Physical-data hybrid driven water supply network operation safety diagnosis method and system

Abstract

This invention provides a physical-data hybrid-driven method and system for diagnosing the operational safety of water supply networks. The safety diagnosis method includes steps such as operation and maintenance data collection, extended physical feature generation, data fusion, physical state deduction, hybrid-driven decision-making, and early warning. The system includes modules corresponding to each step. By extending physical features, this invention enables the system to effectively capture the real physical mechanisms of water supply pipeline damage. It utilizes a category-enhancing intelligent algorithm to fuse physical mechanisms with operational data, constructing a physical enhanced feature space, thus achieving accurate and robust prediction of pipeline damage events. Simultaneously, this method quantitatively evaluates the contribution and mechanism of extended physical features and other operation and maintenance data features to the probability of pipeline damage, enhancing the physical interpretability of the model and making the decision-making process more transparent and consistent with domain knowledge, providing a scientific basis for proactive maintenance and asset management of water supply networks.

Description

Technical Field

[0001] This invention relates to the field of urban lifeline engineering safety technology, specifically to a physical-data hybrid driven method and system for diagnosing the operational safety of water supply networks. Background Technology

[0002] Water supply networks are vital lifelines for maintaining the daily operation and sustainable development of cities. With the continuous advancement of urbanization, leaks and bursts in underground water supply networks occur frequently, causing significant waste of water resources and substantial economic losses. In some cases, they can even lead to cascading damage, resulting in secondary infrastructure safety incidents such as road collapses and traffic disruptions, seriously threatening urban public safety.

[0003] Due to the concealed and uncertain nature of the underground environment, the mechanisms underlying pipe damage in water supply networks are extremely complex, typically resulting from the coupled effects of multiple factors, including pipe material performance degradation, external environmental loads, and internal operating pressures. Currently, pipeline operation safety risk analysis mainly relies on two types of methods: physical models and data-driven machine learning. Physical models can intuitively explain the physical mechanisms of pipe damage, but are limited by idealized mechanical assumptions and cannot fully cover the complex and ever-changing actual operating conditions. Data-driven methods, on the other hand, possess powerful nonlinear modeling capabilities and can effectively extract pipe damage characteristics from operation and maintenance data, but they heavily depend on data quality and sample completeness, and have poor physical interpretability. Summary of the Invention

[0004] This invention is made to solve the above-mentioned problems. Its purpose is to provide a physical-data hybrid driven method and system for diagnosing the operational safety of water supply networks. It deeply integrates physical and mechanical mechanisms with operation and maintenance data characteristics, accurately captures the complex nonlinear coupling laws in the pipeline damage process, accurately predicts and forecasts water supply network damage events, and systematically explains the degree and law of influence of each influencing factor on the probability of pipeline damage.

[0005] This invention provides a physical-data hybrid-driven method for diagnosing the operational safety of water supply networks. The method includes: a maintenance data collection step, which involves determining the scope of the water supply network in a target area, acquiring maintenance data of the network within that area, and extracting maintenance data features; an extended physical feature step, which calculates the structural safety factor of the pipelines in the water supply network using a pre-established physical-mechanical model of the water supply pipelines, and stores the structural safety factor as an extended physical feature in a pre-set extended physical feature set; and a data fusion step, which fuses the maintenance data features and extended physical features to obtain hybrid enhanced features, stores these features in a pre-set hybrid enhanced feature set, performs data preprocessing and correlation analysis on the hybrid enhanced features, obtains the analysis results, and stores them in a pre-set... The established hybrid-driven prediction model includes the following steps: A physical state deduction step, which uses a physical-mechanical model of the water supply pipeline to deduce the dynamic evolution trend of the physical state of the water supply network based on real-time operation and maintenance data; a hybrid-driven decision-making step, which integrates real-time operation and maintenance data with dynamic evolution trend data, extracts the features of the integrated data, and uses the hybrid-driven prediction model to dynamically predict the probability of pipeline loss within the target area, while calculating the marginal contribution of each data feature to the current prediction result in real time; and an early warning step, which determines the risk level of pipeline loss based on the probability of loss and triggers corresponding early warning signals, identifies key risk factors based on marginal contributions, matches corresponding maintenance suggestions from the strategy library in the hybrid-driven prediction model, generates early warning work orders, and pushes them out.

