Leakage detection method and system for water supply pipeline, and storage medium and smart terminal

Through the support vector machine algorithm, the high correlation characteristics in water supply pipeline leakage detection are screened, and the training and test sets are constructed, which solves the problems of high accuracy and cost in the existing technology under complex conditions, and realizes real-time and efficient detection of the water supply pipeline network .

WO2024244556A1PCT designated stage expired Publication Date: 2024-12-05NINGBO DONGTAI WATER TECHNOLOGY CO LTD +1

Patent Information

Application Number
PCT/CN2024/078823
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2023-05-30
Filing Date
2024-02-27
Publication Date
2024-12-05

AI Technical Summary

Technical Problem

The accuracy of existing water supply pipeline leak detection methods is affected under complex conditions, and the calculation and time cost are high, making it difficult to achieve real-time detection.

Method used

By obtaining the training sample data set, calculating the feature correlation weight, filtering features with high correlation, building sample training sets and test sets, training leak detection models, reducing interference, improving detection accuracy, and Signal quality is enhanced by pre-emphasis and standardization.

Benefits of technology

It improves the accuracy and speed of leakage detection, reduces the calculation and time cost of model training, and realizes real-time detection of the water supply pipeline network.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN2024078823_05122024_PF_FP_ABST
    Figure CN2024078823_05122024_PF_FP_ABST
Patent Text Reader

Abstract

A leakage detection method and system for a water supply pipeline, and a storage medium and a smart terminal. The method comprises: acquiring a training sample data set (100); performing model training on the training sample data set, so as to acquire a model output result (103); calculating the sum of feature relevance weights (104); calculating a relevance weight of each group of sample data features (105); calculating relevance information, and acquiring selected sample data features (106); and constructing a sample training set and a sample test set, and performing training according to the sample training set, so as to acquire a leakage detection model (108). By means of only selecting features having a relatively high relevance, the interference of other features is reduced, the interference and constraints of complex conditions on the leakage detection method are overcome, and the detection accuracy is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Water supply pipeline leakage detection method, system, storage medium and intelligent terminal Technical Field

[0001] The present application relates to the field of water leakage detection and positioning technology, and in particular to a water supply pipeline leakage detection method, system, storage medium and intelligent terminal. Background Art

[0002] Statistics show that leaks in water supply pipelines are widespread, resulting in significant water waste and enormous economic losses. Furthermore, large amounts of water seeping through pipelines and underground spaces not only waste water resources but also alter the surrounding soil structure, potentially causing large-scale collapses and posing a significant threat to life and health. Therefore, addressing this issue is urgent.

[0003] The traditional method for detecting leaks in urban water supply pipelines involves having experienced workers use specialized listening equipment in the dead of night to listen to the sound of water flowing through the pipeline section by section, thereby determining the presence and location of leaks. This method remains the mainstream leak detection method due to its high accuracy. However, the complex conditions in the real world significantly impact its effectiveness. Factors such as pipe material, pipe diameter, ambient noise, soil structure, and burial depth can all affect the listening equipment's effectiveness and the workers' judgment.

[0004] The following problems exist in the existing technology. Model detection is currently basically adopted. However, due to the influence of complex conditions, the computational cost and time cost of the model training process are greatly increased, and there is still room for improvement.

[0005] Summary of the Invention

[0006] In order to improve the problem that complex conditions in real environments will greatly affect the effectiveness of the method, the effect of listening equipment and the judgment of workers, the present application provides a water supply pipeline leakage detection method, system, storage medium and intelligent terminal.

