as follows, below by way of
It is illustrated for this signal model:
Wherein, aiAnd wiI-th of amplitude and frequency of oscillation are respectively indicated, R indicates leakage point to the distance of collector, and V is to let out
The spread speed of leakage signal in the duct.
Time frequency analysis is carried out to the pipeline mixed signal comprising leakage signal, that is, uses Short Time Fourier Transform, is obtained short
Time-frequency spectrum:
The STFT of mixed signal is defined as:
Above-mentioned leakage signal combination interfering noise signal is obtained into mixed signal x (t)=s (t)+n (t) and substitutes into above formula, is adopted
STFT is carried out with non-point-by-point stepping sliding window mode, obtains the STFT of x (t) are as follows:
The time-frequency spectrum of mixed signal is the mould square of STFT, is denoted as Hx(t, f) has:
Wherein, h () is a Gauss function with minimum Timed automata, is h (t)=exp (- t2/2σ2)。
The final result of STFT is a function about time and frequency, and compared with traditional FFT, STFT not only has traditional FFT
Spectral characteristic, and by the movement of window function h (), available signal x (t) local spectrum near point t in the specific time
Feature can retain original signal x (t) information as much as possible in this way, be able to ascend signal-to-noise ratio to a certain extent.
102, time-frequency spectrum is converted into two-dimensional time-frequency image.
In 102, the time-frequency spectrum of the mixed signal that above-mentioned 101 step obtains is the function about time domain and frequency domain, can be
It is shown in two-dimensional coordinate with time domain coordinate and frequency domain coordinates, so that the abscissa ordinate with two dimensional image is corresponding, and will
Time-frequency spectrum is converted into the two-dimensional time-frequency image that convolutional neural networks model can identify.
103, time-frequency image is input to preset convolutional neural networks model, according to the output of convolutional neural networks model
As a result, obtaining leakage type corresponding with mixed signal.
In 103, preset convolutional neural networks model is by obtaining after the training of sample mixed signal.Sample mixing
Signal is to have learned that it reveals the mixed signal of type in advance, and mix using corresponding known leaks type as each sample
The label of signal.After establishing convolutional neural networks model, it is trained by a large amount of such sample mixed signal, to obtain
Preset convolutional neural networks model, for subsequently received mixed signal to be detected, the correspondence time-frequency image that will acquire is defeated
Enter to preset convolutional neural networks model, can quick and precisely obtain the output result for revealing type accordingly.Correspondingly, it uses
In the acquisition methods of the time-frequency image of trained mixed signal sample, using the same procedure in step 101.By utilizing depth
CNN network model in study realizes the high-rise attributive character constant from low-level Feature Extraction Abstract, realizes the non-of complexity
Linear function approaches, and features the essential information of leakage signal, to guarantee the accuracy of the leakage type obtained.
Leakage type label can be defined according to demand, for example, there is a leakage, No leakage, or for No leakage, small-sized let out
Dew, medium-sized leakage and large-scale leakage.Allotment of the type for man power and material in subsequent maintenance process is revealed, pass can be played
The directive function of key.
Pipeline leakage detection method provided in an embodiment of the present invention is believed due to being mixed by Short Time Fourier Transform to pipeline
Number time frequency analysis is carried out, the corresponding time-frequency spectrum of mixed signal is obtained, without extracting leakage signal from mixed signal, thus not
Interference vulnerable to mixed noise.In addition, the time-frequency spectrum is converted into two-dimentional time-frequency image, it is input to preset convolutional Neural net
Network model, preset neural network model according to leakage type mixed signal sample training after obtain, can export and let out
The recognition result for revealing type, so that detection process is quick and accurate.
Content based on the above embodiment, as a kind of alternative embodiment, the embodiment of the present invention is not converted to by time-frequency spectrum
Make specific restriction at the method for two-dimensional time-frequency image, including but not limited to: frequency division when corresponding according to mixed signal time-frequency spectrum
Cloth, the energy intensity of each T/F point and different gray values is corresponding, obtain the two-dimentional time-frequency with different gray values
Image.
The time-frequency spectrum of mixed signal, can correspond to the two-dimensional coordinate of T/F, and each point has corresponding energy strong
Degree.A gray value (RGB) can be correspondingly generated according to the energy intensity of the m- Frequency point of each clock synchronization.With the figure of 8bit
For piece, gray value corresponding to maximum energy value is 255, and the corresponding gray value of minimum energy value is 0, and other energy values are pressed
Size carries out the division of ladder section, respectively corresponds 0-255.Time frequency point of the mixed signal on time-frequency coordinate plane corresponds to time-frequency
Pixel in gray level image so as to handle the time-frequency image of mixed signal as gray level image, and is had
There is the two-dimentional time-frequency image of different gray values, remembers that two-dimentional time-frequency gray level image is Γ.It, can be affected to image before inputting neural network
Normalized, by the gray value of all pixels point divided by 255, decimally come indicate normalization after data as initially two
Dimensional feature vector.
Specifically, in order to quickly calculate the STFT of data slot in short time-window, window function length is typically selected to 2 power
It is secondary, it is assumed that Gaussian window length is 2P, sliding window stepping is Q, and the sampling number of x (t) is Nx, PxFor the energy intensity of corresponding points, Γ's
Line number, columns are respectively M, N, and the sample rate for receiving the receiver of mixed signal is fs, the row, column resolution ratio of time-frequency image, i.e.,
The resolutions of the time domain and the frequency domain is Δ t, Δ f, then can be obtained by the following steps two-dimentional time-frequency image:
Γ=fix((Nx-(2P-Q))/Q)； (6)
N=2P-1+1； (7)
Δ t=Q/fs, Δ f=fs/2P (8)
Γ (m, n)=Px((m-1) Δ t, (n-1) Δ f), m=1,2 ..., M；N=1,2 ..., N. (9)
Pipeline leakage detection method provided in an embodiment of the present invention, by the energy intensity of each T/F point from it is different
Gray value is corresponding, obtains the two-dimentional time-frequency image with different gray values, and two-dimentional time-frequency image accurate response is enable to mix
The information of signal time-frequency spectrum, and the calculation amount of neural network model can be reduced.
