A method for monitoring leakage of a pressurized pipeline based on transfer learning, an electronic device and a storage medium

Abstract

A kind of pressure pipeline leakage monitoring method based on transfer learning, electronic equipment and storage medium belong to water supply and drainage pipeline leakage detection field.To solve the effective monitoring pipeline leakage situation, the present application collects pipeline leakage data, the collected leakage sound is pretreated, then leakage sound signal feature extraction is carried out, leakage sound feature vector dataset is constructed, neural network model for pressure pipeline leakage monitoring based on transfer learning is constructed, the constructed training set is used to train the neural network model for pressure pipeline leakage monitoring based on transfer learning, the trained neural network model for pressure pipeline leakage monitoring based on transfer learning is used to transfer neural network parameters, then new leakage sound feature vector is input, and the result of pressure pipeline leakage monitoring based on transfer learning is obtained.The present application effectively monitors pipeline leakage situation, reduces the loss of water resources;understand and master the state of water pipeline leakage, provide strong support for the maintenance of water supply network.

Description

Technical Field

[0001] This invention belongs to the field of water supply and drainage pipeline leakage detection, specifically involving a pressure pipeline leakage monitoring method, electronic equipment, and storage medium based on transfer learning. Background Technology

[0002] Water is the source of life and an indispensable resource in daily life. However, leaks in water pipes not only waste water resources but also cause numerous inconveniences. Currently, the most widely used software-based method is transient flow analysis. However, transient flow analysis relies heavily on prior information about the target pipeline system or model, which is difficult to obtain completely, limiting the practical application of this method. With the improvement of computing power and the rapid development of deep learning technology, data-driven pipeline leak detection and location algorithms have been widely researched and applied, becoming the main technological trend for solving leak detection problems. However, data-driven pipeline leak detection still has significant room for improvement in areas such as detection accuracy.

[0003] Data-driven neural networks can effectively identify leakage sounds, avoiding the need for human judgment based on experience. However, labeling and acquiring leakage sounds in new pipelines is extremely expensive and difficult. The idea behind transfer learning is to use existing source domain knowledge to help learn new target domain knowledge. The core is to find the similarity between the learned source domain knowledge and the target domain knowledge. Usually, learning directly from scratch in the target domain is too costly. For example, if there is only a small amount of leakage sound data in a new pipeline, it is usually not feasible to directly train a leakage monitoring neural network in such a low-resource situation. Summary of the Invention

[0004] The problem this invention aims to solve is to effectively monitor pipeline leakage. It proposes a method, electronic device, and storage medium for monitoring pressurized pipeline leakage based on transfer learning.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A method for monitoring leakage in pressurized pipelines based on transfer learning includes the following steps:

[0007] S1. Based on the public pipeline leakage sound database, collect pipeline leakage data, including leakage flow rate and corresponding leakage sound;

[0008] S2. Preprocess the leakage sound collected in step S1, then extract the leakage sound signal features and construct a leakage sound feature vector dataset;

[0009] S3. Divide the leakage sound feature vector dataset obtained in step S2 into a training set and a validation set;

[0010] S4. Construct a neural network model for monitoring leaks in pressurized pipelines based on transfer learning. Use the training set constructed in step S3 to train the neural network model for monitoring leaks in pressurized pipelines based on transfer learning to obtain a trained neural network model for monitoring leaks in pressurized pipelines based on transfer learning.

[0011] S5. Collect pipeline leakage data under different leakage conditions in the new pipeline seepage detection project, process the pipeline leakage data using the method in step S2, and obtain a new leakage sound feature vector;

[0012] S6. Transfer the neural network parameters of the pressure pipeline leakage monitoring neural network model trained in step S4, and then input the new leakage sound feature vector obtained in step S5 to obtain the pressure pipeline leakage monitoring results based on transfer learning.

