Diaphragm wall defect detection and diagnosis scheme recommendation method based on machine learning

Abstract

The invention provides an impervious wall defect detection and diagnosis scheme recommendation method based on machine learning, and the method comprises the steps: calculating seepage fields and leakage under different defect conditions through a numerical simulation method, and forming a data set as a machine learning sample; based on PSO and SVM, training to obtain a high earth and rockfill dam diaphragm wall defect diagnosis model so as to fit implicit relationships among defect parts, defect types and leakage conditions of the high earth and rockfill dam diaphragm wall; constructing an anti-seepage wall defect and detection case library according to engineering data and literature research, and forming an anti-seepage wall defect detection rule library based on the case library; and forming a knowledge graph according to the constructed case library and rule library, and realizing defect diagnosis scheme recommendation for a specific defect position and a specific damage type. The method effectively solves the problem that the selection of the high earth and rockfill dam diaphragm wall defect detection scheme is lack of basis, effectively improves the diaphragm wall detection efficiency, and improves the diaphragm wall detection effect.

Description

Technical Field

[0001] The present application relates to the field of earth-rock dam construction in water conservancy projects, and in particular to a method for detecting and recommending a diagnosis solution for an anti-seepage wall defect based on machine learning. Background Art

[0002] As one of the important water conservancy engineering structures, the integrity of the cut-off wall of a high earth-rock dam is crucial to the stability and safety of the dam body. However, due to the influence of various factors (rainfall, water level changes, structural defects, aging of the cut-off wall, etc.), the cut-off wall of a high earth-rock dam may have defects, such as leakage, cracking, damage, etc. Accurately identifying the defect type and locating the defect position is of great significance to ensure the normal operation and safety of the high earth-rock dam.

[0003] The intelligent diagnosis technology of defects in the cut-off wall of high earth-rock dam is developing in a more digital and intelligent direction. The cut-off wall of high earth-rock dam is a concealed project. In view of the concealed characteristics of the cut-off wall, the reasonable selection of the detection scheme is crucial to the detection results. The selection of traditional detection schemes has problems such as lack of reference standards, strong subjectivity, strong blindness, weak economic efficiency and failure to closely integrate with the actual project, which limits the early discovery and positioning of defects, thus affecting the effective implementation of the entire maintenance and reinforcement work. Summary of the invention

[0004] The present application aims to solve one of the technical problems in the related art at least to some extent.

[0005] To this end, the purpose of this application is to propose a method for recommending anti-seepage wall defect detection and diagnosis schemes based on machine learning. First, it solves the problem that the location and form of defects in the anti-seepage wall of a high earth-rockfill dam cannot be determined, and thus an appropriate defect detection method cannot be selected; second, it aims to address the drawbacks in the selection of defect detection schemes for the anti-seepage wall of a high earth-rockfill dam, namely the lack of reference basis, which leads to the problem of blind selection of schemes, and proposes improvement measures.

[0006] To achieve the above objectives, the present application embodiment proposes a method for detecting and diagnosing impermeable wall defects based on machine learning, including:

[0007] Calculating the seepage field and leakage under different defect conditions by numerical simulation method to form a data set as a machine learning sample, wherein the defect conditions include defect location and defect form;

[0008] Based on the particle swarm optimization algorithm PSO and the support vector machine algorithm SVM, a high earth-rock dam cut-off wall defect diagnosis model is obtained according to the machine learning sample training to fit the implicit relationship between the defect location, defect type and leakage situation of the high earth-rock dam cut-off wall;

[0009] According to engineering data and literature research, a case library of impermeable wall defects and detection is constructed, and a rule library for impermeable wall defect detection is formed based on the case library;

[0010] Based on the constructed case library and rule library, a knowledge graph for defect diagnosis of the anti-seepage wall of high earth-rock dam is formed to recommend defect diagnosis solutions for specific defect locations and specific damage types.