Time-frequency image is input to preset convolutional neural networks model, according to convolution by content based on the above embodiment
The output of neural network model is as a result, obtain leakage type corresponding with mixed signal, comprising: time-frequency image is input to convolution
The convolutional layer and pond layer of neural network model carry out feature extraction to time-frequency image using convolutional layer and pond layer, when output
The corresponding two-dimensional feature vector of frequency image；Two-dimensional characteristics vector is input to the full articulamentum of convolutional neural networks model, by two
Dimensional feature vector is converted into one-dimensional characteristic vector and exports；One-dimensional characteristic vector is input to the logic of convolutional neural networks model
Layer is returned, the corresponding prediction probability of output leakage type obtains the corresponding leakage type of mixed signal according to prediction probability.
Specifically, the process flow of the time-frequency image input CNN neural network model of mixed signal to be detected can be led to
Following method is crossed to realize:
Firstly, time-frequency image is passed through (each feature extraction packet of feature extraction layer twice from the input layer input of network
Include one layer of convolutional layer and one layer of pond layer) feature extraction is carried out to the time-frequency image of mixed signal, it exports corresponding with time-frequency image
Multiple two-dimensional feature vectors.Secondly, by the full connection of obtained multiple two-dimensional feature vectors input convolutional neural networks model
Layer carries out feature self study by full articulamentum, and the two dimensional character that feature extraction layer is obtained integrates, and exports comprising more
The one-dimensional characteristic vector of a feature.Then, one-dimensional characteristic vector is input to the logistic regression layer of convolutional neural networks model, it is defeated
Prediction probability corresponding with leakage type out.The realization of softmax classifier, the one-dimensional spy that will be extracted can be used in logistic regression layer
It levies vector and inputs softmax classifier, the corresponding prediction probability of leakage type can be obtained.To reveal type as No leakage, small
For type leakage, medium-sized leakage and large-scale leakage, the output network element of softmax classifier is 4, and 4 output network elements are right respectively
Above-mentioned four kinds of leakages type is answered, therefrom selects that mixed signal is corresponding to let out according to the corresponding probability of leakage type that network element exports
Reveal type.Optionally, between each convolutional layer and pond layer, using ReLU (The Rectified Linear Unit, amendment
Linear unit) it is used as activation primitive, its main feature is that convergence is fast, ask gradient simple.
Pipeline leakage detection method provided in an embodiment of the present invention extracts feature by preset convolutional neural networks, and
The corresponding prediction probability of leakage type is exported by logistic regression layer, Accurate classification can be carried out to leakage type, ensure that inspection
Survey result efficiently and accurately.
Content based on the above embodiment is also wrapped before time-frequency image is input to preset convolutional neural networks model
It includes: obtaining multiple mixed signal samples and the corresponding leakage type label of each mixed signal sample；Become based on Fourier in short-term
It changes and time frequency analysis is carried out to each mixed signal sample, obtain the corresponding time-frequency spectrum of each mixed signal, and by each time-frequency
Spectrum is converted into two-dimensional time-frequency image；By the group of each mixed signal sample corresponding two-dimentional time-frequency image and leakage type label
Cooperation is a training sample, to obtain multiple training samples, using multiple training samples to convolutional neural networks model into
Row training.
Before time-frequency image is input to preset convolutional neural networks model, also need to be trained the neural network,
To obtain being able to carry out the default neural network model of leakage type detection, the specific steps are as follows:
Firstly, obtaining multiple mixed signal samples, and obtain each mixed signal sample in multiple mixed signal sample
Corresponding leakage type, the leakage type that each mixed signal is had determined is as the label of the mixed signal.For example, can lead to
Cross the mixed signal sample for obtaining in history leak data and determining leakage type.For each mixed signal sample, it is based on short
When each mixed signal sample of Fourier transform pairs carry out time frequency analysis, to obtain the corresponding time-frequency spectrum of each mixed signal,
And the corresponding time-frequency spectrum of each mixed signal sample is converted into two-dimensional time-frequency image.To in Fu in short-term of mixed signal sample
The conversion process of leaf transformation processing and time-frequency image can be found in above-mentioned corresponding embodiment.
Secondly, using the combination of the corresponding two-dimentional time-frequency image of each mixed signal sample and leakage type label as one
Sample, to obtain multiple training samples.The corresponding time-frequency image of mixed signal in each sample is input to the convolution of building
Neural network model, and according to the relevant parameter of output result adjustment convolutional neural networks model, it realizes to convolutional neural networks
The training process of model, to obtain above-mentioned preset convolved data network model.
Pipeline leakage detection method provided in an embodiment of the present invention, by obtaining multiple mixed signal samples and each mixing
The corresponding leakage type label of sample of signal is to obtain multiple training samples, using multiple training samples to convolutional neural networks
Model is trained, and hence for the mixed signal that is detected of the convolutional neural networks model is inputted, can be obtained accurately
Corresponding leakage type.
Content based on the above embodiment is trained convolutional neural networks model using multiple training samples, comprising:
The time-frequency image of any one sample mixed signal is input to convolutional neural networks model, output sample mixed signal is corresponding
Reveal type prediction probability；Using default loss function according to the corresponding prediction probability of sample mixed signal and sample mixed signal
Leakiness label calculate penalty values；If penalty values are less than preset threshold, convolutional neural networks model training is completed.