[0006] The physical-data hybrid-driven water supply network operation safety diagnosis method provided by this invention may also have the following features: the establishment and training process of the hybrid-driven prediction model includes: based on operation and maintenance data and an extended physical feature set, using whether a pipeline has experienced a pipe loss event as the sample classification criterion, classifying pipe loss sample types, and storing them in a preset pipe loss sample dataset; using a portion of the pipe loss sample dataset as a training set and the remaining portion as a test set, and performing data augmentation on the training set; using a class boosting algorithm to capture the nonlinear relationships between the fused features, storing them in the hybrid-driven prediction model, and iteratively optimizing the hyperparameters in the hybrid-driven prediction model using a Bayesian optimization algorithm; evaluating the reliability and cost-effectiveness of completing the training of the hybrid-driven prediction model on the test set; calculating the marginal contribution of each feature of each pipe loss sample in the test set to the prediction result, determining the direction of influence of each feature on the pipe loss probability, ranking the importance of each feature on the pipe loss probability, judging the effectiveness of the pipe loss prediction based on the importance of the influence, and comparing the consistency of the influence direction with the objective physical laws to evaluate the physical interpretability of the hybrid-driven prediction model.

[0007] The physical-data hybrid-driven water supply network operation safety diagnosis method provided by this invention may also have the following features: the operation and maintenance data collection step includes: defining the target area range and dividing the water supply network in the target area into several independent evaluation units according to pipe sections; collecting and organizing historical pipe damage repair records and daily operation and maintenance data of the water supply network in the target area; extracting three types of features of the operation and maintenance data, namely, basic network attributes, operating parameters and surrounding environment, and constructing an operation and maintenance data feature set.

[0008] The physical-data hybrid-driven water supply network operation safety diagnosis method provided by this invention may also have the following features: the historical pipe damage maintenance record includes: the specific time, detailed location, pipe damage type and accident cause of the pipe damage event; and the daily operation and maintenance data includes: one or more of the following: pipe network geographic information system data, pipe network monitoring and data acquisition system data, pipe perimeter geological condition data, road traffic data, and meteorological condition data.

[0009] The physical-data hybrid-driven water supply network operation safety diagnosis method provided by this invention may also have the following features: among the three types of features of operation and maintenance data, the basic attribute features include pipe material, pipe diameter, pipe burial depth, pipe length and pipe service life; the operation parameter features include historical pipe damage frequency, pipe management office and water treatment plant; and the surrounding environment features include the type of soil around the pipe, the grade of the overlying road, the distance to the nearest intersection and the angle of intersection with the road.

[0010] The physical-data hybrid-driven water supply network operation safety diagnosis method provided by this invention may also have the following features: the extended physical feature step includes: calculating the structural stress of the pipeline under five operating loads using a physical and mechanical model of the water supply pipeline, where the five loads are: soil pressure, traffic load, internal water pressure, temperature load, and freezing load; determining the combined effect of the five operating loads, and calculating the biaxial stress components of the circumferential and axial stresses of the pipeline under the combined action of the five operating loads; calculating the depth of the corrosion pit and the residual wall thickness after corrosion of the pipeline, and estimating the remaining strength of the pipeline after corrosion; comparing the biaxial stress components and the remaining strength of the pipeline, calculating the structural safety factor of the pipeline, and storing it in the extended physical feature set.

[0011] The physical-data hybrid-driven water supply network operation safety diagnosis method provided by this invention may also have the following features: in the data fusion step, the hybrid enhancement feature set includes discrete features and continuous features in terms of data type; data preprocessing of the hybrid enhancement features includes label encoding of discrete features and standardization of continuous features; correlation analysis includes evaluating the correlation between each hybrid enhancement feature and identifying and filtering out strongly correlated features.

[0012] The physical-data hybrid-driven water supply network operation safety diagnosis method provided by this invention may also have the following features: discrete features include: pipe material, pipe management office, water treatment plant, surrounding soil type, and overlying road grade; continuous features include structural safety factor, pipe diameter, pipe burial depth, pipe length, pipe service life, historical pipe damage frequency, distance to adjacent intersections, and intersection angle with roads.