[0011] Optionally, the method of calculating the seepage field and leakage under different defect conditions by a numerical simulation method to form a data set as a machine learning sample includes:

[0012] With reference to relevant literature, engineering data and the damage mechanism of the cut-off wall, two defect forms and five defect positions were proposed to simulate the leakage under different cut-off wall defects. Among them, the two defect forms are local leakage and structural damage, and the five defect positions are: the connection between the cut-off wall and the bedrock, 1 / 4 of the height of the cut-off wall above the bedrock, 1 / 2 of the height of the cut-off wall above the bedrock, 3 / 4 of the height of the cut-off wall above the bedrock, and the connection between the cut-off wall and the dam body;

[0013] According to the calculation conditions of different combination results, the finite element calculation model of high earth-rock dam impermeable wall defect leakage is established and the calculation parameters are set;

[0014] The finite element calculation model of defect leakage in the high earth-rockfill dam cut-off wall is numerically simulated, and the calculation results of the seepage field and leakage amount are extracted and sorted to form a data set to obtain a machine learning sample.

[0015] Optionally, the numerical simulation of the finite element calculation model of defect leakage of the high earth-rockfill dam cut-off wall also includes:

[0016] Before calculation, determine the results to be extracted under different leakage conditions, and use the leakage volume and seepage field as the main reference indicators of the leakage situation;

[0017] After the calculation is completed, under the finite element calculation model of defect leakage of high earth-rock dam cut-off wall under different calculation conditions, the pore water pressure of n characteristic points in the space is extracted to characterize the overall seepage field characteristics, and combined with the leakage amount, it jointly reflects the leakage status under the calculation conditions.

[0018] Optionally, the defect diagnosis model of the high earth-rock dam cut-off wall is obtained by training the machine learning samples based on the particle swarm optimization algorithm and the support vector machine algorithm, including:

[0019] Before optimizing SVM parameters, determine the parameters that need to be optimized, including regularization parameters and kernel function parameters;

[0020] Confirm the optimization range and different levels of the above parameters;

[0021] The root mean square error is used as the objective function, the PSO-SVM classification model is trained according to the machine learning samples, and the generalization ability of the model is verified according to the test set. After the training, the defect diagnosis model of the high earth-rock dam anti-seepage wall is obtained.

[0022] Optionally, the training of the PSO-SVM classification model according to the machine learning sample includes:

[0023] Dividing the machine learning samples into a training set and a test set in a preset ratio, taking the leakage amount and the seepage field as the input of the PSO-SVM classification model, and taking the defect position and the defect type as the output of the PSO-SVM classification model;

[0024] Training the PSO-SVM classification model, establishing a mapping relationship, optimizing the parameters of the SVM using the PSO algorithm until an optimal model is obtained, and using the optimal model as a defect diagnosis model for the high earth-rock dam anti-seepage wall;

[0025] The model capability of the high earth-rock dam cut-off wall defect diagnosis model is verified based on the test set.

[0026] Optionally, the construction of an anti-seepage wall defect and detection case library based on engineering data and literature research, and the formation of an anti-seepage wall defect detection rule library based on the case library, include:

[0027] Through case analysis and mechanism analysis, the possible defects of the anti-seepage wall are summarized;

[0028] Through literature research, expert consultation and engineering data collection methods, we obtained cases of impermeable wall defects in high earth-rock dams and determined the main influencing factors of impermeable wall defects.

[0029] Through literature research, the characteristics, applicability and application cases of high earth-rock dam detection schemes are summarized, and a case library for high earth-rock dam cut-off wall defect detection is constructed;

[0030] A high earth-rock dam cut-off wall defect detection rule library is constructed based on the high earth-rock dam cut-off wall defect detection case library.

[0031] Optionally, the step of constructing a high earth-rock dam cut-off wall defect detection rule library based on the high earth-rock dam cut-off wall defect detection case library includes:

[0032] Obtaining defect detection related information based on the high earth-rock dam cut-off wall defect detection case library;

[0033] Based on the defect detection related information, determine the detection requirements and detection objectives, and analyze the applicability of each detection method;

[0034] Based on the above analysis results, the relationship between the defect location and the detection scheme is analyzed to obtain a relationship analysis result, and based on the relationship analysis result, an anti-seepage wall detection rule is proposed to complete the construction of the high earth-rock dam anti-seepage wall defect detection rule base.