Firstly, any mixed signal (i.e. sample mixed signal) is chosen from mixed signal sample, by the mixed signal
Time-frequency image is input to preset convolutional neural networks model, extracts two-dimensional feature vector through feature extraction layer, full articulamentum obtains
The prediction probability of each leakage type is exported to one-dimensional vector and logistic regression layer.According to the prediction probability and sample mixing letter
The label of number leakage type, calculates the corresponding penalty values of loss function.For example, selecting loss function to select intersects entropy loss letter
Number.The relevant parameter of convolutional neural networks is updated using back propagation.Secondly, judging whether the penalty values are less than preset threshold
Value terminates training process if being less than, if being trained not less than new sample mixed signal is chosen.
Specifically, following scheme can be used in the design realization of preset convolutional neural networks model:
The parameter of convolutional layer includes the number of the image number inputted, the number of characteristic image and the two pixel.This
Each layer of image is sized in embodiment identical, is all (Mx, My)；Wherein convolution kernel size is (Kx, Ky), size is
Kx, KyConvolution kernel do convolution operation in the useful block of original image.Each convolutional layer is directed to its upper one layer Mn-1Width figure
As being K using sizex×KyConvolution kernel do convolution, for extract piece image in feature substantially process.MnThe image of a output
It is to pass through Mn-1It not is that linear excitation function obtains that the convolution operation of a image, which is added,.
In above formula, n indicates that the number of plies, image size are (Mx, My), WijIndicating size isCNN filter to defeated
The weighting parameter that j-th of picture of the image of i-th of picture and output of the image entered is operated, * indicate convolution
Operation.It is for pixel sizeInput picture Yn-1, it is by coreFilter W processing after
The output image of acquisition is Yn, size isFor Mn-1Width
The convolution of input picture responds,For Mn-1The biasing of width input picture.
For pond layer using the method for maximizing pondization sampling, output is that size is (Kx, Ky) non-overlap matrix maximum
Value.This method, which embodies for part, has the characteristics that displacement will not become, and passes through the factor (Kx, Ky) by each difference
The input picture in direction does down-sampled processing.
During being trained by above-mentioned steps, the process of undated parameter includes:
Firstly, carrying out parameter initialization before first sample mixed signal inputs.Pass through random initializtion network
Weight.Secondly, being input to default neural network model for each sample mixed signal obtains the process and calculating of prediction probability
During penalty values, by operating undated parameter as follows:
Propagated forward: the propagated forward of convolutional neural networks is specially the operation of convolutional layer and pond layer.Assuming that l layers are
Convolutional layer, l-1 layers are pond layer or input layer, then are to output before can calculating from l-1 layers to l layers of neuronTable
What is shown is l layers of ith feature figure, whereinL is the total number of plies of neural network structure, NlFor
All characteristic patterns sum, i.e. neuron number in l layers.
Backpropagation: error function is calculated when backpropagation to the weight of neuron and the local derviation of biasing.The two knot
Fruit all with inputSensitive error it is related, be typically expressed asCalculate the error of output layer and backpropagation
To first hidden layer, to calculate the residual error of each layer
Loss function employed in the embodiment of the present invention is to intersect entropy function, and E is loss function.Calculate network weight
With the gradient of biasing, weight w and biasing b are constantly updated in each study using stochastic gradient descent algorithm, until error amount
Less than preset threshold.Optionally, it is assumed that ε is learning rate, then it is as follows specifically to update iterative definition:
After calculating weight and biasing, the weight and partially eventually for update is generated in conjunction with error function and learning rate
It sets, accelerates convergent effect to reach.
Time-frequency image is input to preset convolution mind as a kind of alternative embodiment by content based on the above embodiment
Before network model, further includes: enhancing pretreatment is carried out to the time-frequency image of sample mixed signal, after obtaining enhancing pretreatment
Time-frequency image；Correspondingly, time-frequency image is input to preset convolutional neural networks model, it specially will be after enhancing pretreatment
Time-frequency image be input to preset convolutional neural networks model；Wherein, enhancing pretreatment includes median filter process.
When noise is relatively low, time-frequency image ambient noise is higher, the degradation in contrast of ambient noise and leakage signal；It will
When time-frequency spectrum is directly changed into time-frequency image, since time-frequency spectrum intensity itself is poor, it is difficult to observe leakage signal, it is therefore desirable to
Enhancing pretreatment is carried out to the two-dimentional time-frequency image of mixed signal.Correspondingly, mixed signal is input to preset convolutional Neural
Network model is input to preset convolutional neural networks model for that will enhance pretreated mixed signal.In addition, for training
Sample mixed signal also use identical preprocess method, its corresponding time-frequency image is carried out to input mind after enhancing pretreatment
It is trained through network model.Pipeline leakage detection method provided in an embodiment of the present invention, by enhancing time-frequency image
Pretreatment, accurate detection result can be obtained when noise is relatively low.
As a kind of alternative embodiment, enhancing pretreatment includes median filter process.Median filtering is a kind of non-thread mild-natured
Sliding technology, the basic principle is that the value of the certain point in Serial No. is replaced with the intermediate value of point value each in the neighborhood of a point,
Make the pixel value of surrounding close to true value, to eliminate isolated noise spot.Two-dimensional median filtering output are as follows:
G (x, y)=mid { f (x_k, y_1), (k, 1 ∈ W) } (13)
Wherein f (x, y), g (x, y) are respectively original image and treated image, and W is two dimension pattern plate, preferably when for 3 ×
3,5 × 5 region can be different shape, such as threadiness, circle, cross, circular ring shape.