[0013] This invention also provides a physical-data hybrid driven water supply network operation safety diagnosis system for performing the aforementioned water supply network operation safety diagnosis method. It features the following components: an operation and maintenance data collection module for determining the scope of the water supply network in a target area, acquiring operation and maintenance data of the water supply network within that area, and extracting operation and maintenance data features; an extended physical feature module for calculating the structural safety factor of the pipelines in the water supply network using a pre-established physical and mechanical model of the water supply pipelines, and storing the calculated structural safety factor in a pre-set extended physical feature set; and a data fusion module for fusing the operation and maintenance data features and extended physical features, and performing data preprocessing and correlation analysis on the fused features. The system obtains analysis results and stores them in a pre-established hybrid-driven prediction model; the physical state simulation module is used to simulate the evolution trend of the physical state of the water supply network; the hybrid-driven decision-making module integrates real-time operation and maintenance data and physical state evolution trend data, calls the hybrid-driven prediction model to dynamically predict the probability of pipe loss in the target area, and calculates the marginal contribution of each feature to the current prediction result in real time; the advanced early warning module is used to determine the risk level of pipe loss based on the probability of pipe loss and trigger corresponding early warning signals, and identify key risk factors based on the marginal contribution of features, match corresponding maintenance suggestions from the strategy library in the hybrid-driven prediction model, generate early warning work orders, and push them.

[0014] Compared with the prior art, the functions and effects of the present invention include:

[0015] This invention collects operation and maintenance data of water supply networks and calculates the safety factor of pipeline structures based on structural mechanics principles. This enables the system to effectively capture the real physical mechanisms of water supply pipeline damage, overcoming the prediction bias caused by sparse, noisy, or poor-quality network operation and maintenance data in existing technologies. It achieves accurate and robust prediction of pipeline damage events under limited data conditions.

[0016] This invention integrates physical modeling and data mining techniques by combining operational data features and extended physical features, significantly improving the physical interpretability of data-driven models. It effectively solves the "black box" problem of traditional data-driven models being disconnected from actual working conditions, making the decision-making process more transparent and in line with domain knowledge.

[0017] Therefore, this invention can provide reliable technical support for water industry and city managers to formulate scientific proactive maintenance strategies and optimize resource allocation. Attached Figure Description

[0018] Figure 1 This is a flowchart of the physical-data hybrid driven water supply network operation safety diagnosis method of the present invention;

[0019] Figure 2 This is a schematic diagram of the water supply network and the distribution of historical pipe loss events in an embodiment of the present invention;

[0020] Figure 3 This is a schematic diagram of the elastic foundation beam-soil interaction model in an embodiment of the present invention;

[0021] Figure 4 This is a feature correlation heatmap in an embodiment of the present invention;

[0022] Figure 5 This is a graph showing the marginal contribution of features to the probability of pipe loss in an embodiment of the present invention.

[0023] Figure 6 This is an architecture diagram of the water supply network operation safety diagnosis system in an embodiment of the present invention.

[0024] In the diagram, 1. Operation and maintenance data collection module; 2. Extended physical characteristics module; 3. Data fusion module; 4. Physical state inference module; 5. Hybrid-driven decision-making module; 6. Advanced early warning module; 10. System. Detailed Implementation

[0025] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection, an electrical connection, or a connection that allows communication between them; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication between two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0026] To make the technical means, creative features, objectives and effects of the present invention easy to understand, the following embodiments, in conjunction with the accompanying drawings, specifically illustrate the physical-data hybrid driven water supply network operation safety diagnosis method and system of the present invention.

[0027] Figure 1 This is a flowchart of the physical-data hybrid driven water supply network operation safety diagnosis method of the present invention. Figure 1 As shown: This embodiment provides a physical-data hybrid driven method for diagnosing the operational safety of a water supply network, including the following steps S1 to S6.

[0028] Step S1, the operation and maintenance data collection step, involves determining the scope of the water supply network in the target area, acquiring the operation and maintenance data of the water supply network within that area, and extracting the operation and maintenance data features. This step S1 includes steps S1-1 to S1-3.

[0029] Specifically, in this embodiment, step S1-1: taking the water supply network of a central urban area of ​​a certain city as the object, the water supply network of the target area is divided into 103,791 pipes according to the pipe segments, that is, 103,791 evaluation units.