[0013] Furthermore, the specific implementation method of step S2 includes the following steps:

[0014] S2.1. Preprocess the leakage sound collected in step S1:

[0015] S2.1.1. Normalization: Normalization is used to eliminate the influence of intensity differences in the leakage sound collected in step S1. This is the original sound signal of leakage. The normalized leakage sound signal is calculated using the following expression:

[0016]

[0017] in, The minimum leakage sound signal, This is the maximum sound signal of leakage;

[0018] S2.1.2. The normalized leakage sound signal from step S2.1.1 is pre-emphasized, and the calculation expression is as follows:

[0019]

[0020] in, This is the leakage sound signal after the nth pre-emphasis. Let be the nth normalized leakage sound signal, and 'a' be the pre-emphasis coefficient, with a value ranging from 0.99 to 0.95. For the first A normalized leak sound signal;

[0021] S2.1.3. Perform frame segmentation on the pre-emphasized leakage sound signal from step S2.1.2, and then segment the nth leakage sound signal into frames. A frame is defined as ;

[0022] S2.1.4. Perform windowing processing on the leakage sound signal after framing in step S2.1.3, and set the windowing coefficient. The calculation expression is:

[0023]

[0024] Where N is the length of a frame after framing;

[0025] Then the sound signal of leakage after adding a window is obtained. The expression is:

[0026] ;

[0027] S2.1.5. Perform a Fast Fourier Transform on the windowed leakage sound signal obtained in step S2.1.4. The expression is:

[0028]

[0029] in, Leakage sound signal frequency domain signal, To detect the sound signal of leakage Perform a Fast Fourier Transform.

[0030] S2.2. The first step after windowing in step S2.1 is... Frame leakage sound signal Perform feature extraction;

[0031] S2.2.1. Extract the fundamental period. short-time autocorrelation function Defined as:

[0032]

[0033] Then the fundamental period The calculation expression is:

[0034] ;

[0035] S2.2.2. Extract short-time energy; the calculation expression is as follows:

[0036]

[0037] in, For the first Frame Short Time Energy

[0038] S2.2.3. Extracting Short-Time Energy: Short-Time Zero-Crossing Rate refers to the number of times the signal crosses the horizontal axis within one frame. Definition: The expression is:

[0039]

[0040] Then the first Frame leakage sound signal short-time zero crossing rate The expression is:

[0041] ;

[0042] S2.2.4. Extracting Feature Frequencies Depending on the environment, take the first 2 to 6 values; the expression is:

[0043] ;

[0044] in, The m-th order characteristic frequency;

[0045] S2.2.5. Extracting the first-order difference of feature frequencies The expression is:

[0046] ;

[0047] S2.2.6. Extract statistical features, including maximum values. Minimum value Arithmetic mean The expression is:

[0048]

[0049]

[0050]

[0051] Extract the relative positions of the maximum and minimum values, and define a position function. for lie in The expression for the nth position is:

[0052]

[0053] Then the relative position of the maximum value The relative position of the minimum value The expression is:

[0054]

[0055]

[0056] Extracting standard deviation skewness and kurtosis The expression is:

[0057]

[0058]

[0059]

[0060] in, For the first Frame leakage sound signal The mean;

[0061] S2.3. Based on the leakage sound signal features extracted in step S2.2, construct a leakage sound feature vector. The feature vector dataset of leakage sounds is obtained, and the expression is:

[0062] , , , , , ).

[0063] Furthermore, in step S3, the leakage sound feature vector dataset is divided into a training set and a validation set in an 8:2 ratio.

[0064] Furthermore, the specific implementation method of step S4 includes the following steps:

[0065] S4.1. Construct a neural network model for monitoring leaks in pressurized pipelines based on transfer learning. The model is constructed using a multilayer perceptron with 6 to 8 layers, where the shallowest layer is the input layer and the deepest layer is the output layer. The activation function is Sigmoid.

[0066] S4.2. The training set is used for model training using the ten-fold crossover method. The backpropagation algorithm is used to iteratively train the neural network model for monitoring leaks in pressurized pipelines based on transfer learning constructed in step S4.1 based on the training set, and to determine the values ​​of each parameter of the neural network model for monitoring leaks in pressurized pipelines based on transfer learning.

[0067] S4.3. The validation set is used to evaluate the performance of the neural network model for monitoring leaks in pressurized pipelines based on transfer learning after the iterative training is completed, and to obtain the final performance evaluation result of the neural network model for monitoring leaks in pressurized pipelines based on transfer learning.

[0068] Furthermore, the specific implementation method of step S6 includes the following steps:

[0069] S6.1. Freeze the first 3 to 5 layers of the neural network model for monitoring leaks in pressurized pipelines based on transfer learning trained in step S4, release some hidden and output layers, initialize 2 to 3 hidden layers, and create a new output layer. The activation function of the output layer is Sigmoid.