Time-frequency image is input to preset convolution mind as a kind of alternative embodiment by content based on the above embodiment
Before network model, further includes: dimension-reduction treatment is carried out to the time-frequency image of sample mixed signal, after obtaining dimension-reduction treatment when
Frequency image；Correspondingly, time-frequency image is input to preset convolutional neural networks model, specifically: by after dimension-reduction treatment when
Frequency image is input to preset convolutional neural networks model.
From the angle of signal processing, dimension-reduction treatment can under the premise of keeping mixing leakage signal time-frequency characteristics, by when
Frequency image carries out dimension, makes it possible lightweight convolutional neural networks and small sample training.For example, by averaging
The time-frequency image of 128 × 16000 pixels is reduced to 28 × 140 pixels by method.Correspondingly, time-frequency image is input to default
Convolutional neural networks model, for the time-frequency image after dimension-reduction treatment is input to preset convolutional neural networks model.Pass through
Dimension-reduction treatment is carried out to the time-frequency image of mixed signal, to reduce the computation complexity of convolutional neural networks.
Based on above-mentioned each method embodiment, rolled up using Short Time Fourier Transform, median filtering, signal grade dimensionality reduction and lightweight
The method that product neural network combines realizes leakage type detection.Using reveal type as No leakage, small-sized leakage, medium-sized leakage and
For large size leakage, every kind of leakage situation carries out CNN neural metwork training using 600 groups of samples.Later, using 120 groups of samples
It is tested.Table 1 is pipe leakage type detection classification results, is leaked for No leakage, osculum leakage, the leakage of middle mouth, big mouth
In the case of four kinds, the four kinds of leakage signals identified are detected.As shown in Table 1, due to using Short Time Fourier Transform, intermediate value filter
Wave, signal grade dimensionality reduction design the optimization method combined with lightweight convolutional neural networks, and present invention obtains preferable leakages
Signal detection classification results.
1 pipe leakage type detection classification results of table
Fig. 2 is pipeline leakage structure diagram of detection device provided in an embodiment of the present invention, as shown in Fig. 2, the pipeline leakage is examined
Surveying device includes: time frequency analysis module 201, image generation module 202 and type detection module 203.Wherein, time frequency analysis module
201, for carrying out time frequency analysis to pipeline mixed signal based on Short Time Fourier Transform, obtain the corresponding time-frequency spectrum of mixed signal；
Image generation module 202 is used to time-frequency spectrum being converted into two-dimensional time-frequency image；Type detection module 203 is used for time-frequency image
It is input to preset convolutional neural networks model, according to the output of convolutional neural networks model as a result, obtaining and mixed signal pair
The leakage type answered；It wherein, include leakage signal in mixed signal, convolutional neural networks model is according to leakage type label
Mixed signal sample training after obtain, leakage type includes without leakage, osculum leakage, the leakage of middle mouth and the leakage of big mouth.
Leak detection apparatus actual acquisition to signal include leakage signal and random disturbances noise signal mixing letter
Number, generally it is regarded as the additive combination of the two, it may be assumed that
X (t)=s (t)+n (t) (14)
Wherein, n (t) is random disturbances noise, and s (t) is leakage signal.Mixed signal is leakage signal and noise signal
Combination, belongs to nonstationary random signal.
Time frequency analysis module 201 carries out time frequency analysis to the pipeline mixed signal comprising leakage signal, using in Fu in short-term
Leaf transformation obtains short-term spectrum.The time-frequency spectrum of obtained mixed signal is the function about time domain and frequency domain, image generation module
202 can be shown in two-dimensional coordinate with time domain coordinate and frequency domain coordinates, so that the abscissa ordinate with two dimensional image is opposite
It answers, and time-frequency spectrum is converted into the two-dimensional time-frequency image that convolutional neural networks model can identify.
Type detection module 203 passes through the time-frequency that preset convolutional neural networks model obtains image generation module 202
Image is handled.Sample mixed signal is to have learned that it reveals the mixed signal of type in advance, and known let out corresponding
Reveal label of the type as each sample mixed signal.It is mixed by a large amount of such sample after establishing convolutional neural networks model
It closes signal to be trained, so that obtaining preset convolutional neural networks model will acquire subsequently received mixed signal
Corresponding time-frequency image is input to handled in the default convolutional neural networks model of type detection module 203 after, can be quick
Accurately obtain the output result for revealing type accordingly.
Leakage type label can be defined according to demand, for example, there is a leakage, No leakage, or for No leakage, small-sized let out
Dew, medium-sized leakage and large-scale leakage.Allotment of the type for man power and material in subsequent maintenance process is revealed, pass can be played
The directive function of key.
Installation practice provided in an embodiment of the present invention is the detailed process and in detail in order to realize above-mentioned each method embodiment
Thin content please refers to above method embodiment, and details are not described herein again.
Pipeline leakage detection device provided in an embodiment of the present invention is believed due to being mixed by Short Time Fourier Transform to pipeline
Number time frequency analysis is carried out, the corresponding time-frequency spectrum of mixed signal is obtained, without extracting leakage signal from mixed signal, thus not
Interference vulnerable to mixed noise.In addition, the time-frequency spectrum is converted into two-dimentional time-frequency image, it is input to preset convolutional Neural net
Network model, preset neural network model according to leakage type mixed signal sample training after obtain, can export and let out
The recognition result for revealing type, so that detection process is quick and accurate.