[0030] Step S1-2: Obtain 5942 historical pipe damage repair records from the water industry department, as well as pipeline geographic information system (GIS) data, pipeline monitoring and data acquisition system (SCADA) data, pipeline surrounding geological condition data, and road traffic data. The distribution of water supply network and pipe damage events provided in this embodiment is as follows: Figure 2 As shown;

[0031] Steps S1-3: Combining pipeline network operation and maintenance data, construct an operation and maintenance data feature set, including five basic attribute features, three operational parameter features, and four surrounding environment features, totaling twelve operation and maintenance data features. The basic attribute features are pipeline material, pipeline diameter, pipeline burial depth, pipeline length, and pipeline service life. The operational parameter features are historical pipeline damage frequency, pipeline management office, and water treatment plant. The surrounding environment features are the type of soil around the pipeline, the grade of the overlying road, the distance to the nearest intersection, and the angle of intersection with the road.

[0032] Step S2, the extended physical feature step, involves calculating the structural safety factor of the pipelines in the water supply network using a pre-established physical and mechanical model of the water supply pipeline, and storing this structural safety factor as an extended physical feature in a preset extended physical feature set. This step S2 includes steps S2-1 to S2-4.

[0033] Specifically, Figure 3 This is a schematic diagram of the elastic foundation beam-tube-soil interaction model in this embodiment. Figure 3 As shown, step S2-1 involves constructing an elastic foundation beam-pipe-soil interaction model, i.e., a physical and mechanical model of the water supply pipeline, to calculate the structural stress of the pipeline under five operating loads: earth pressure, traffic load, internal water pressure, temperature load, and freezing load. During the calculation, the soil unit weight is set to 18.5. The traffic load for main and secondary roads is set at 70 kN, and for branch roads at 52.5 kN. The internal water pressure range is set at 0.18 to 0.24 MPa, with variations in internal water pressure across different street blocks. The temperature difference between the water inside the pipe and the surrounding soil is set at -4.24℃. The freezing load factor is set at 0.1. Other parameters are determined based on pipeline operation and maintenance data. The calculation formula is as follows:

[0034]

[0035]

[0036]

[0037]

[0038]

[0039]

[0040] in, and Structural stress generated by internal water pressure; The structural stress generated by earth pressure; Structural stress generated by traffic loads; Structural stress generated by temperature load; Structural stress generated by freezing load; Where is Poisson's ratio; p is the internal water pressure; D is the pipe diameter; t is the pipe wall thickness; γ is the bending moment coefficient; γ is the weight of the soil around the pipe. The width of the trench; To calculate the coefficients; The elastic modulus of the pipe; The deflection coefficient; Impact factor; Traffic load factor; F is traffic load; A is pipe length; The coefficient of thermal expansion of the pipeline; This refers to the temperature difference between the water inside the pipe and the surrounding soil. This is the freezing load factor.

[0041] Step S2-2: Earth pressure, traffic load, internal water pressure, and temperature load generate circumferential stress on the pipeline structure, while freezing load generates axial stress. Considering the Poisson effect of pipe deformation, the biaxial stress components of the circumferential and axial stresses under combined loads are calculated. The calculation formula is as follows:

[0042]

[0043]

[0044] in, It is circumferential stress; This is axial stress.

[0045] Step S2-3: Set the soil corrosivity of the target area to the lowest level, with a minimum corrosion rate of 0.0042. The pitting depth scaling constant is 1.95 mm, and the corrosion rate inhibition factor is 0.058. Based on this, the depth of the corrosion pit and the residual wall thickness after corrosion occur in the pipeline are calculated, and the remaining strength of the defective pipeline is estimated by combining fracture mechanics. The calculation formula is expressed as follows:

[0046]

[0047]

[0048]

[0049] Where d is the depth of the corrosion pit; This represents the residual wall thickness. α represents the residual strength; T represents the service life of the pipeline; a represents the minimum corrosion rate; b represents the pitting depth scaling constant; c represents the corrosion rate inhibition factor; α and S represent the constants of the fracture toughness equation. For fracture toughness; This refers to the lateral dimension of the corrosion pit; and For a double-notched tensile specimen, the geometric factor constant is given.

[0050] Step S2-4: Compare the biaxial stress of the pipe , and residual strength Calculate the safety factor of the pipeline structure and store it in the extended physical feature set. The calculation formula is as follows:

[0051]

[0052] Wherein, FOS is the safety factor of the pipeline structure.

[0053] Step S3, the data fusion step, fuses the operation and maintenance data features and extended physical features to obtain hybrid enhanced features, which are then stored in a preset hybrid enhanced feature set. The hybrid enhanced features are preprocessed and subjected to correlation analysis to obtain the analysis results, which are then stored in a pre-established hybrid-driven prediction model. This step S3 includes steps S3-1 to S3-3.