[0070] S6.2. Input the transferred neural network obtained in step S6.1 into the new leakage sound feature vector obtained in step S5 to obtain the pressure pipeline leakage monitoring results based on transfer learning.

[0071] An electronic device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the described method for monitoring leaks in pressurized pipelines based on transfer learning.

[0072] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the aforementioned method for monitoring leaks in pressurized pipelines based on transfer learning.

[0073] The beneficial effects of this invention are:

[0074] The present invention discloses a method for monitoring leakage in pressurized pipelines based on transfer learning. This method can accurately determine the occurrence of pipeline leakage and detect the leakage flow rate of pressurized pipelines.

[0075] This invention discloses a method for monitoring leaks in pressurized pipelines based on transfer learning. Leakage sound recognition systems require processing large amounts of sound data, but in practical applications, datasets are often limited. This method utilizes existing leakage sound data to pre-train the model, thereby improving model performance and reducing the difficulty of collecting data from new pipelines to be detected.

[0076] The present invention discloses a pressurized pipeline leakage monitoring method based on transfer learning, which adopts a general leakage sound feature extraction method, simplifies the leakage sound feature extraction process, and reduces the model's data requirements due to the complexity of neural networks.

[0077] The present invention discloses a pressurized pipeline leakage monitoring method based on transfer learning, which effectively enables the pipeline leakage detection model to learn and adjust according to real-time pipeline operation data, thereby better adapting to pipeline changes and leakage occurrences, and improving its detection accuracy and real-time performance. Transfer learning can reduce the data requirements for new pipeline leakage detection projects by approximately 50%.

[0078] The present invention discloses a pressurized pipeline leakage monitoring method based on transfer learning, which reduces water resource loss by effectively monitoring pipeline leakage; understands and grasps the leakage status of water transmission pipelines, providing strong support for the maintenance of water supply networks; and utilizes the noise generated when pipeline leakage occurs to propose an effective leakage monitoring method for long-distance water transmission pipelines, ensuring the normal operation of water transmission pipelines. Attached Figure Description

[0079] Figure 1 This is a flowchart of a pressurized pipeline leakage monitoring method based on transfer learning as described in this invention;

[0080] Figure 2 This is a flowchart of the ten-fold cross method described in this invention;

[0081] Figure 3 This is a schematic diagram of the frozen migration network described in this invention. Detailed Implementation

[0082] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only for explaining the invention and are not intended to limit the invention; that is, the described specific embodiments are merely a part of the embodiments of the invention, and not all of them. The components of the specific embodiments of the invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations, and the invention may also have other embodiments.

[0083] Therefore, the following detailed description of specific embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected specific embodiments of the invention. All other specific embodiments obtained by those skilled in the art based on these specific embodiments without inventive effort are within the scope of protection of this invention.

[0084] To further understand the invention's content, features, and effects, the following specific embodiments are provided, along with accompanying drawings. Figure 1 -Appendix Figure 3 Detailed explanation is as follows: Specific implementation method one:

[0086] A method for monitoring leakage in pressurized pipelines based on transfer learning includes the following steps:

[0087] S1. Based on the public pipeline leakage sound database, collect pipeline leakage data, including leakage flow rate and corresponding leakage sound;

[0088] S2. Preprocess the leakage sound collected in step S1, then extract the leakage sound signal features and construct a leakage sound feature vector dataset;

[0089] Furthermore, the specific implementation method of step S2 includes the following steps:

[0090] S2.1. Preprocess the leakage sound collected in step S1:

[0091] S2.1.1. Normalization: Normalization is used to eliminate the influence of intensity differences in the leakage sound collected in step S1. This is the original sound signal of leakage. The normalized leakage sound signal is calculated using the following expression:

[0092]

[0093] in, The minimum leakage sound signal, This is the maximum sound signal of leakage;

[0094] Furthermore, because the leakage sound is affected by distance and environmental factors during the recording process, the recorded sound intensity may deviate significantly. Normalization is used to eliminate the influence of these differences that reduce sound intensity.

[0095] S2.1.2. The normalized leakage sound signal from step S2.1.1 is pre-emphasized, and the calculation expression is as follows:

[0096]

[0097] in, This is the leakage sound signal after the nth pre-emphasis. Let be the nth normalized leakage sound signal, and 'a' be the pre-emphasis coefficient, with a value ranging from 0.99 to 0.95. For the first A normalized leak sound signal;

[0098] Furthermore, high-frequency sounds are more easily absorbed and attenuated than low-frequency sounds during propagation. Pre-emphasis can compensate for this high-frequency loss, making the signal more uniform.