[0007] In a first aspect, the present application provides a water supply pipeline leakage detection method, which adopts the following technical solution:

[0008] A water supply pipeline leakage detection method comprising:

[0009] Obtaining a training sample data set, wherein the training sample data set includes several groups of sample data features;

[0010] Use the preset initial support vector machine algorithm to train the model on the training sample data set to obtain the model output results and retain the various model parameter information during the training model process;

[0011] Calculate the sum of feature correlation weights based on each model parameter information;

[0012] Calculate the correlation weight of each set of sample data features based on the model output results and various model parameter information;

[0013] Calculate the ratio of the correlation weight of each set of sample data features to the total of the feature correlation weights to obtain the correlation information of each set of sample data features;

[0014] Filter the relevance information according to a preset filtering method to obtain the selected sample data features;

[0015] A sample training set and a sample test set are constructed based on the selected sample data features, and a leakage detection model is trained and obtained based on the sample training set.

[0016] By adopting the above technical solution, the correlation of various features of the leakage sound sample signal in the leakage detection task is calculated, a feature subset is screened, a support vector machine model is trained, and then the leakage detection results are obtained. On the one hand, only features with high correlation are selected to reduce the interference of other features, overcome the interference and constraints of complex conditions on the leakage detection method, and improve the accuracy of detection; on the other hand, the reduction of features effectively reduces the computational cost and time cost in the model training process, thereby realizing real-time detection of the water supply network.

[0017] Optionally, the following steps may be included before the step of using a preset initial support vector machine algorithm to perform model training on the training sample dataset:

[0018] Pre-emphasize the sample data features. The calculation method of pre-emphasis is: y[n]=x[n]-αx[n-1]

[0019] Wherein, y represents the output sample data feature after pre-emphasis is completed, x represents the original sample data feature obtained, n represents the nth original sample data feature, and α is the parameter of the n-1th sampling point of the original sample data feature.

[0020] By adopting the above technical solution, the monitoring signal is pre-emphasized, the high-frequency components of the signal are enhanced, the original waveform when the leakage occurs is restored, the signal with energy loss during the transmission process is compensated, and the excessive attenuation of the high-frequency component signal is compensated in the compensation process, while the noise is not affected by the pre-emphasis, thereby improving the signal-to-noise ratio.

[0021] Optionally, before the step of performing model training on the training sample data set using a preset initial support vector machine algorithm after pre-emphasis of the sample data features, the following steps are further included:

[0022] Standardize the sample data features in the training sample data set. The standardization formula is:

[0023] Among them, x represents the original single sample data feature, x new Represents the processed single sample data feature, μ represents the mean of the original sample data feature of the same group, and σ represents the standard deviation of the original sample data feature of the same group.

[0024] By adopting the above technical solution, the feature data is standardized so that all data are at the same dimensional level, the influence of different dimensions between different features is eliminated, and the comparability of the data is improved.

[0025] Optionally, the step of calculating the correlation weight of each set of sample data features based on the model output results and various model parameter information includes:

[0026] Get the kernel function type of the support vector machine;

[0027] Determine the kernel function calculation formula based on the kernel function type;

[0028] Fill the kernel function calculation formula according to the retained model parameter information to obtain the adapted kernel function calculation formula;

[0029] The model output results are input into the adaptive kernel function calculation formula to calculate the correlation weight of each set of sample data features.

[0030] By adopting the above technical solution and calculating the feature correlation, a numerical comparison is made between each feature vector, thereby obtaining the feature contribution, which makes it easier for users to screen out the features that best reflect the leakage situation and improves the accuracy of feature identification of leakage situations.

[0031] Optionally, the steps of constructing a sample training set and a sample test set based on the selected sample data features, and training and obtaining a leakage detection model based on the sample training set include:

[0032] Performing model training on the feature training subset to correct original parameter information, and defining the corrected original parameter information as corrected parameter information;

[0033] Adjusting the model parameters based on the correction parameter information to obtain a corrected detection model;

[0034] Model verification is performed based on the modified detection model and the feature test subset to obtain the best detection model, which is then output as a leak detection model.

[0035] By adopting the above technical solution, the support vector machine model is trained using the training set, the model parameters are continuously adjusted in the process, and the model effectiveness is verified through the test set, and finally the optimal leakage detection model is obtained, which improves the detection accuracy and the speed of model training.