[0054] Specifically, step S3-1: Using the unique identifier code of the pipeline as an index, twelve operation and maintenance data features and one extended physical feature of the pipeline are associated and fused to construct a hybrid enhanced feature set containing thirteen features. This hybrid enhanced feature set includes five discrete features and eight continuous features in terms of data type. Among them, the discrete features are pipeline material, pipeline management office, water treatment plant, surrounding soil type and overlying road grade, and the continuous features are structural safety factor, pipeline diameter, pipeline burial depth, pipeline length, pipeline service life, historical pipeline damage frequency, distance to adjacent intersections, and intersection angle with roads.

[0055] Step S3-2: Perform data preprocessing on the hybrid enhanced features, including filling in missing values ​​and removing outliers. Align all types of data in terms of time and spatial scale. Label and encode discrete features to make them analyzable, and standardize continuous features to eliminate differences in scale and dimension. After processing, a standardized dynamic operation and maintenance data stream is formed.

[0056] Step S3-3: Evaluate the correlation between the various hybrid enhancement features, identify and remove strongly correlated features with an absolute correlation coefficient higher than 0.7 to suppress the impact of multicollinearity on model performance. The feature correlation heatmap provided in this embodiment is shown below. Figure 4 The correlation coefficient between pipe burial depth and pipe diameter is 0.78, indicating a strong correlation. Therefore, in this embodiment, pipe diameter is excluded from the hybrid enhancement feature set.

[0057] Furthermore, in this embodiment, the process of establishing and training the hybrid-driven prediction model includes the following steps (1)-(3):

[0058] (1) Based on operation and maintenance data and extended physical feature set, the pipelines that have experienced pipe damage events are used as the basis for sample classification. Pipes that have experienced pipe damage in the target area are marked as positive samples, and pipes that have not experienced pipe damage are marked as negative samples. A pipe damage sample dataset covering all pipelines in the target area is constructed. The dataset is divided into training set and test set according to the proportion, and data augmentation is performed on the training set.

[0059] Specifically, in this embodiment, the 5942 pipe damage events in the target area are distributed across 5051 pipes. Some of these pipes experienced multiple pipe damage events, while the remaining 98740 pipes did not experience any damage. The 5942 pipe damage events are used as positive samples, and the 98740 pipes that did not experience any damage are used as negative samples, thus constructing a pipe damage sample dataset with a sample size of 104682.

[0060] Then, the pipeline damage sample dataset was randomly divided into a training set and a test set in a 4:1 ratio. The number of positive samples and negative samples in the training set were 4762 and 78983, respectively, while the number of positive samples and negative samples in the test set were 1180 and 19757, respectively.

[0061] The training set was augmented using a synthetic minority oversampling method. 74,221 new synthetic tube loss positive samples were generated by interpolation to balance the class distribution of positive and negative samples in the training set. The number of positive and negative samples in the augmented training set was 78,983 each.

[0062] (2) The category boosting algorithm is used to capture the nonlinear relationship between the hybrid boosting features, store them in the hybrid driven prediction model, and evaluate the reliability and applicability of the model.

[0063] Specifically, based on the enhanced training set, the hybrid-driven prediction model with improved categories is trained. The classification threshold is set to 0.5, and the Bayesian optimization algorithm is used to optimize the model hyperparameters. The optimal parameter combination is as follows: 1000 iterations, tree depth of 4, learning rate of 0.1, L2 regularization coefficient of 3, and loss function of Logloss.

[0064] The reliability of the trained hybrid-driven prediction model was evaluated on the test set. Three performance metrics were calculated based on the confusion matrix: accuracy, recall, and AUC value under the ROC curve. The values ​​of these three metrics range from 0 to 1, with higher values ​​indicating better model reliability. In this embodiment, the calculated results for accuracy, recall, and AUC value under the ROC curve were 0.9497, 0.8602, and 0.9714, respectively, verifying that the constructed hybrid model possesses excellent classification accuracy and the ability to identify positive samples with poor performance.