[0035] Optionally, the defect diagnosis knowledge graph of the high earth-rock dam cut-off wall is formed according to the constructed case library and rule library to recommend defect diagnosis solutions for specific defect locations and specific damage types, including:

[0036] The data in the constructed case base and rule base are stored in a Neo4j graph database to form a knowledge graph for defect diagnosis of the high earth-rock dam impermeable wall;

[0037] The leakage situation of the high earth-rock dam is input into the trained high earth-rock dam anti-seepage wall defect diagnosis model, and the defect location and defect type are output;

[0038] According to the output defect locations and defect types, the high earth-rock dam cut-off wall defect diagnosis knowledge graph is searched and scheme similarity matching is performed, and a recommended detection scheme is obtained based on the matching results.

[0039] Optionally, the data in the constructed case base and rule base are stored in a Neo4j graph database to form a knowledge graph for defect diagnosis of the high earth-rock dam cut-off wall, including:

[0040] Based on the unstructured data in the relevant data information in the constructed case library and rule library, data structuring is performed to construct the model layer of the knowledge graph for defect diagnosis of the anti-seepage wall of high earth-rock dam;

[0041] Based on the pattern layer, data structuring is performed on the constructed case library to construct a data layer of the knowledge graph for defect diagnosis of the anti-seepage wall of a high earth-rock dam;

[0042] Based on the data layer, the extracted knowledge is stored in the Neo4j database in the form of triples, and finally a complete knowledge graph for defect diagnosis of the high earth-rock dam cut-off wall is constructed.

[0043] Optionally, searching and matching scheme similarity on the defect diagnosis knowledge graph of the high earth-rockfill dam cut-off wall according to the output defect location and defect type includes:

[0044] The calculation formula for the similarity of the characteristic attributes of numerical data is:

[0045]

[0046] Among them, C ik For the i-th item C i The kth characteristic attribute, C 0k is the kth feature attribute of the reference item C0, aik is the value of the kth feature attribute of the ith item, a 0k The value of the kth feature attribute of the reference item;

[0047] The formula for calculating the similarity of character-type feature attributes is:

[0048]

[0049] The similarity calculation formula for the project is:

[0050]

[0051] Among them, g k represents the weight of the kth feature attribute, sim k Represents the similarity of the kth feature attribute.

[0052] The technical solution provided by the embodiments of the present application brings at least the following beneficial effects:

[0053] 1. This application introduces the PSO-SVM algorithm to solve the problem that the location and type of defects in traditional anti-seepage walls cannot be determined, and introduces the particle swarm algorithm into the machine learning model to optimize the parameter selection of the SVM model, improve the model performance, and optimize the most accurate anti-seepage wall defect classification results.

[0054] 2. This application constructs a knowledge graph for diagnosing defects in the cut-off wall of a high earth-rock dam, matches the detection schemes similarly, and recommends detection schemes based on the matching results, thereby improving the level of construction management of defect detection and repair of the cut-off wall of a high earth-rock dam.

[0055] 3. In summary, this application proposes a method for recommending detection schemes for anti-seepage walls of high earth-rock dams based on machine learning and knowledge graphs, and constructs a machine learning model optimized using a particle swarm algorithm. This method can be used for preliminary judgment of the form of defects in the anti-seepage wall, and is helpful for selecting applicable detection schemes based on the classification results of anti-seepage wall defects, thereby improving the safety, design optimization, and economic benefits of earth-rock dam projects, and providing reliable data support for engineering decision-making.

[0056] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0058] Figure 1 It is a flow chart of a method for recommending a defect detection and diagnosis scheme of an anti-seepage wall based on machine learning according to an embodiment of the present application;

[0059] Figure 2 It is an overall flow chart of a method for detecting and diagnosing impermeable wall defects based on machine learning according to an embodiment of the present application;

[0060] Figure 3 is a schematic diagram of numerical simulation results of leakage of impermeable wall defects according to an embodiment of the present application;

[0061] Figure 4 is a particle scatter diagram of the PSO algorithm shown in an embodiment of the present application;

[0062] Figure 5 is a schematic diagram of a knowledge graph for diagnosing defects in an anti-seepage wall of a high earth-rockfill dam according to an embodiment of the present application; DETAILED DESCRIPTION

[0063] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0064] In view of the problems that the selection of the above-mentioned traditional detection schemes has, such as lack of reference standards, strong subjectivity and blindness, this application proposes a method for recommending anti-seepage wall defect detection and diagnosis schemes based on machine learning. First, the position and form of the anti-seepage wall defects are judged by the PSO-SVM algorithm, and then a knowledge graph is established and similarity is calculated, finally realizing the recommendation of the detection method for the anti-seepage wall defects of high earth-rock dams.