Fig. 3 is the entity structure schematic diagram of a kind of electronic equipment provided in an embodiment of the present invention, as shown in figure 3, the service
Device may include: processor (processor) 301,302, memory communication interface (Communications Interface)
(memory) 303 and bus 304, wherein processor 301, communication interface 302, memory 303 are completed mutually by bus 304
Between communication.Communication interface 302 can be used for the information transmission of electronic equipment.Processor 301 can call in memory 303
Logical order includes following method to execute: carrying out time frequency analysis to pipeline mixed signal based on Short Time Fourier Transform, obtains
Take the corresponding time-frequency spectrum of mixed signal；Time-frequency spectrum is converted into two-dimensional time-frequency image；Time-frequency image is input to preset volume
Product neural network model, according to the output of convolutional neural networks model as a result, obtaining leakage type corresponding with mixed signal；Its
In, it include leakage signal in mixed signal, convolutional neural networks model is according to the sample mixed signal with leakage type label
It is obtained after sample training, leakage type includes without leakage, osculum leakage, the leakage of middle mouth and big mouth leakage.
In addition, the logical order in above-mentioned memory 303 can be realized by way of SFU software functional unit and conduct
Independent product when selling or using, can store in a computer readable storage medium.Based on this understanding, originally
Substantially the part of the part that contributes to existing technology or the technical solution can be in other words for the technical solution of invention
The form of software product embodies, which is stored in a storage medium, including some instructions to
So that a computer equipment (can be personal computer, server or the network equipment etc.) executes the above-mentioned each side of the present invention
The all or part of the steps of method embodiment.And storage medium above-mentioned include: USB flash disk, mobile hard disk, read-only memory (ROM,
Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic or disk etc. it is various
It can store the medium of program code.
The apparatus embodiments described above are merely exemplary, wherein unit can be as illustrated by the separation member
Or may not be and be physically separated, component shown as a unit may or may not be physical unit, i.e.,
It can be located in one place, or may be distributed over multiple network units.It can select according to the actual needs therein
Some or all of the modules realizes the purpose of the embodiment of the present invention.Those of ordinary skill in the art are not paying creativeness
Labour in the case where, it can understand and implement.
Through the above description of the embodiments, those skilled in the art can be understood that each embodiment can
It realizes by means of software and necessary general hardware platform, naturally it is also possible to pass through hardware.Based on this understanding, on
Stating technical solution, substantially the part that contributes to existing technology can be embodied in the form of software products in other words, should
Computer software product may be stored in a computer readable storage medium, such as ROM/RAM, magnetic disk, CD, including several fingers
It enables and using so that a computer equipment (can be personal computer, server or the network equipment etc.) executes each implementation
The method of certain parts of example or embodiment.
Finally, it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations；Although
Present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: it still may be used
To modify the technical solutions described in the foregoing embodiments or equivalent replacement of some of the technical features；
And these are modified or replaceed, technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution spirit and
Range.
Claims (10)
1. a kind of pipeline leakage detection method characterized by comprising
Time frequency analysis is carried out to pipeline mixed signal based on Short Time Fourier Transform, obtains the corresponding time-frequency of the mixed signal
Spectrum；
The time-frequency spectrum is converted into two-dimensional time-frequency image；
The time-frequency image is input to preset convolutional neural networks model, according to the output of the convolutional neural networks model
As a result, obtaining leakage type corresponding with the mixed signal；
It wherein, include leakage signal in the mixed signal, the convolutional neural networks model is according to leakage type label
Mixed signal sample training after obtain, the leakage type includes without leakage, osculum leakage, the leakage of middle mouth and the leakage of big mouth.
2. the method according to claim 1, wherein described be converted into two-dimensional time-frequency figure for the time-frequency spectrum
Picture, comprising:
According to the corresponding time-frequency distributions of mixed signal time-frequency spectrum, by the energy intensity of each T/F point and different gray values
It is corresponding, obtain the two-dimentional time-frequency image with different gray values.
3. the method according to claim 1, wherein described be input to preset convolution mind for the time-frequency image
Through network model, according to the output of the convolutional neural networks model as a result, obtaining leakage class corresponding with the mixed signal
Type, comprising:
The time-frequency image is input to the convolutional layer and pond layer of the convolutional neural networks model, using the convolutional layer and
The pond layer carries out feature extraction to the time-frequency image, exports the corresponding two-dimensional feature vector of the time-frequency image；
The two-dimensional characteristics vector is input to the full articulamentum of the convolutional neural networks model, by the two-dimensional feature vector
It is converted into one-dimensional characteristic vector and exports；
The one-dimensional characteristic vector is input to the logistic regression layer of the convolutional neural networks model, exports the leakage type
Corresponding prediction probability obtains the corresponding leakage type of the mixed signal according to the prediction probability.
4. the method according to claim 1, wherein described be input to preset convolution mind for the time-frequency image
Before network model, further includes:
Obtain multiple mixed signal samples and the corresponding leakage type label of each mixed signal sample；
Based on Short Time Fourier Transform to each mixed signal sample carry out time frequency analysis, obtain each mixed signal it is corresponding when
Frequency spectrum, and each time-frequency spectrum is converted into two-dimensional time-frequency image；
Using the combination of the corresponding two-dimentional time-frequency image of each mixed signal sample and leakage type label as a training sample,
To obtain multiple training samples, the convolutional neural networks model is trained using the multiple training sample.
5. according to the method described in claim 4, it is characterized in that, described refreshing to the convolution using the multiple training sample
It is trained through network model, comprising:
The time-frequency image of any one sample mixed signal is input to the convolutional neural networks model, it is mixed to export the sample
Close the corresponding leakage type prediction probability of signal；
Utilize default loss function letting out according to the corresponding prediction probability of the sample mixed signal and the sample mixed signal
Leakage degree label calculates penalty values；
If the penalty values are less than preset threshold, the convolutional neural networks model training is completed.