[0036] Optionally, the step of determining a kernel function calculation formula based on the kernel function type includes:

[0037] Determine whether the kernel function type is a preset and decomposable type;

[0038] If so, the kernel function calculation formula is as follows:

[0039] Among them, x i is the support vector, k(*,*) represents the kernel function, and d is the feature number;

[0040] If not, the kernel function calculation formula is as follows:

[0041] Here, x0 satisfies f(x0)=0, f(x) is the pre-training result, and the distance value from x0 to x is less than a preset magnitude.

[0042] By adopting the above technical solution and analyzing different kernel function types, an adaptive calculation method can be adopted for each data type, thereby improving the accuracy of the calculation results of different data.

[0043] Optionally, the step of calculating the correlation weight of each set of sample data features based on the model output results and various model parameter information further includes:

[0044] Determine the protection parameter based on the adaptive kernel function calculation formula and the sum of the feature correlation weights;

[0045] Calculate the correlation information of each sample data feature in the signal feature set. The calculation formula of the correlation information is as follows:

[0046] Among them, R d is the feature correlation of a feature in the feature vector, k d is the calculation result obtained by the corresponding kernel function calculation formula, f(x) is the pre-training result, and ε is the protection parameter.

[0047] By adopting the above technical solution, by setting the protection parameters, when the sum of the feature correlation weights is too small and the feature correlation is too large, the feature correlation is protected, thereby improving the feature correlation R d The rationality of the output; at the same time, the protection parameter settings have the effect of suppressing noise interference.

[0048] In a second aspect, the present application provides a water supply pipeline leakage detection system, which adopts the following technical solutions:

[0049] A water supply pipeline leakage detection system, comprising:

[0050] Acquisition module, used to obtain training sample data set and kernel function type;

[0051] A memory for storing a program of any one of the above-mentioned control methods for detecting water supply pipeline leakage;

[0052] The program in the processor memory can be loaded and executed by the processor to implement any of the above-mentioned control methods for detecting water supply pipeline leakage.

[0053] By adopting the above technical solution, the correlation of various features of the leakage sound signal in the leakage detection task is calculated, a feature subset is screened, a support vector machine model is trained, and then the leakage detection results are obtained. On the one hand, only features with high correlation are selected to reduce the interference of other features, overcome the interference and constraints of complex conditions on the leakage detection method, and improve the accuracy of detection; on the other hand, the reduction of features effectively reduces the computational cost and time cost in the model training process, thereby realizing real-time detection of the water supply network.

[0054] In a third aspect, the present application provides a smart terminal, which adopts the following technical solutions:

[0055] The intelligent terminal includes a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and execute any of the above-mentioned water supply pipeline leakage detection methods.

[0056] By adopting the above technical solution, the correlation of various features of the leakage sound signal in the leakage detection task is calculated, a feature subset is screened, a support vector machine model is trained, and then the leakage detection results are obtained. On the one hand, only features with high correlation are selected to reduce the interference of other features, overcome the interference and constraints of complex conditions on the leakage detection method, and improve the accuracy of detection; on the other hand, the reduction of features effectively reduces the computational cost and time cost in the model training process, thereby realizing real-time detection of the water supply network.

[0057] Fourthly, the present application provides a computer storage medium that can store corresponding programs and has the characteristics of fast, accurate analysis and processing and efficient training.

[0058] Computer-readable storage medium, using the following technical solution:

[0059] A computer-readable storage medium stores a computer program that can be loaded by a processor and execute any of the above-mentioned water supply pipeline leakage detection methods.

[0060] By adopting the above technical solution, the correlation of various features of the leakage sound signal in the leakage detection task is calculated, a feature subset is screened, a support vector machine model is trained, and then the leakage detection results are obtained. On the one hand, only features with high correlation are selected to reduce the interference of other features, overcome the interference and constraints of complex conditions on the leakage detection method, and improve the accuracy of detection; on the other hand, the reduction of features effectively reduces the computational cost and time cost in the model training process, thereby realizing real-time detection of the water supply network.