[0065] The cost-effectiveness of the trained hybrid-driven prediction model is evaluated on the test set. Two performance metrics are calculated: maintenance rate and average maintenance rate under the maintenance rate-recall curve. Both metrics range from 0 to 1, with lower values ​​indicating better cost-effectiveness. In this embodiment, the maintenance rate and average maintenance rate are calculated to be 0.0909 and 0.0553, respectively, verifying that the constructed hybrid model has high cost-effectiveness and economic applicability in real-world maintenance scenarios. The maintenance rate calculation formula is expressed as:

[0066]

[0067] in, TP represents the maintenance rate; TN represents the number of pipe loss samples correctly predicted by the model; FP represents the number of pipe loss samples incorrectly predicted by the model; and FN represents the number of pipe loss samples incorrectly predicted by the model.

[0068] (3) Perform interpretability analysis on the constructed hybrid-driven prediction model, quantify the marginal contribution of each hybrid enhancement feature to the pipe loss probability, and evaluate the effectiveness of the extended physical features.

[0069] Specifically, using a game theory-based additive feature interpretation method, the marginal contribution of twelve features in the 20,937 samples of the test set to the probability of pipe loss is calculated, and the quantitative value of the role of each feature in the model decision for each sample is obtained.

[0070] Figure 5This visualization shows the distribution relationship between each feature value and its quantized impact value. Based on the average absolute value of each feature's quantized impact value across all samples in the test set, the features are ranked from highest to lowest importance, as shown in the following figure. Figure 5 As shown from top to bottom. In this embodiment, the three most important features are the structural safety factor, pipeline length, and pipeline service life. Among them, pipeline length and pipeline service life are positively correlated with the probability of pipeline loss, while the structural safety factor is negatively correlated with the probability of pipeline loss.

[0071] In this embodiment, the structural safety factor feature exhibits the highest degree of influence, confirming the effectiveness of integrating extended physical features into the data-driven model to predict pipeline loss. At the same time, the structural safety factor feature value is negatively correlated with its quantified value, which is consistent with objective physical laws. That is, the higher the structural safety factor, the higher the structural reliability and the lower the probability of pipeline loss, thus confirming the physical interpretability of the constructed hybrid model.

[0072] S4, Physical State Deduction Step: Based on real-time operation and maintenance data, the dynamic evolution trend of the physical state of the water supply network is deduced through the physical and mechanical model of the water supply pipeline.

[0073] Specifically, based on the real-time dynamic operation and maintenance data stream, the physical and mechanical model of the water supply pipeline uses hourly steps to extrapolate the dynamic evolution trend of the pipeline's physical state in future periods, based on the pipeline's current structural safety status.

[0074] S5, the hybrid-driven decision-making step, integrates real-time operation and maintenance data streams and dynamic evolution trend data, extracts the features of the integrated data, and calls the hybrid-driven prediction model to dynamically predict the pipe loss probability of 103,791 pipelines in the target area, and calculates the marginal contribution of each data feature to the current prediction result in real time.

[0075] S6, the proactive warning step, determines the risk level of pipeline loss based on the probability of pipeline loss and triggers the corresponding warning signal, and identifies key risk factors based on the marginal contribution calculated in step S5, matches the corresponding maintenance suggestions from the strategy library in the hybrid-driven prediction model, generates a warning work order and pushes it.

[0076] Specifically, four warning thresholds are pre-set: blue warning (0-0.3), yellow warning (0.3-0.5), orange warning (0.5-0.7), and red warning (0.7-1). Based on the threshold range of the probability of pipe loss, the system automatically determines the corresponding risk level and triggers the corresponding warning signal. At the same time, it identifies key risk factors by combining feature marginal contribution, matches corresponding maintenance suggestions from the strategy library, generates a warning work order containing the risk location, level, key factors, and maintenance measures, and pushes it to the water management terminal.

[0077] Figure 6This is an architecture diagram of the water supply network operation safety diagnosis system 10 in an embodiment of the present invention. Figure 6 As shown: This embodiment also provides a physical-data hybrid driven water supply network operation safety diagnosis system 10, which is used to perform the above-mentioned water supply network operation safety diagnosis method. It includes an operation and maintenance data collection module 1, an extended physical feature module 2, a data fusion module 3, a physical state inference module 4, a hybrid driven decision module 5, and an early warning module 6.

[0078] Among them, the operation and maintenance data collection module 1 is used to implement the above step S1, the extended physical feature module 2 is used to implement the above step S2, the data fusion module 3 is used to implement the above step S3, the physical state inference module 4 is used to implement the above step S4, the hybrid driving decision module 5 is the core computing unit of this system 10 and is used to implement the above step S5, and the advanced early warning module 6 is used to implement the above step S6.