6. the method according to claim 1, wherein described be input to preset convolution mind for the time-frequency image
Before network model, further includes:
Enhancing pretreatment is carried out to the time-frequency image of the mixed signal, obtaining enhances pretreated time-frequency image；
Correspondingly, the time-frequency image is input to preset convolutional neural networks model, specifically:
Pretreated time-frequency image will be enhanced and be input to preset convolutional neural networks model；
Wherein, the enhancing pretreatment includes median filter process.
7. method according to claim 1-6, which is characterized in that it is described the time-frequency image is input to it is default
Convolutional neural networks model before, further includes:
Dimension-reduction treatment is carried out to the time-frequency image of the mixed signal, the time-frequency image after obtaining dimension-reduction treatment；
Correspondingly, the time-frequency image is input to preset convolutional neural networks model, specifically:
Time-frequency image after dimension-reduction treatment is input to preset convolutional neural networks model.
8. a kind of pipeline leakage detection device characterized by comprising
Time frequency analysis module obtains described mixed for carrying out time frequency analysis to pipeline mixed signal based on Short Time Fourier Transform
Close the corresponding time-frequency spectrum of signal；
Image generation module, for the time-frequency spectrum to be converted into two-dimensional time-frequency image；
Type detection module, for the time-frequency image to be input to preset convolutional neural networks model, according to the convolution
The output of neural network model is as a result, obtain leakage type corresponding with the mixed signal；
It wherein, include leakage signal in the mixed signal, the convolutional neural networks model is according to leakage type label
Mixed signal sample training after obtain, the leakage type includes without leakage, osculum leakage, the leakage of middle mouth and the leakage of big mouth.
9. a kind of electronic equipment including memory, processor and stores the calculating that can be run on a memory and on a processor
Machine program, which is characterized in that realize that the pipeline as described in any one of claim 1 to 7 is let out when the processor executes described program
The step of revealing detection method.
10. a kind of non-transient computer readable storage medium, is stored thereon with computer program, which is characterized in that the computer
It is realized when program is executed by processor as described in any one of claim 1 to 7 the step of pipeline leakage detection method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910192757.8A CN109931506A (en) | 2019-03-14 | 2019-03-14 | Pipeline leakage detection method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910192757.8A CN109931506A (en) | 2019-03-14 | 2019-03-14 | Pipeline leakage detection method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109931506A true CN109931506A (en) | 2019-06-25 |
Family
ID=66987312
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910192757.8A Pending CN109931506A (en) | 2019-03-14 | 2019-03-14 | Pipeline leakage detection method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109931506A (en) |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
| CN110440148A (en) * | 2019-07-15 | 2019-11-12 | 清华大学 | A kind of leakage loss acoustical signal classifying identification method, apparatus and system |
| CN110553151A (en) * | 2019-07-17 | 2019-12-10 | 石化盈科信息技术有限责任公司 | pipeline leakage monitoring method and system |
| CN110969123A (en) * | 2019-12-02 | 2020-04-07 | 集美大学 | Electromagnetic information leakage detection method based on frequency domain, terminal equipment and storage medium |
| WO2020102846A1 (en) * | 2018-11-23 | 2020-05-28 | The University Of Adelaide | Method and system to analyse pipeline condition |
| CN111975638A (en) * | 2020-08-18 | 2020-11-24 | 青岛海纳水下信息技术有限公司 | Water grinding and cutting system |
| WO2021022315A1 (en) * | 2019-08-02 | 2021-02-11 | The University Of Adelaide | Method and system to monitor pipeline condition |
| CN112560806A (en) * | 2021-01-26 | 2021-03-26 | 华东交通大学 | Artificial intelligence identification method for natural gas pipeline leakage signal |
| CN112613431A (en) * | 2020-12-28 | 2021-04-06 | 中北大学 | Automatic identification method, system and device for leaked gas |
| CN112762362A (en) * | 2021-01-15 | 2021-05-07 | 中国海洋石油集团有限公司 | Underwater pipeline leakage acoustic emission detection method based on convolutional neural network |
| CN112949817A (en) * | 2020-12-24 | 2021-06-11 | 长江勘测规划设计研究有限责任公司 | Water supply pipe leakage edge equipment detection method based on time convolution neural network |
| CN113392705A (en) * | 2021-05-13 | 2021-09-14 | 电子科技大学 | Method for identifying pipeline leakage target in desert area based on convolutional neural network |
| CN113606502A (en) * | 2021-07-16 | 2021-11-05 | 青岛新奥燃气设施开发有限公司 | Method for judging whether operator performs pipeline air leakage detection based on machine vision |
| CN113719764A (en) * | 2021-08-04 | 2021-11-30 | 中国人民解放军陆军勤务学院 | Pipeline leakage detection method |
| CN113781513A (en) * | 2021-08-19 | 2021-12-10 | 广东能源集团科学技术研究院有限公司 | Method and system for detecting leakage of water supply pipeline of power plant |
| CN113792585A (en) * | 2021-08-03 | 2021-12-14 | 重庆兆光科技股份有限公司 | Pipeline vibration signal monitoring method and system, electronic equipment and medium |