[0065] The following describes a method for recommending anti-seepage wall defect detection and diagnosis solutions based on machine learning in an embodiment of the present application with reference to the accompanying drawings.

[0066] Figure 1 and Figure 2 A simplified flowchart and a detailed flowchart of a method for detecting and diagnosing impermeable wall defects based on machine learning are shown in the embodiments of the present application, as shown in FIG. Figure 1 As shown, the method comprises the following steps:

[0067] Step 101, calculate the seepage field and leakage under different defect conditions by numerical simulation method, and form a data set as a machine learning sample, wherein the defect condition includes defect position and defect form.

[0068] Specifically, step 101 also includes:

[0069] Step 11. Referring to relevant literature, engineering data and the damage mechanism of the cut-off wall, two defect forms and five defect positions are proposed to be combined to simulate the leakage under different cut-off wall defects.

[0070] In a possible embodiment, the two defect forms are local leakage and structural damage, and the five defect locations are respectively the connection between the anti-seepage wall and the bedrock, 1 / 4 of the height of the anti-seepage wall above the bedrock, 1 / 2 of the height of the anti-seepage wall above the bedrock, 3 / 4 of the height of the anti-seepage wall above the bedrock, and the connection between the anti-seepage wall and the dam body.

[0071] Step 12: According to the calculation conditions of different combination results, establish the finite element calculation model of defect leakage of high earth-rock dam cut-off wall respectively, and set the calculation parameters.

[0072] It should be noted that the calculation parameters need to be set according to the actual scenario, and this application does not make any specific limitations on this.

[0073] Step 13: numerically simulate the finite element calculation model of defect leakage in the high earth-rockfill dam cut-off wall, extract and organize the calculation results of the seepage field and leakage to form a data set, and obtain a machine learning sample.

[0074] It can be understood that before numerically simulating the finite element calculation model of defect leakage in the high earth-rock dam cut-off wall, it is necessary to determine the results to be extracted under different leakage conditions.

[0075] In particular, in engineering practice, leakage volume and seepage field are generally used as the main reference for leakage conditions. Accordingly, in this application, leakage volume and seepage field are also used as the main reference indicators for leakage conditions.

[0076] Furthermore, after the calculation is completed, the pore water pressure of n characteristic points in the space is extracted under the finite element calculation model of defect leakage of the high earth-rock dam anti-seepage wall under different calculation conditions to characterize the characteristics of the overall seepage field, and together with the seepage volume, it represents the leakage situation under the calculation conditions.

[0077] In a possible embodiment, a schematic diagram of numerical simulation results of leakage of impermeable wall defects is shown in FIG. Figure 3 shown.

[0078] Step 102, based on the particle swarm optimization algorithm PSO and the support vector machine algorithm SVM, a high earth-rock dam cut-off wall defect diagnosis model is obtained according to machine learning sample training to fit the implicit relationship between the defect location, defect type and leakage situation of the high earth-rock dam cut-off wall.

[0079] In the embodiment of the present application, before performing SVM parameter optimization, the parameters that need to be optimized are determined.

[0080] In a possible embodiment, the SVM parameters may be parameters such as regularization parameters and kernel function parameters, which are not specifically limited in the present application.

[0081] Then, the optimization range and different levels of the above parameters are confirmed, and the root mean square error is used as the objective function. The PSO-SVM classification model is trained according to the machine learning samples, and the model capability of the model is verified according to the test set. After the training, the defect diagnosis model of the anti-seepage wall of the high earth-rock dam is obtained.

[0082] It can be understood that by inputting the leakage situation of the high earth-rock dam into the trained high earth-rock dam anti-seepage wall defect diagnosis model, the defect location and defect type can be output, and then based on the knowledge graph constructed in the subsequent steps, the defect diagnosis plan for the specific defect location and specific damage type can be recommended.