[0061] In summary, this application includes at least the following beneficial technical effects:

[0062] 1. By selecting only the features with high correlation, the interference of other features is reduced, the interference and constraints of complex conditions on the leak detection method are overcome, and the accuracy of detection is improved;

[0063] 2. By using the training set to train the support vector machine model, and continuously adjusting the model parameters in the process, and verifying the model effectiveness through the test set, the optimal leak detection model is finally obtained, which improves the detection accuracy and the speed of model training. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] FIG1 is a flow chart of a water supply pipeline leakage detection method according to an embodiment of the present application.

[0065] FIG2 is a flowchart of the steps of calculating the relevance weight of each set of sample data features based on the model output results and various model parameter information in an embodiment of the present application.

[0066] FIG3 is a flowchart of the steps of constructing a sample training set and a sample test set based on selected sample data features, and training and acquiring a leakage detection model according to the sample training set in an embodiment of the present application.

[0067] FIG4 is a flowchart of the steps of determining a kernel function calculation formula based on the kernel function type in an embodiment of the present application.

[0068] FIG5 is a flowchart of further steps for calculating the relevance weight of each set of sample data features based on the model output results and various model parameter information in an embodiment of the present application.

[0069] FIG6 is a system module diagram of a water supply pipeline leakage detection method in an embodiment of the present application. DETAILED DESCRIPTION

[0070] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to Figures 1-6 and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0071] The present invention discloses a method for detecting leakage in a water supply pipeline. Referring to FIG1 , the method comprises:

[0072] Step 100: Obtain a training sample data set, where the training sample data set includes several groups of sample data features.

[0073] The training sample dataset consists of user-collected sound signals from water supply pipes, used to train the detection model. Sample data features include time domain features (skewness, kurtosis, form factor, crest factor, impulse factor, margin factor, zero-crossing rate, root mean square (RMS), power spectrum entropy, singular spectrum entropy, etc.), frequency domain features (linear prediction coefficient, signal-to-noise ratio, spectrum centroid, spectrum bandwidth, spectrum roll-off, spectrum contrast, spectrum flatness, etc.), and time-frequency domain features (short-time Fourier spectrum, mel spectrum, mel cepstral coefficients, etc.). All data is collected and acquired using listening devices.

[0074] Step 101: Pre-emphasize sample data features.

[0075] Pre-emphasis enhances the high-frequency components of the signal characteristic of the sample data to compensate for the excessive attenuation of high-frequency components during transmission. Since the energy of the signal characteristic of the sample data gradually decreases during transmission, and the higher the signal frequency, the greater the loss of sound energy, in order to restore the original waveform when the leakage occurred, it is necessary to compensate for the damaged signal. The calculation method for pre-emphasis is: y[n] = x[n] - αx[n-1]

[0076] Where y represents the output sample data feature after pre-emphasis is completed, x represents the received sample data feature to be processed, n represents the nth sampling point of the sample data feature to be processed, and the value of α ranges from 0.9 to 1 and is determined by those skilled in the art based on experience. In the embodiment of the present invention, the value of α is 0.97.

[0077] Step 102: Standardize the sample data features in the training sample data set.

[0078] The purpose of normalization is to eliminate the impact of different dimensions between different features.

[0079] For example, the most commonly used z-score standardization method is calculated as follows:

[0080] Where x represents the original data, x new represents the processed data, μ represents the mean of the original data, and σ represents the standard deviation of the original data. Finally, a data distribution with a mean of 0 and a variance of 1 is obtained.

[0081] Step 103: Use the preset initial support vector machine algorithm to perform model training on the training sample data set to obtain model output results and retain various model parameter information during the model training process.

[0082] The pre-training result is the result obtained after the model is trained, and the model output result is regarded as the correlation R of each feature. d The sum of , the formula is:

[0083] Where d is the number of the feature, R d The correlation of each eigenvector will be explained in the subsequent steps. At this time, the model is not modified, that is, the parameters in the model do not need to be modified.