[0099] S2.1.3. Perform frame segmentation on the pre-emphasized leakage sound signal from step S2.1.2, and then segment the nth leakage sound signal into frames. A frame is defined as ;

[0100] Furthermore, the sound signal of pipe leakage exhibits a relatively stable and periodic state in the short term, with a frame length of 10~30ms, and consecutive frames need to overlap by 30%~50% to ensure the complete preservation of effective information.

[0101] S2.1.4. Perform windowing processing on the leakage sound signal after framing in step S2.1.3, and set the windowing coefficient. The calculation expression is:

[0102]

[0103] Where N is the length of a frame after framing;

[0104] Then the sound signal of leakage after adding a window is obtained. The expression is:

[0105] ;

[0106] Furthermore, the framed audio signal is multiplied by a window function to emphasize the signal data in the middle of the frame, while also satisfying the continuity requirement of the discrete Fourier transform for signal extension.

[0107] S2.1.5. Perform a Fast Fourier Transform on the windowed leakage sound signal obtained in step S2.1.4. The expression is:

[0108]

[0109] in, Leakage sound signal frequency domain signal, To detect the sound signal of leakage Perform a Fast Fourier Transform.

[0110] S2.2. The first step after windowing in step S2.1 is... Frame leakage sound signal Perform feature extraction;

[0111] S2.2.1. Extract the fundamental period. short-time autocorrelation function Defined as:

[0112]

[0113] Then the fundamental period The calculation expression is:

[0114] ;

[0115] S2.2.2. Extracting short-time energy. Short-time energy refers to the energy contained in a single frame of signal. The calculation expression is as follows:

[0116]

[0117] in, For the first Frame short-time energy;

[0118] S2.2.3. Extracting Short-Time Energy: Short-Time Zero-Crossing Rate refers to the number of times the signal crosses the horizontal axis within one frame. Definition: The expression is:

[0119]

[0120] Then the first Frame leakage sound signal short-time zero crossing rate The expression is:

[0121] ;

[0122] S2.2.4. Extracting Feature Frequencies Depending on the environment, take the first 2 to 6 values; the expression is:

[0123] ;

[0124] in, The m-th order characteristic frequency;

[0125] Furthermore, the frequency of the leakage sound is mainly concentrated between 20 and 300 Hz, and extraction... The characteristic frequencies are between 20 and 300 Hz, and the first 2 to 6 are selected depending on the environment.

[0126] S2.2.5. Extracting the first-order difference of feature frequencies The expression is:

[0127] ;

[0128] S2.2.6. Extract statistical features, including maximum values. Minimum value Arithmetic mean The expression is:

[0129]

[0130]

[0131]

[0132] Extract the relative positions of the maximum and minimum values, and define a position function. for lie in The expression for the nth position is:

[0133]

[0134] Then the relative position of the maximum value The relative position of the minimum value The expression is:

[0135]

[0136]

[0137] Extracting standard deviation skewness and kurtosis The expression is:

[0138]

[0139]

[0140]

[0141] in, For the first Frame leakage sound signal The mean;

[0142] S2.3. Based on the leakage sound signal features extracted in step S2.2, construct a leakage sound feature vector. The feature vector dataset of leakage sounds is obtained, and the expression is:

[0143] , , , , , );

[0144] S3. Divide the leakage sound feature vector dataset obtained in step S2 into a training set and a validation set;

[0145] Furthermore, in step S3, the leakage sound feature vector dataset is divided into a training set and a validation set in an 8:2 ratio.

[0146] S4. Construct a neural network model for monitoring leaks in pressurized pipelines based on transfer learning. Use the training set constructed in step S3 to train the neural network model for monitoring leaks in pressurized pipelines based on transfer learning to obtain a trained neural network model for monitoring leaks in pressurized pipelines based on transfer learning.

[0147] Furthermore, the specific implementation method of step S4 includes the following steps:

[0148] S4.1. Construct a neural network model for monitoring leaks in pressurized pipelines based on transfer learning. The model is constructed using a multilayer perceptron with 6 to 8 layers, where the shallowest layer is the input layer and the deepest layer is the output layer. The activation function is Sigmoid.

