Urban underground pipeline health management method and system based on width learning

Abstract

The embodiment of the invention provides an urban underground pipeline health management method and system based on width learning, and the method comprises the steps: constructing a health evaluation model based on width learning, and carrying out the model training according to historical data, and obtaining an applicable health evaluation model; performing grid division on the urban underground pipeline; acquiring urban underground pipeline multi-source data of the target grid, and converting the urban underground pipeline multi-source data into a pipeline feature data matrix; inputting the pipeline feature data matrix into an applicable health evaluation model to obtain a health degree score; performing adaptive adjustment on the health degree score to obtain a comprehensive health degree score; and determining the health level of the urban underground pipeline according to the comprehensive health degree score. According to the technical scheme, the width learning model is adopted for health management of the urban underground pipeline, efficient, real-time and dynamic evaluation of the health state of the underground pipeline can be achieved in combination with a data integration and cooperation mechanism of urban grid management, and technical support is provided for scientific management and precise maintenance of the urban underground pipeline.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of underground pipeline management, and particularly relates to a city underground pipeline health management method and system based on width learning. BACKGROUND

[0002] City underground pipelines are called "city lifelines", and their health conditions are directly related to the safe operation of the city. With the acceleration of urbanization, the scale of underground pipelines is continuously expanding, the service life is increasing, and problems such as pipeline aging and damage are increasingly prominent, which brings great challenges to city operation and management.

[0003] The existing underground pipeline health evaluation methods mainly have the following problems: insufficient data integration and standardization: city underground pipeline data sources are diverse, formats are different, and there is a lack of unified data standards and integration mechanisms, which leads to difficulties in data sharing and one-sided evaluation results. Real-time monitoring and intelligent analysis capability is weak: the existing monitoring system mainly focuses on functional indicators (such as water pressure and flow), lacks real-time perception of pipeline safety status, and relies on manual inspection, which cannot efficiently process large-scale and high-noise data. Poor model updating and adaptability: traditional methods (such as TOPSIS and analytic hierarchy process) require manual adjustment of parameters and are difficult to cope with dynamic factors such as pipeline aging and environmental changes, and deep learning models are difficult to realize real-time incremental updating due to high computational complexity.

[0004] Therefore, how to realize a new underground pipeline health evaluation method, and further realize efficient, real-time and dynamic evaluation of the health status of underground pipelines, is a problem to be solved. SUMMARY

[0005] The embodiment of the present application provides a city underground pipeline health management method and system based on width learning, which is used to realize efficient, real-time and dynamic evaluation of the health status of underground pipelines.

[0006] To achieve the above purpose, on the one hand, the embodiment of the present application provides a city underground pipeline health management method based on width learning, which comprises the following steps: constructing a health evaluation model based on width learning, and training the health evaluation model according to historical data to obtain an applicable health evaluation model; grid division is performed on the city underground pipeline, so that the size of each grid meets the preset size; obtaining the city underground pipeline multi-source data of the target grid, and obtaining the pipeline feature data matrix according to the city underground pipeline multi-source data; inputting the pipeline feature data matrix into the applicable health evaluation model to obtain the health degree score corresponding to the pipeline feature data matrix; adaptively adjusting the health degree score to obtain a comprehensive health degree score; determining the health grade of the city underground pipeline of the target grid according to the comprehensive health degree score; and performing corresponding health management according to the health grade.

[0007] In another aspect, the embodiment of the present application also provides a city underground pipeline health management system based on width learning, comprising: a model construction module, configured to construct a health evaluation model based on width learning, and train the health evaluation model according to historical data to obtain an applicable health evaluation model; a grid division module, configured to divide the city underground pipeline into grids, so that the size of each grid meets a preset size; a data acquisition module, configured to acquire city underground pipeline multi-source data of a target grid, and obtain a pipeline feature data matrix according to the city underground pipeline multi-source data; a health degree score generation module, configured to input the pipeline feature data matrix into the applicable health evaluation model to obtain a health degree score corresponding to the pipeline feature data matrix; a health grade evaluation module, configured to adaptively adjust the health degree score to obtain a comprehensive health degree score value; determine the health grade of the city underground pipeline of the target grid according to the comprehensive health degree score value; and a health management module, configured to perform corresponding health management according to the health grade.

[0008] The above technical solution has the following beneficial effects:

[0009] The technical solution adopts a width learning model to perform health management of the city underground pipeline, can fully exert the rapid training, small sample adaptability and incremental learning capability of the width learning system, and can realize efficient, real-time and dynamic evaluation of the health state of the underground pipeline by combining the data integration and collaborative mechanism of city grid management, thereby providing technical support for scientific management and precise maintenance of the city underground pipeline. BRIEF DESCRIPTION OF DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained from these drawings without creative labor.