[0083] Specifically, the training process includes the following steps:

[0084] Step 21: Divide the machine learning samples into a training set and a test set in a preset ratio, use the leakage amount and seepage field as the input of the PSO-SVM classification model, and use the defect location and defect type as the output of the PSO-SVM classification model.

[0085] In a possible embodiment, the machine learning samples are divided into a training set and a test set in a ratio of 4:1.

[0086] It is understood that the test set should contain data distribution similar to the training set, but it must be new data that the model has not seen during training.

[0087] Step 21, train the PSO-SVM classification model, establish a mapping relationship, use the PSO algorithm to optimize the parameters of the SVM until the optimal model is obtained, and use the optimal model as the defect diagnosis model of the anti-seepage wall of the high earth-rock dam.

[0088] In a possible embodiment, the particle scatter diagram of the PSO algorithm is as follows: Figure 4 shown.

[0089] As a possible implementation method, the PSO algorithm is used to find the optimal SVM parameter combination in the set search space, and each particle represents a set of possible SVM parameter configurations. In addition, in each iteration, the position and velocity of the particle are updated according to the SVM model performance (such as cross-validation accuracy) corresponding to the current position of the particle. The iteration stops when the predetermined maximum number of iterations is reached or other stopping conditions are met, thereby obtaining the optimal model.

[0090] Step 23: Verify the model capability of the high earth-rock dam cut-off wall defect diagnosis model based on the test set.

[0091] In the embodiment of the present application, the accuracy of the model is evaluated by comparing the model prediction results with the actual defect locations and defect types in the test set. Among them, a variety of evaluation indicators, such as accuracy, precision, recall, F1 score, etc., can be used to comprehensively evaluate the model performance.

[0092] It should be noted that if the model performs well on the test set, it means that the model has good generalization ability and can be used to diagnose defects in high earth-rock dam cut-off walls in practical applications. If the model performs poorly, it is necessary to readjust the model parameters or optimize the algorithm until satisfactory performance is achieved.

[0093] Step 103, constructing a cut-off wall defect and detection case library based on engineering data and literature research, and forming a cut-off wall defect detection rule library based on the case library.

[0094] Specifically, step 103 further includes the following steps:

[0095] Step 31: Through case analysis and mechanism analysis, sort out and summarize the possible defects of the anti-seepage wall.

[0096] In this step, we collected several cases of anti-seepage wall failure in actual projects, combined with physical and chemical mechanism analysis, analyzed the failure causes and processes in detail, and summarized the common defect forms, such as leakage, cracks, structural deformation, material degradation, local defects, etc.

[0097] Step 32: Obtain cases of defects in the cut-off wall of high earth-rock dams through literature research, expert consultation and engineering data collection methods, and determine the main influencing factors of the cut-off wall defects.

[0098] In this step, we collected relevant literature on the research and construction of cut-off walls at home and abroad, analyzed typical cases of cut-off wall defects, and summarized the problems in its design, construction and operation. We also interviewed experts in the field to obtain in-depth insights into common problems and technical experience in the operation of cut-off walls. We also combined first-hand information such as engineering construction files and test reports to clarify the main factors affecting the performance of cut-off walls.

[0099] Among them, the main factors include geological conditions, construction conditions, material selection, construction technology, operating environment, etc.

[0100] Step 33: Summarize the characteristics, applicability and application cases of high earth-rock dam detection schemes through literature research, and build a case library for high earth-rock dam anti-seepage wall defect detection.

[0101] Among them, the high earth-rock dam anti-seepage wall defect detection case library includes dam height, anti-seepage wall type, defect form, defect cause, detection method and related detection parameters, and the related detection parameters are point spacing, measuring line length, position, etc.

[0102] Step 34: construct a high earth-rock dam cut-off wall defect detection rule library based on the high earth-rock dam cut-off wall defect detection case library.

[0103] Among them, based on the defect detection case library of high earth-rock dam cut-off wall, defect detection related information is obtained; based on the defect detection related information, the detection requirements and detection targets are determined, and the applicability of each detection method is analyzed; based on the above analysis results, the relationship between the defect location and the detection scheme is analyzed to obtain the relationship analysis results, and based on the relationship analysis results, the cut-off wall detection rules are proposed to complete the construction of the defect detection rule library for the high earth-rock dam cut-off wall.