| CN114001880A (en) * | 2021-10-29 | 2022-02-01 | 清华大学合肥公共安全研究院 | Water supply pipeline leakage amount evaluation method based on in-pipe leakage acoustic wave analysis |
| CN114429504A (en) * | 2022-04-06 | 2022-05-03 | 北京理工大学 | Trace leakage diagnosis method for hydrogen system of fuel cell automobile |
| 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 |
| CN117668623A (en) * | 2024-02-02 | 2024-03-08 | 中国海洋大学 | Multi-sensor cross-domain fault diagnosis method for leakage of ship pipeline valve |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2000221020A (en) * | 1999-02-01 | 2000-08-11 | Mitsubishi Electric Corp | Abnormal place detecting device |
| CN106446782A (en) * | 2016-08-29 | 2017-02-22 | 北京小米移动软件有限公司 | Image identification method and device |
| CN106909784A (en) * | 2017-02-24 | 2017-06-30 | 天津大学 | Epileptic electroencephalogram (eeg) recognition methods based on two-dimentional time-frequency image depth convolutional neural networks |
| CN107977456A (en) * | 2017-12-15 | 2018-05-01 | 清华大学 | A kind of multi-source big data analysis method based on multitask depth network |
| CN108304770A (en) * | 2017-12-18 | 2018-07-20 | 中国计量大学 | A method of the flow pattern of gas-liquid two-phase flow based on time frequency analysis algorithm combination deep learning theory |
| CN108564005A (en) * | 2018-03-26 | 2018-09-21 | 电子科技大学 | A kind of human body tumble discrimination method based on convolutional neural networks |
| CN108731886A (en) * | 2018-05-30 | 2018-11-02 | 北京工业大学 | A kind of more leakage point acoustic fix ranging methods of water supply line based on iteration recursion |
| CN108896296A (en) * | 2018-04-18 | 2018-11-27 | 北京信息科技大学 | A kind of wind turbine gearbox method for diagnosing faults based on convolutional neural networks |
| CN109001557A (en) * | 2018-06-11 | 2018-12-14 | 西北工业大学 | A kind of aircraft utilities system fault recognition method based on random convolutional neural networks |
| CN109274621A (en) * | 2018-09-30 | 2019-01-25 | 中国人民解放军战略支援部队信息工程大学 | Communication protocol signals recognition methods based on depth residual error network |
| CN109271643A (en) * | 2018-08-08 | 2019-01-25 | 北京捷通华声科技股份有限公司 | A kind of training method of translation model, interpretation method and device |
| CN109340586A (en) * | 2018-11-06 | 2019-02-15 | 三川智慧科技股份有限公司 | A kind of detection method and system of water supply line leakage |
-
2019
- 2019-03-14 CN CN201910192757.8A patent/CN109931506A/en active Pending
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2000221020A (en) * | 1999-02-01 | 2000-08-11 | Mitsubishi Electric Corp | Abnormal place detecting device |
| CN106446782A (en) * | 2016-08-29 | 2017-02-22 | 北京小米移动软件有限公司 | Image identification method and device |
| CN106909784A (en) * | 2017-02-24 | 2017-06-30 | 天津大学 | Epileptic electroencephalogram (eeg) recognition methods based on two-dimentional time-frequency image depth convolutional neural networks |
| CN107977456A (en) * | 2017-12-15 | 2018-05-01 | 清华大学 | A kind of multi-source big data analysis method based on multitask depth network |
| CN108304770A (en) * | 2017-12-18 | 2018-07-20 | 中国计量大学 | A method of the flow pattern of gas-liquid two-phase flow based on time frequency analysis algorithm combination deep learning theory |
| CN108564005A (en) * | 2018-03-26 | 2018-09-21 | 电子科技大学 | A kind of human body tumble discrimination method based on convolutional neural networks |
| CN108896296A (en) * | 2018-04-18 | 2018-11-27 | 北京信息科技大学 | A kind of wind turbine gearbox method for diagnosing faults based on convolutional neural networks |
| CN108731886A (en) * | 2018-05-30 | 2018-11-02 | 北京工业大学 | A kind of more leakage point acoustic fix ranging methods of water supply line based on iteration recursion |
| CN109001557A (en) * | 2018-06-11 | 2018-12-14 | 西北工业大学 | A kind of aircraft utilities system fault recognition method based on random convolutional neural networks |
| CN109271643A (en) * | 2018-08-08 | 2019-01-25 | 北京捷通华声科技股份有限公司 | A kind of training method of translation model, interpretation method and device |
| CN109274621A (en) * | 2018-09-30 | 2019-01-25 | 中国人民解放军战略支援部队信息工程大学 | Communication protocol signals recognition methods based on depth residual error network |
| CN109340586A (en) * | 2018-11-06 | 2019-02-15 | 三川智慧科技股份有限公司 | A kind of detection method and system of water supply line leakage |
Non-Patent Citations (3)
| Title |
|---|
| 徐明远: "《MATLAB仿真在电子测量中的应用》", 31 July 2013, 西安电子科技大学 * |
| 王新颖等: "城市燃气管道泄漏诊断方法", 《消防科学与技术》 * |
| 陈雯柏: "《人工神经网络原理与实践》", 31 January 2016, 西安电子科技大学出版社 * |
Cited By (26)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020102846A1 (en) * | 2018-11-23 | 2020-05-28 | The University Of Adelaide | Method and system to analyse pipeline condition |
| CN110440148A (en) * | 2019-07-15 | 2019-11-12 | 清华大学 | A kind of leakage loss acoustical signal classifying identification method, apparatus and system |
| 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 |
| CN110553151B (en) * | 2019-07-17 | 2021-05-07 | 石化盈科信息技术有限责任公司 | Pipeline leakage monitoring method and system |
| CN110553151A (en) * | 2019-07-17 | 2019-12-10 | 石化盈科信息技术有限责任公司 | pipeline leakage monitoring method and system |
| WO2021022315A1 (en) * | 2019-08-02 | 2021-02-11 | The University Of Adelaide | Method and system to monitor pipeline condition |
| CN110969123A (en) * | 2019-12-02 | 2020-04-07 | 集美大学 | Electromagnetic information leakage detection method based on frequency domain, terminal equipment and storage medium |