[0084] Step 104: Calculate the sum of feature relevance weights based on each model parameter information.

[0085] The calculation formula is as follows:

[0086] Where S represents the number of support vectors, x i ,y i ,α i denote the support vector, the label corresponding to the support vector, and the original parameter parameters, respectively, and k(·,·) denotes the kernel function.

[0087] Step 105: Calculate the correlation weight of each set of sample data features based on the model output results and various model parameter information.

[0088] The feature relevance is the contribution of each feature vector in the classification task. The classification task here is to classify whether it is a leakage signal.

[0089] Referring to FIG2 , the specific steps for calculating the relevance weight are as follows:

[0090] Step 1051: Obtain the kernel function type of the support vector machine.

[0091] The kernel function type is the type of support vector machine algorithm, such as the linear kernel function, the histogram intersection kernel function, the chi-square kernel function, and the Gaussian kernel function. Different kernel function types are used based on different features, and the specific method for obtaining them is to search the database.

[0092] Step 1052: Determine a kernel function calculation formula based on the kernel function type.

[0093] The kernel function calculation formula is a calculation formula for calculating the kernel function of a certain dimension in the feature vector.

[0094] Referring to Figure 3, the specific steps for determining the kernel function calculation formula are as follows:

[0095] Step 10521: Determine whether the kernel function type is a preset and decomposable type.

[0096] A decomposable type is a type in which multiple values ​​can be added and then decomposed.

[0097] Step 105211: If yes, the kernel function calculation formula is as follows:

[0098] Among them, x i is the support vector, k(*,*) represents the kernel function, and d is the feature number.

[0099] If so, for example, the linear kernel function and the histogram cross kernel function, then it means that the kernel function that satisfies the eigenvector is equal to the sum of the kernel functions of each feature dimension in the eigenvector.

[0100] Step 105212: If not, the kernel function calculation formula is as follows:

[0101] Here, x0 satisfies f(x0)=0, f(x) is the pre-training result, and the distance value from x0 to x is less than a preset magnitude.

[0102] If not, for example, the chi-square kernel function and the Gaussian kernel function, it means that the kernel function can be differentiated, and the above formula can be used. In order to make the kernel function differentiable, the distance value from x0 to x is less than the preset magnitude, and the distance from x0 to x is very close.

[0103] Step 10522: Fill in the kernel function calculation formula according to the retained model parameter information to obtain the adapted kernel function calculation formula.

[0104] In order to reduce the computational complexity caused by the change of the kernel function calculation formula due to parameter changes, the parameters in the kernel function calculation formula remain unchanged, and only their contribution is calculated.

[0105] Step 10523: Input the model output results into the adaptive kernel function calculation formula to calculate the correlation weight of each set of sample data features.

[0106] Step 106: Calculate the ratio of the correlation weight of each set of sample data features to the total of the feature correlation weights to obtain correlation information of each set of sample data features.

[0107] 4 , the specific steps for calculating the correlation weight of each set of sample data features by inputting the model output results into the adaptive kernel function calculation formula are as follows:

[0108] Step 1061: Determine the protection parameter according to the adaptive kernel function calculation formula and the sum of the feature correlation weights.

[0109] The protection parameter is to prevent the sum of feature correlation weights k s The parameter corresponding to the protective measure taken due to the small value can also have the effect of suppressing noise interference. The value can be set by looking up the table, or it can be a fixed value set manually, or it can be a fixed ratio set manually, so that the protection parameter and the sum of the kernel function and feature correlation weights are at the same level.

[0110] Step 1062: Calculate the correlation information of each sample data feature in the signal feature set. The calculation formula of the correlation information is as follows:

[0111] Among them, R d is the feature correlation of a feature in the feature vector, k d is the calculation result obtained by the corresponding kernel function calculation formula, f(x) is the pre-training result, and ε is the protection parameter.

[0112] Step 107: Filter the correlation information according to a preset filtering method to obtain selected sample data features.