[0104] Step 104, forming a knowledge graph for defect diagnosis of the cut-off wall of a high earth-rockfill dam based on the constructed case library and rule library, and recommending defect diagnosis solutions for specific defect locations and specific damage types.

[0105] Specifically, step 104 also includes:

[0106] Step 41: Store the data in the constructed case base and rule base through the Neo4j graph database to form a knowledge graph for defect diagnosis of high earth-rock dam anti-seepage wall.

[0107] In a possible embodiment, a schematic diagram of a knowledge graph for diagnosing defects in an anti-seepage wall of a high earth-rock dam is as follows: Figure 5 shown.

[0108] Furthermore, the process of constructing a knowledge graph for high earth-rock dam impermeable wall defect diagnosis includes:

[0109] Step 411: perform data structuring processing on the unstructured data in the relevant data information in the constructed case library and rule library to construct a model layer of the knowledge graph for defect diagnosis of high earth-rock dam anti-seepage wall.

[0110] As a possible implementation method, text descriptions, pictures, reports and other forms of unstructured data are extracted from the case base and rule base, and these data are converted into structured information using natural language processing technology, data mining and machine learning methods. The basic pattern layer of the knowledge graph is defined, including nodes (such as defect type, influencing factors, detection methods), relationships (such as "cause", "detection"), attributes (such as "severity", "frequency of occurrence"), etc. Ensure that the pattern layer can fully cover all aspects of the defects of the high earth-rock dam cut-off wall, laying the foundation for the construction of the knowledge graph.

[0111] Step 412: Based on the pattern layer, data structuring is performed on the constructed case library to construct a data layer of the knowledge graph for defect diagnosis of the anti-seepage wall of a high earth-rock dam.

[0112] As a possible implementation method, the instance data in the case library is structured according to the definition of the pattern layer. This includes extracting and standardizing key elements in the data, such as defect type, test results, repair measures, etc., and matching this information with the definition in the pattern layer. Using data cleaning, integration and transformation tools, the structured data is input into the data layer of the knowledge graph to ensure the accuracy and consistency of the information. This process may involve data annotation, classification and normalization operations to facilitate the subsequent construction of the knowledge graph.

[0113] Step 413: Based on the data layer, the extracted knowledge is stored in the Neo4j database in the form of triples, and finally a complete knowledge graph for defect diagnosis of the anti-seepage wall of a high earth-rock dam is constructed.

[0114] Finally, the information in the data layer is converted into triples and stored in the knowledge graph database. The subject in the triples can be a specific defect type or detection method, the relationship is the defined operation or impact, and the object is the related attribute or result.

[0115] Step 42: input the leakage situation of the high earth-rock dam into the trained high earth-rock dam anti-seepage wall defect diagnosis model, and output the defect location and defect type.

[0116] Step 43: Based on the output defect location and defect type, the high earth-rock dam cut-off wall defect diagnosis knowledge graph is searched and scheme similarity matching is performed, and a recommended detection scheme is obtained based on the matching results.

[0117] In the embodiment of the present application, based on the identified defect location and defect type, the query function of the knowledge graph is used to search. The search will identify historical cases, influencing factors, and existing detection solutions similar to the current defect through the triple relationship in the graph. Then, the similarity matching algorithm is used to calculate the similarity between the defect characteristics and the existing cases to determine the solution closest to the current defect. Finally, based on the similarity matching results, a set of detection solutions that best suit the current defect is recommended, thereby realizing the recommendation of defect diagnosis solutions for specific defect locations and specific damage types.

[0118] It should be noted that multiple groups of defect diagnosis solutions can be recommended according to actual scenario needs, and this application does not make any specific limitations on this.

[0119] Specifically, the characteristic attribute similarity calculation formula of numerical data is:

[0120]

[0121] Among them, C ik For the i-th item C i The kth characteristic attribute, C 0k is the kth feature attribute of the reference item C0, a ik is the value of the kth feature attribute of the ith item, a 0k The value of the kth feature attribute of the reference item, maxa ik and mina ik It represents the maximum and minimum values ​​of the kth feature attribute in all cases or specific cases, and is used for normalization to calculate the relative difference of the values.