[0079] The role and effect of the embodiments

[0080] This embodiment collects operation and maintenance data of the water supply network and calculates the safety factor of the pipeline structure based on the principles of structural mechanics. This enables System 10 to effectively capture the real physical mechanisms of water supply pipeline damage, overcoming the limitations of traditional pure data-driven models and pure physical models. By utilizing a category-enhancing intelligent algorithm to fuse physical mechanisms with operational data, a physical augmentation feature space is constructed, achieving accurate and robust prediction of pipeline damage events. Simultaneously, this technology quantitatively evaluates the contribution and mechanism of extended physical features and other operation and maintenance data features to the probability of pipeline damage, enhancing the physical interpretability of the model. This effectively solves the "black box" problem of traditional data-driven models being disconnected from actual operating conditions, making the decision-making process more transparent and consistent with domain knowledge, and providing a scientific basis for proactive maintenance and asset management of the water supply network.

[0081] Those skilled in the art should understand that this invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to this invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A physical-data hybrid-driven method for diagnosing the operational safety of a water supply network, characterized in that, include: The steps for collecting operation and maintenance data are as follows: determine the scope of the water supply network in the target area, obtain the operation and maintenance data of the water supply network in that area, and extract the characteristics of the operation and maintenance data. The extended physical feature step involves calculating the structural safety factor of the pipeline in the water supply network using a pre-established physical and mechanical model of the water supply pipeline, and storing the structural safety factor as an extended physical feature in a preset extended physical feature set. The data fusion step involves fusing the operation and maintenance data features and the extended physical features to obtain hybrid enhanced features, which are then stored in a preset hybrid enhanced feature set. The hybrid enhanced features are preprocessed and subjected to correlation analysis to obtain the analysis results, which are then stored in a pre-established hybrid-driven prediction model. The physical state deduction steps involve using real-time operation and maintenance data to deduce the dynamic evolution trend of the physical state of the water supply network through the physical and mechanical model of the water supply pipeline. The hybrid-driven decision-making process integrates the real-time operation and maintenance data and the dynamic evolution trend data, extracts the integrated data features, and uses the hybrid-driven prediction model based on these data features to dynamically predict the pipeline loss probability within the target area, and calculates the marginal contribution of each data feature to the current prediction result in real time. The advanced early warning step involves determining the risk level of pipeline damage based on the pipeline damage probability and triggering a corresponding early warning signal. It also identifies key risk factors based on the marginal contribution, matches corresponding maintenance suggestions from the strategy library in the hybrid-driven prediction model, generates an early warning work order, and pushes it out.

2. The physical-data hybrid driven method for diagnosing the operation safety of water supply networks according to claim 1, Its features are: in, The operation and maintenance data collection steps include: Define the target area and divide the water supply network of the target area into several independent evaluation units according to pipe sections; Collect and organize historical pipe damage repair records and daily operation and maintenance data of the water supply network in the target area; The three types of features of the operation and maintenance data are extracted, namely the basic attributes of the pipeline network, the operating parameters and the surrounding environment, and stored in a preset operation and maintenance data feature set.

3. The physical-data hybrid driven method for diagnosing the operational safety of water supply networks according to claim 2, characterized in that: in, The historical pipeline damage maintenance records include: the specific time, detailed location, type of pipeline damage, and cause of the incident. The daily operation and maintenance data includes one or more of the following: pipeline geographic information system data, pipeline monitoring and data acquisition system data, pipeline surrounding geological conditions data, road traffic data, and meteorological conditions data.

4. The physical-data hybrid driven method for diagnosing the operation safety of water supply networks according to claim 2, characterized in that: in, Among the three types of characteristics of the operation and maintenance data, the basic attribute characteristics include pipe material, pipe diameter, pipe burial depth, pipe length, and pipe service life; The operational parameter characteristics include historical pipeline damage frequency, pipeline management office, and water treatment plant; The surrounding environmental characteristics include the type of soil around the pipe, the grade of the overlying road, the distance to the nearest intersection, and the angle of intersection with the road.