[0113] The screening method is a manually set method for screening feature relevance, for example, a threshold, which means that features can be selected when the feature relevance is greater than a certain threshold; or the number of features, for example, 20, which means that 20 features with the highest relevance are screened out starting from the number.

[0114] Step 108: construct a sample training set and a sample test set based on the selected sample data features, and train and obtain a leakage detection model based on the sample training set.

[0115] The feature training subset is the set of feature vectors in the signal feature subset used to train the support vector machine model. The feature testing subset is the set of feature vectors in the signal feature subset used to verify the validity of the trained detection model. Allocation can be done by artificial proportions. Here, the number of feature training subsets is greater than the number of feature vectors in the feature testing subset, as an example to ensure a more accurate final leak detection model after training. A leak detection model is a model that can detect leak signals.

[0116] Referring to Figure 5, the specific steps are as follows:

[0117] Step 1081: Perform model training on the feature training subset to correct the original parameter information, and define the corrected original parameter information as the corrected parameter information.

[0118] The modified parameter information is the parameters in the support vector machine model adjusted after training.

[0119] Step 1082: Adjust the model parameters based on the correction parameter information to obtain a corrected detection model.

[0120] The modified detection model is a support vector machine model after the model parameters are adjusted.

[0121] Step 1083: Perform model verification based on the modified detection model and the feature test subset to obtain the optimal detection model, and output the optimal detection model as a leakage detection model.

[0122] The purpose of validation is to test the effectiveness of all adjusted models and then output the best detection model based on each validation result.

[0123] Based on the same inventive concept, an embodiment of the present invention provides a water supply pipeline leakage detection system.

[0124] 6 , a water supply pipeline leakage detection system includes:

[0125] Acquisition module, used to obtain training sample data set and kernel function type;

[0126] A memory for storing a program for a control method of a water supply pipeline leakage detection method;

[0127] The program in the memory can be loaded and executed by the processor to implement a control method for detecting leakage of a water supply pipeline.

[0128] Those skilled in the art will clearly understand that for the sake of convenience and brevity, the division of the above-mentioned functional modules is only used as an example for illustration. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. The specific working processes of the above-mentioned systems, devices, and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0129] An embodiment of the present invention provides a computer-readable storage medium storing a computer program capable of being loaded and executed by a processor for a water supply pipeline leakage detection method.

[0130] Computer storage media include, for example, various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0131] Based on the same inventive concept, an embodiment of the present invention provides an intelligent terminal including a memory and a processor, wherein the memory stores a computer program that can be loaded and executed by the processor for a water supply pipeline leakage detection method.

[0132] The above are all preferred embodiments of the present application and are not intended to limit the scope of protection of this application. Unless otherwise stated, any feature disclosed in this specification (including the abstract and drawings) may be replaced by other equivalent or similar features. In other words, unless otherwise stated, each feature is merely an example of a series of equivalent or similar features.

Claims

1. A water supply pipeline leakage detection method, characterized in that: include: Acquire a training sample data set, wherein the training sample data set includes several groups of sample data features; Use the preset initial support vector machine algorithm to train the model on the training sample data set to obtain the model output results and retain the model parameter information during the training model process; Calculate the sum of feature relevance weights based on each model parameter information; Calculate the relevance weight of each set of sample data features based on the model output results and various model parameter information; Calculate the ratio of the relevance weight of each set of sample data features to the sum of the feature relevance weights to obtain the relevance information of each set of sample data features; Filter the relevance information according to a preset filtering method to obtain the selected sample data features; A sample training set and a sample test set are constructed based on the selected sample data features, and a leakage detection model is trained and obtained based on the sample training set.

2. A water supply pipeline leakage detection method according to claim 1, characterized in that: The step of using the preset initial support vector machine algorithm to train the model on the training sample data set also includes the following steps: Pre-emphasize the sample data features. The calculation method of pre-emphasis is: y[n]=x[n]-αx[n-1] Wherein, y represents the output sample data feature after pre-emphasis is completed, x represents the original sample data feature obtained, n represents the nth original sample data feature, and α is the parameter of the n-1th sampling point of the original sample data feature.