[0122] The formula for calculating the similarity of character-type feature attributes is:

[0123]

[0124] It should be noted that a ik and a 0k It still represents the numerical value of the feature attribute, but here it refers to character data. And the similarity sim(C ik ,C 0k ) is a binary function in character data. If the two attribute values ​​are equal, it is 1 (completely similar); if they are not equal, it is 0 (not similar).

[0125] The similarity calculation formula for the project is:

[0126]

[0127] Among them, g k Represents the weight of the kth feature attribute, which is used to weight the importance of different features when calculating the overall similarity; sim k Indicates the similarity of the kth feature attribute, which may be a numeric or character similarity, depending on the type of the feature; SIM(C0,C i ) represents the overall similarity between two items, which is the weighted sum of the similarities of all feature attributes.

[0128] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this application can be executed in parallel, sequentially or in different orders, as long as the expected results of the technical solution of this application can be achieved, and this document is not limited here.

[0129] The above specific implementations do not constitute a limitation on the protection scope of this application. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application should be included in the protection scope of this application.

Claims

1. A method for detecting and diagnosing impermeable wall defects based on machine learning, characterized in that: include: Calculating the seepage field and leakage under different defect conditions by numerical simulation method to form a data set as a machine learning sample, wherein the defect conditions include defect location and defect form; Based on the particle swarm optimization algorithm PSO and the support vector machine algorithm SVM, a high earth-rock dam cut-off wall defect diagnosis model is obtained according to the machine learning sample training to fit the implicit relationship between the defect location, defect type and leakage situation of the high earth-rock dam cut-off wall; According to engineering data and literature research, a case library of impermeable wall defects and detection is constructed, and a rule library for impermeable wall defect detection is formed based on the case library; Based on the constructed case library and rule library, a knowledge graph for defect diagnosis of the anti-seepage wall of high earth-rock dam is formed to recommend defect diagnosis solutions for specific defect locations and specific damage types.

2. The method according to claim 1, characterized in that The method of calculating the seepage field and leakage under different defect conditions by numerical simulation to form a data set as a machine learning sample includes: With reference to relevant literature, engineering data and the damage mechanism of the cut-off wall, two defect forms and five defect positions were proposed to simulate the leakage under different cut-off wall defects. Among them, the two defect forms are local leakage and structural damage, and the five defect positions are the connection between the cut-off wall and the bedrock, 1 / 4 of the height of the cut-off wall above the bedrock, 1 / 2 of the height of the cut-off wall above the bedrock, 3 / 4 of the height of the cut-off wall above the bedrock, and the connection between the cut-off wall and the dam body. According to the calculation conditions of different combination results, the finite element calculation model of high earth-rock dam impermeable wall defect leakage is established and the calculation parameters are set; The finite element calculation model of defect leakage in the high earth-rockfill dam cut-off wall is numerically simulated, and the calculation results of the seepage field and leakage amount are extracted and sorted to form a data set to obtain a machine learning sample.

3. The method according to claim 2, characterized in that The method of numerically simulating the finite element calculation model of the defect leakage of the high earth-rock dam cut-off wall also includes: Before calculation, determine the results to be extracted under different leakage conditions, and use the leakage volume and seepage field as the main reference indicators for evaluating the leakage situation; After the calculation is completed, under the finite element calculation model of defect leakage of high earth-rock dam cut-off wall under different calculation conditions, the pore water pressure of n characteristic points in the space is extracted to characterize the overall seepage field characteristics, and combined with the leakage amount, it jointly reflects the leakage status under the calculation conditions.

4. The method according to claim 1, characterized in that: The defect diagnosis model of the high earth-rock dam cut-off wall is obtained by training the machine learning samples based on the particle swarm optimization algorithm and the support vector machine algorithm, including: Before optimizing SVM parameters, determine the parameters that need to be optimized, including regularization parameters and kernel function parameters; Confirm the optimization range and different levels of the above parameters; The root mean square error is used as the objective function, the PSO-SVM classification model is trained according to the machine learning samples, and the generalization ability of the model is verified according to the test set. After the training, the defect diagnosis model of the high earth-rock dam anti-seepage wall is obtained.