| CN111975638B (en) * | 2020-08-18 | 2022-03-29 | 青岛海纳水下信息技术有限公司 | Water grinding and cutting system |
| CN111975638A (en) * | 2020-08-18 | 2020-11-24 | 青岛海纳水下信息技术有限公司 | Water grinding and cutting system |
| CN112949817A (en) * | 2020-12-24 | 2021-06-11 | 长江勘测规划设计研究有限责任公司 | Water supply pipe leakage edge equipment detection method based on time convolution neural network |
| CN112613431A (en) * | 2020-12-28 | 2021-04-06 | 中北大学 | Automatic identification method, system and device for leaked gas |
| CN112762362A (en) * | 2021-01-15 | 2021-05-07 | 中国海洋石油集团有限公司 | Underwater pipeline leakage acoustic emission detection method based on convolutional neural network |
| CN112560806A (en) * | 2021-01-26 | 2021-03-26 | 华东交通大学 | Artificial intelligence identification method for natural gas pipeline leakage signal |
| CN112560806B (en) * | 2021-01-26 | 2022-06-21 | 华东交通大学 | Artificial intelligence identification method for natural gas pipeline leakage signal |
| CN113392705A (en) * | 2021-05-13 | 2021-09-14 | 电子科技大学 | Method for identifying pipeline leakage target in desert area based on convolutional neural network |
| CN113606502A (en) * | 2021-07-16 | 2021-11-05 | 青岛新奥燃气设施开发有限公司 | Method for judging whether operator performs pipeline air leakage detection based on machine vision |
| CN113792585A (en) * | 2021-08-03 | 2021-12-14 | 重庆兆光科技股份有限公司 | Pipeline vibration signal monitoring method and system, electronic equipment and medium |
| CN113792585B (en) * | 2021-08-03 | 2023-06-27 | 重庆兆光科技股份有限公司 | Pipeline vibration signal monitoring method, system, electronic equipment and medium |
| CN113719764A (en) * | 2021-08-04 | 2021-11-30 | 中国人民解放军陆军勤务学院 | Pipeline leakage detection method |
| CN113781513A (en) * | 2021-08-19 | 2021-12-10 | 广东能源集团科学技术研究院有限公司 | Method and system for detecting leakage of water supply pipeline of power plant |
| CN113781513B (en) * | 2021-08-19 | 2023-11-28 | 广东能源集团科学技术研究院有限公司 | Leakage detection method and system for water supply pipeline of power plant |
| CN114001880A (en) * | 2021-10-29 | 2022-02-01 | 清华大学合肥公共安全研究院 | Water supply pipeline leakage amount evaluation method based on in-pipe leakage acoustic wave analysis |
| CN114429504A (en) * | 2022-04-06 | 2022-05-03 | 北京理工大学 | Trace leakage diagnosis method for hydrogen system of fuel cell automobile |
| 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 |
| CN117668623A (en) * | 2024-02-02 | 2024-03-08 | 中国海洋大学 | Multi-sensor cross-domain fault diagnosis method for leakage of ship pipeline valve |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109931506A (en) | Pipeline leakage detection method and device | |
| Bao et al. | Threat of adversarial attacks on DL-based IoT device identification | |
| CN107220606B (en) | Radar radiation source signal identification method based on one-dimensional convolutional neural network | |
| CN111178260B (en) | Modulation signal time-frequency diagram classification system based on generation countermeasure network and operation method thereof | |
| CN112884059B (en) | Small sample radar working mode classification method fusing priori knowledge | |
| CN110147812A (en) | Recognition Method of Radar Emitters and device based on expansion residual error network | |
| CN109932708A (en) | A method of the underwater surface class object based on interference fringe and deep learning | |
| Zhang et al. | A GANs-based deep learning framework for automatic subsurface object recognition from ground penetrating radar data | |
| CN111126134A (en) | Radar radiation source deep learning identification method based on non-fingerprint signal eliminator | |
| AU2021101336A4 (en) | A Classification System Of Modulation Signal Time-Frequency Image Based On Generative Adversarial Network And Its Operation Method | |
| CN113673312B (en) | Deep learning-based radar signal intra-pulse modulation identification method | |
| Li et al. | IncepTCN: A new deep temporal convolutional network combined with dictionary learning for strong cultural noise elimination of controlled-source electromagnetic data | |
| CN112990082B (en) | Detection and identification method of underwater sound pulse signal | |
| CN111931820A (en) | Water target radiation noise LOFAR spectrogram spectrum extraction method based on convolution residual error network | |
| Ni et al. | LPI radar waveform recognition based on multi-resolution deep feature fusion | |
| CN112036239A (en) | Radar signal working mode identification method and system based on deep learning network | |
| CN113780242A (en) | Cross-scene underwater sound target classification method based on model transfer learning | |
| Ye et al. | Recognition algorithm of emitter signals based on PCA+ CNN | |
| Zhou et al. | MSAR‐DefogNet: Lightweight cloud removal network for high resolution remote sensing images based on multi scale convolution | |
| CN114117912A (en) | Sea clutter modeling and inhibiting method under data model dual drive | |
| CN115114949A (en) | Intelligent ship target identification method and system based on underwater acoustic signals | |
| CN112766381A (en) | Attribute-guided SAR image generation method under limited sample | |
| CN113343924B (en) | Modulation signal identification method based on cyclic spectrum characteristics and generation countermeasure network | |
| CN109654383A (en) | A kind of leak degree of water supply line determines method and system | |
| CN109945075A (en) | A kind of water supply line leakiness detection method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190625 |
|
| RJ01 | Rejection of invention patent application after publication |