5. The physical-data hybrid driven method for diagnosing the operation safety of water supply networks according to claim 1. Its features are: The extended physical feature step includes: Using a physical and mechanical model of the water supply pipeline, the structural stress of the pipeline under five operating loads was calculated. The five operating loads are: earth pressure, traffic load, internal water pressure, temperature load, and freezing load. The combined effect of the five operating loads is determined, and the biaxial stress components of the circumferential and axial stresses of the pipeline under the combined action of the five operating loads are calculated. Calculate the depth of the corrosion pit and the residual wall thickness after corrosion occurs in the pipeline, and estimate the remaining strength of the pipeline after corrosion occurs; By comparing the biaxial stress components and the residual strength, the structural safety factor of the pipeline is calculated and stored in the extended physical feature set.

6. The physical-data hybrid driven method for diagnosing the operational safety of water supply networks according to claim 1, characterized in that: In the data fusion step, the hybrid enhanced feature set includes discrete and continuous features in terms of data type. Preprocessing the hybrid enhanced features includes label encoding of the discrete features and standardization of the continuous features. The correlation analysis includes assessing the correlation between the individual hybrid enhancement features, identifying and filtering out strongly correlated hybrid enhancement features.

7. The physical-data hybrid driven method for diagnosing the operational safety of water supply networks according to claim 6, characterized in that: in, The discrete features include: pipeline material, pipeline management station, water treatment plant, surrounding soil type, and overlying road grade; The continuous characteristics include structural safety factor, pipe diameter, pipe burial depth, pipe length, pipe service life, historical pipe damage frequency, distance to adjacent intersections, and intersection angle with roads.

8. The physical-data hybrid driven method for diagnosing the operational safety of water supply networks according to claim 1, characterized in that: In the data fusion step, the process of establishing and training the hybrid-driven prediction model includes: Based on the operation and maintenance data and the extended physical feature set, the pipeline damage event is used as the basis for sample classification. The types of pipeline damage samples are divided and stored in a preset pipeline damage sample dataset. A portion of the pipeline damage sample dataset is used as the training set and the remaining portion is used as the test set. Data augmentation is performed on the training set. A category boosting algorithm is used to capture the nonlinear relationship between the hybrid enhancement features, which is then stored in the hybrid-driven prediction model. The hyperparameters in the hybrid-driven prediction model are iteratively optimized using a Bayesian optimization algorithm. The reliability and cost-effectiveness of training the hybrid-driven prediction model are evaluated on the test set. The marginal contribution of each of the hybrid enhancement features for each pipe loss sample in the test set to the prediction result is calculated. The direction of influence of each of the hybrid enhancement features on the pipe loss probability is determined. The importance of the influence of each of the hybrid enhancement features on the pipe loss probability is ranked. The effectiveness of the pipe loss prediction is judged based on the importance of the influence. The consistency of the influence direction with the objective physical laws is compared, and the physical interpretability of the hybrid-driven prediction model is evaluated.

9. A physical-data hybrid-driven water supply network operation safety diagnosis system, used for performing the physical-data hybrid-driven water supply network operation safety diagnosis method according to any one of claims 1-8, characterized in that, include: The operation and maintenance data collection module is used to determine the scope of the water supply network in the target area, obtain the operation and maintenance data of the water supply network in that area, and extract the characteristics of the operation and maintenance data. The extended physical feature module calculates the structural safety factor of the pipeline in the water supply network using a pre-established physical and mechanical model of the water supply pipeline, and stores the structural safety factor as an extended physical feature in a preset extended physical feature set. The data fusion module is used to fuse the operation and maintenance data features and extended physical features to obtain hybrid enhanced features, store them in a preset hybrid enhanced feature set, and perform data preprocessing and correlation analysis on the hybrid enhanced features to obtain analysis results and store them in a pre-established hybrid-driven prediction model. The physical state deduction module is used to deduce the dynamic evolution trend of the physical state of the water supply network based on real-time operation and maintenance data and the physical and mechanical model of the water supply pipeline. The hybrid-driven decision-making module is used to fuse the real-time operation and maintenance data and the dynamic evolution trend data, extract the fused data features, and call the hybrid-driven prediction model based on the data features to dynamically predict the pipeline loss probability in the target area, and calculate the marginal contribution of each data feature to the current prediction result in real time. The advanced early warning module is used to determine the risk level of pipeline loss based on the pipeline loss probability and trigger the corresponding early warning signal, identify key risk factors based on the marginal contribution, match the corresponding maintenance suggestions from the strategy library in the hybrid-driven prediction model, generate an early warning work order and push it.

Images