[0149] S4.2. The training set is used for model training using the ten-fold crossover method. The backpropagation algorithm is used to iteratively train the neural network model for monitoring leaks in pressurized pipelines based on transfer learning constructed in step S4.1 based on the training set, and to determine the values ​​of each parameter of the neural network model for monitoring leaks in pressurized pipelines based on transfer learning.

[0150] S4.3. The validation set is used to evaluate the performance of the neural network model for monitoring leaks in pressurized pipelines based on transfer learning after the iterative training is completed, and to obtain the final performance evaluation result of the neural network model for monitoring leaks in pressurized pipelines based on transfer learning.

[0151] S5. Collect pipeline leakage data under different leakage conditions in the new pipeline seepage detection project, process the pipeline leakage data using the method in step S2, and obtain a new leakage sound feature vector;

[0152] S6. Transfer the neural network parameters of the pressure pipeline leakage monitoring neural network model trained in step S4, and then input the new leakage sound feature vector obtained in step S5 to obtain the pressure pipeline leakage monitoring results based on transfer learning.

[0153] Furthermore, the specific implementation method of step S6 includes the following steps:

[0154] S6.1. Freeze the first 3 to 5 layers of the neural network model for monitoring leaks in pressurized pipelines based on transfer learning trained in step S4, release some hidden and output layers, initialize 2 to 3 hidden layers, and create a new output layer. The activation function of the output layer is Sigmoid.

[0155] S6.2. Input the transferred neural network obtained in step S6.1 into the new leakage sound feature vector obtained in step S5 to obtain the pressure pipeline leakage monitoring results based on transfer learning. Specific Implementation Method Two:

[0157] An electronic device includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the pressure pipeline leakage monitoring method based on transfer learning as described in Specific Embodiment 1.

[0158] The computer device of the present invention may include a processor and a memory, such as a microcontroller containing a central processing unit. The processor executes the computer program stored in the memory to implement the steps of the aforementioned method for monitoring leaks in pressurized pipelines based on transfer learning. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created based on the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device. Specific implementation method three:

[0160] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the pressure pipeline leakage monitoring method based on transfer learning as described in Specific Embodiment 1.

[0161] The computer-readable storage medium of the present invention can be any form of storage medium that can be read by the processor of a computer device, including but not limited to non-volatile memory, volatile memory, ferroelectric memory, etc. The computer-readable storage medium stores a computer program. When the processor of the computer device reads and executes the computer program stored in the memory, the steps of the aforementioned method for monitoring leaks in pressurized pipelines based on transfer learning can be implemented. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0162] It should be noted that relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0163] Although this application has been described above with reference to specific embodiments, various modifications can be made and components can be replaced with equivalents without departing from the scope of this application. In particular, as long as there is no structural conflict, the features in the specific embodiments disclosed in this application can be combined with each other in any way. The lack of an exhaustive description of these combinations in this specification is merely for the sake of brevity and resource conservation. Therefore, this application is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A method for monitoring leakage in pressurized pipelines based on transfer learning, characterized in that, Includes the following steps: S1. Based on the public pipeline leakage sound database, collect pipeline leakage data, including leakage flow rate and corresponding leakage sound; S2. Preprocess the leakage sound collected in step S1, then extract the leakage sound signal features and construct a leakage sound feature vector dataset; The specific implementation method of step S2 includes the following steps: S2.

1. Preprocess the leakage sound collected in step S1: S2.1.

1. Normalization: Normalization is used to eliminate the influence of intensity differences in the leakage sound collected in step S1. This is the original sound signal of leakage. The normalized leakage sound signal is calculated using the following expression: in, The minimum leakage sound signal, This is the maximum sound signal of leakage; S2.1.

2. The normalized leakage sound signal from step S2.1.1 is pre-emphasized, and the calculation expression is as follows: in, This is the leakage sound signal after the nth pre-emphasis. Let be the nth normalized leakage sound signal, and 'a' be the pre-emphasis coefficient, with a value ranging from 0.99 to 0.

95. For the first A normalized leak sound signal; S2.1.

3. Perform frame segmentation on the pre-emphasized leakage sound signal from step S2.1.2, and then segment the nth leakage sound signal into frames. A frame is defined as ; S2.1.

4. Perform windowing processing on the leakage sound signal after framing in step S2.1.3, and set the windowing coefficient. The calculation expression is: Where N is the length of a frame after framing; Then the sound signal of leakage after adding a window is obtained. The expression is: ； S2.1.