5. The method according to claim 4, characterized in that The training of the PSO-SVM classification model according to the machine learning sample includes: Dividing the machine learning samples into a training set and a test set in a preset ratio, taking the leakage amount and the seepage field as the input of the PSO-SVM classification model, and taking the defect position and the defect type as the output of the PSO-SVM classification model; Training the PSO-SVM classification model, establishing a mapping relationship, optimizing the parameters of the SVM using the PSO algorithm until an optimal model is obtained, and using the optimal model as a defect diagnosis model for the high earth-rock dam anti-seepage wall; The model capability of the high earth-rock dam cut-off wall defect diagnosis model is verified based on the test set.

6. The method according to claim 1, characterized in that The impermeable wall defect and detection case library is constructed based on engineering data and literature research, and an impermeable wall defect detection rule library is formed based on the case library, including: Through case analysis and mechanism analysis, the possible defects of the anti-seepage wall are summarized; Through literature research, expert consultation and engineering data collection methods, we obtained cases of defects in the cut-off wall of high earth-rock dams and determined the main influencing factors of the defects in the cut-off wall. Through literature research, we summarized the characteristics, applicability and application cases of the detection scheme of high earth-rock dams and built a case library of defects detection of the cut-off wall of high earth-rock dams. A high earth-rock dam cut-off wall defect detection rule library is constructed based on the high earth-rock dam cut-off wall defect detection case library.

7. The method according to claim 6, characterized in that The method of constructing a high earth-rock dam cut-off wall defect detection rule library based on the high earth-rock dam cut-off wall defect detection case library comprises: Obtaining defect detection related information based on the high earth-rock dam cut-off wall defect detection case library; Based on the defect detection related information, determine the detection requirements and detection objectives, and analyze the applicability of each detection method; Based on the above analysis results, the relationship between the defect location and the detection scheme is analyzed to obtain a relationship analysis result, and based on the relationship analysis result, an anti-seepage wall detection rule is proposed to complete the construction of the high earth-rock dam anti-seepage wall defect detection rule base.

8. The method according to claim 1, characterized in that The defect diagnosis knowledge graph of the high earth-rock dam cut-off wall is formed according to the constructed case library and rule library, so as to recommend defect diagnosis solutions for specific defect locations and specific damage types, including: The data in the constructed case base and rule base are stored in a Neo4j graph database to form a knowledge graph for defect diagnosis of the high earth-rock dam impermeable wall; The leakage situation of the high earth-rock dam is input into the trained high earth-rock dam anti-seepage wall defect diagnosis model, and the defect location and defect type are output; According to the output defect locations and defect types, the high earth-rock dam cut-off wall defect diagnosis knowledge graph is searched and scheme similarity matching is performed, and a recommended detection scheme is obtained based on the matching results.

9. The method according to claim 8, characterized in that The data in the constructed case base and rule base are stored in a Neo4j graph database to form a knowledge graph for defect diagnosis of the high earth-rock dam cut-off wall, including: Based on the unstructured data in the relevant data information in the constructed case library and rule library, data structuring is performed to construct the model layer of the knowledge graph for defect diagnosis of the anti-seepage wall of high earth-rock dam; Based on the pattern layer, data structuring is performed on the constructed case library to construct a data layer of the knowledge graph for defect diagnosis of the anti-seepage wall of a high earth-rock dam; Based on the data layer, the extracted knowledge is stored in the Neo4j database in the form of triples, and finally a complete knowledge graph for defect diagnosis of the high earth-rock dam cut-off wall is constructed.

10. The method according to claim 8, characterized in that The step of searching the defect diagnosis knowledge graph of the high earth-rock dam cut-off wall according to the output defect location and defect type and matching the scheme similarity includes: The calculation formula for the similarity of the characteristic attributes of numerical data is: Among them, C ik For the i-th item C i The kth characteristic attribute, C 0k is the kth feature attribute of the reference item C0, a ik is the value of the kth feature attribute of the ith item, a 0k The value of the kth feature attribute of the reference item; The formula for calculating the similarity of character-type feature attributes is: The similarity calculation formula for the project is: Among them, g k represents the weight of the kth feature attribute, sim k Represents the similarity of the kth feature attribute.

Images