3. A water supply pipeline leakage detection method according to claim 2, characterized in that: After pre-emphasis of the sample data features, the following steps are also included before the step of using a preset initial support vector machine algorithm to perform model training on the training sample data set: Standardize the sample data features in the training sample data set. The standardization formula is: Among them, x represents the original single sample data feature, x new represents the processed single sample data feature, μ represents the mean of the original sample data feature of the same group, and σ represents the standard deviation of the original sample data feature of the same group.

4. A water supply pipeline leakage detection method according to claim 1, characterized in that: The steps of calculating the relevance weight of each set of sample data features according to the model output results and various model parameter information include: Get the kernel function type of the support vector machine; Determine the kernel function calculation formula based on the kernel function type; Filling the kernel function calculation formula according to the retained model parameter information to obtain an adapted kernel function calculation formula; The model output results are input into the adaptive kernel function calculation formula to calculate the correlation weight of each set of sample data features.

5. A water supply pipeline leakage detection method according to claim 1, characterized in that: The steps of constructing a sample training set and a sample test set based on the selected sample data features, and training and obtaining a leakage detection model according to the sample training set include: performing model training on the feature training subset to correct original parameter information, and defining the corrected original parameter information as corrected parameter information; Adjusting the model parameters based on the correction parameter information to obtain a corrected detection model; Model verification is performed based on the modified detection model and the feature test subset to obtain the best detection model, and the best detection model is output as a leakage detection model.

6. A water supply pipeline leakage detection method according to claim 4, characterized in that: The steps of determining the kernel function calculation formula based on the kernel function type include: Determine whether the kernel function type is a preset and decomposable type; If so, the kernel function calculation formula is as follows: Among them, x i is the support vector, k(*,*) represents the kernel function, and d is the feature number; If not, the kernel function calculation formula is as follows: Among them, x0 satisfies f(x0)=0, f(x) is the pre-training result, and the distance value from x0 to x is less than a preset magnitude.

7. A water supply pipeline leakage detection method according to claim 4, characterized in that: The step of calculating the relevance weight of each set of sample data features according to the model output results and various model parameter information further includes: Determine the protection parameter according to the adaptive kernel function calculation formula and the sum of the feature correlation weights; Calculate the correlation information of each sample data feature in the signal feature set. The calculation formula of the correlation information is as follows: Among them, R d is the feature correlation of a feature in the feature vector, k d is the calculation result obtained by the corresponding kernel function calculation formula, f(x) is the pre-training result, and ∈ is the protection parameter.

8. A water supply pipeline leakage detection system, characterized in that: include: The acquisition module is used to obtain the training sample data set and kernel function type; A memory for storing a program of a water supply pipeline leakage detection method according to any one of claims 1 to 7; The program in the memory can be loaded and executed by the processor to implement a water supply pipeline leakage detection method as described in any one of claims 1 to 7.

9. An intelligent terminal, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program which can be loaded by the processor and executes a water supply pipeline leakage detection method as claimed in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: A computer program is stored which can be loaded by a processor and executes a water supply pipeline leakage detection method according to any one of claims 1 to 7.

Citation Information

Patent Citations

  • Urban gas pipeline leakage positioning method based on supporting vector machine regression

    CN106845437A

  • Natural gas pipeline leakage detection method

    CN111734961A

  • Overdue risk prediction method for optimizing multi-core support vector machine based on dragonfly algorithm

    CN113239638A

  • Pipeline leakage detection method and device based on support vector machine

    CN115640511A

  • Water supply pipeline leakage detection method and system, storage medium and intelligent terminal

    CN116705057A

Cited By

  • Water supply network leakage identification method and device based on data enhancement and hybrid neural network architecture, equipment and medium

    CN120708653A

  • Short-distance pipeline leakage detecting and positioning method and system based on multiple sensors and alpha evolution

    CN121877297A