5. Perform a Fast Fourier Transform on the windowed leakage sound signal obtained in step S2.1.

4. The expression is: in, Leakage sound signal frequency domain signal, To detect the sound signal of leakage Perform a fast Fourier transform; S2.

2. The first step after windowing in step S2.1 is... Frame leakage sound signal Perform feature extraction; S2.2.

1. Extract the fundamental period. short-time autocorrelation function Defined as: Then the fundamental period The calculation expression is: ； S2.2.

2. Extract short-time energy; the calculation expression is as follows: in, For the first Frame short-time energy; S2.2.

3. Extracting Short-Time Energy: Short-Time Zero-Crossing Rate refers to the number of times the signal crosses the horizontal axis within one frame. Definition: The expression is: Then the first Frame leakage sound signal short-time zero crossing rate The expression is: ； S2.2.

4. Extracting Feature Frequencies Depending on the environment, take the first 2 to 6 values; the expression is: ； in, The m-th order characteristic frequency; S2.2.

5. Extracting the first-order difference of feature frequencies The expression is: ； S2.2.

6. Extract statistical features, including maximum values. Minimum value Arithmetic mean The expression is: Extract the relative positions of the maximum and minimum values, and define a position function. for lie in The expression for the nth position is: Then the relative position of the maximum value The relative position of the minimum value The expression is: Extracting standard deviation skewness and kurtosis The expression is: in, For the first Frame leakage sound signal The mean; S2.

3. Based on the leakage sound signal features extracted in step S2.2, construct a leakage sound feature vector. The feature vector dataset of leakage sounds is obtained, and the expression is: , , , , , )； S3. Divide the leakage sound feature vector dataset obtained in step S2 into a training set and a validation set; S4. Construct a neural network model for monitoring leaks in pressurized pipelines based on transfer learning. Use the training set constructed in step S3 to train the neural network model for monitoring leaks in pressurized pipelines based on transfer learning to obtain a trained neural network model for monitoring leaks in pressurized pipelines based on transfer learning. The specific implementation method of step S4 includes the following steps: S4.

1. Construct a neural network model for monitoring leaks in pressurized pipelines based on transfer learning. The model is constructed using a multilayer perceptron with 6 to 8 layers, where the shallowest layer is the input layer and the deepest layer is the output layer. The activation function is Sigmoid. S4.

2. The training set is used for model training using the ten-fold crossover method. The backpropagation algorithm is used to iteratively train the neural network model for monitoring leaks in pressurized pipelines based on transfer learning constructed in step S4.1 based on the training set, and to determine the values ​​of each parameter of the neural network model for monitoring leaks in pressurized pipelines based on transfer learning. S4.

3. The validation set is used to evaluate the performance of the neural network model for monitoring leaks in pressurized pipelines based on transfer learning after the iterative training is completed, and to obtain the final performance evaluation result of the neural network model for monitoring leaks in pressurized pipelines based on transfer learning. S5. Collect pipeline leakage data under different leakage conditions in the new pipeline seepage detection project, process the pipeline leakage data using the method in step S2, and obtain a new leakage sound feature vector; S6. Transfer the neural network parameters of the pressure pipeline leakage monitoring neural network model trained in step S4, and then input the new leakage sound feature vector obtained in step S5 to obtain the pressure pipeline leakage monitoring results based on transfer learning.

2. The method for monitoring leaks in pressurized pipelines based on transfer learning according to claim 1, characterized in that, Step S3: The leakage sound feature vector dataset is divided into a training set and a validation set in an 8:2 ratio.

3. The method for monitoring leaks in pressurized pipelines based on transfer learning according to claim 2, characterized in that, The specific implementation method of step S6 includes the following steps: S6.

1. Freeze the first 3 to 5 layers of the neural network model for monitoring leaks in pressurized pipelines based on transfer learning trained in step S4, release some hidden and output layers, initialize 2 to 3 hidden layers, and create a new output layer. The activation function of the output layer is Sigmoid. S6.

2. Input the transferred neural network obtained in step S6.1 into the new leakage sound feature vector obtained in step S5 to obtain the pressure pipeline leakage monitoring results based on transfer learning.

4. An electronic device, characterized in that, It includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the pressure pipeline leakage monitoring method based on transfer learning as described in any one of claims 1-3.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the pressurized pipeline leakage monitoring method based on transfer learning as described in any one of claims 1-3.

Images