[0011] Figure 1 is a flow chart of a city underground pipeline health management method based on width learning according to an embodiment of the present application;

[0012] Figure 2 is a structural diagram of a city underground pipeline health management system based on width learning according to an embodiment of the present application. DETAILED DESCRIPTION

[0013] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of the present application.

[0014] As shown in Figure 1 The embodiment of the present application provides a health management method for urban underground pipelines based on width learning, comprising:

[0015] S101, a health evaluation model based on width learning is constructed, and the health evaluation model is trained according to historical data to obtain an applicable health evaluation model;

[0016] S102, the urban underground pipelines are divided into grids, so that the size of each grid meets the preset size;

[0017] S103, the multi-source data of the urban underground pipelines of the target grid are obtained, and pipeline feature data matrix is obtained according to the multi-source data of the urban underground pipelines;

[0018] S104, the pipeline feature data matrix is input into the applicable health evaluation model to obtain a health degree score corresponding to the pipeline feature data matrix;

[0019] S105, the health degree score is adaptively adjusted to obtain a comprehensive health degree score;

[0020] S106, the health grade of the urban underground pipelines of the target grid is determined according to the comprehensive health degree score;

[0021] S107, corresponding health management is performed according to the health grade.

[0022] To solve the above problems, in the technical solution, a width learning model is introduced to manage the health of urban underground pipelines, which can fully exert the rapid training, small sample adaptability and incremental learning ability of the width learning system. At the same time, the urban underground pipelines are managed in a grid manner, which can realize data integration and collaborative mechanism of grid management, so as to complete efficient, real-time and dynamic evaluation of the health state of underground pipelines, and provide technical support for scientific management and precise maintenance of urban underground pipelines. In addition, after outputting the health degree score by the applicable health evaluation model, expert weights and score adjustment rules are introduced to revise the health degree score, so as to make it more meet the actual situation of the health management of urban underground pipelines, and provide more accurate basis for subsequent health grade division.

[0023] Further, the urban underground pipeline multi-source data includes sensor data, pipeline attribute data, GIS data, and maintenance data.

[0024] In the technical solution, multiple types of sensors (pressure, temperature, vibration, etc.) and optical fiber monitoring equipment are deployed in each grid, and the operation data of the urban underground pipeline are collected and transmitted in real time through Zigbee / 5G communication technology, so the sensor data includes pressure sensor data, temperature sensor data, and vibration sensor data; meanwhile, the pipeline attribute data includes pipe age (the difference between the construction year and the current year), pipe diameter, pipe material (such as cast iron, PVC, and steel pipe), and pipe length; the GIS data includes pipeline spatial coordinates (latitude and longitude, buried depth), surrounding soil properties (corrosion grade), underground water level, and ground traffic load (such as traffic flow and road bearing capacity); and the maintenance data includes inspection records (inspection time and discovered abnormal conditions), maintenance history (maintenance frequency, maintenance position, and replaced parts), and overload conditions (such as the number of times of pipe overpressure operation).

[0025] Further, in the S103, the pipeline feature data matrix is obtained according to the urban underground pipeline multi-source data, and specifically includes:

[0026] S1031, the sensor data is subjected to abnormal data processing, and the abnormal data processing includes missing value processing and noise filtering; the missing value processing can adopt a K-nearest neighbor interpolation method, etc.; and the noise filtering can adopt a wavelet denoising method (for continuous data such as pressure and temperature) or a median filter method (for high-frequency noise data such as vibration), etc.

[0027] S1032, the urban underground pipeline multi-source data subjected to the abnormal data processing is subjected to data alignment processing; for example, time dimension alignment is achieved through GPS time stamp synchronization, and spatial dimension alignment is achieved through three-dimensional coordinate conversion;

[0028] S1033, from the urban underground pipeline multi-source data subjected to the data alignment processing, feature data for each feature item (pipe age, pipe diameter, pipe material, soil corrosion, underground water level, maintenance frequency, and overload number, etc.) is extracted according to a preset feature item;

[0029] S1034, Z-score standardization method is adopted to respectively standardize all the feature data;

[0030] S1035, the feature data subjected to the standardization is summarized, and a pipeline feature data matrix is constructed.

[0031] Further, the step S105 specifically includes:

[0032] S1051. Determine the expert weight evaluation items (such as pipe age, burial depth, and soil corrosivity), and determine the expert weight values ​​corresponding to the expert weight evaluation items (such as pipe age accounting for 30%, burial depth accounting for 20%, and soil corrosivity accounting for 25%). Each expert weight evaluation item is a type of data from the multi-source data of urban underground pipelines.

[0033] S1052. Match the expert weight evaluation items with the multi-source data of urban underground pipelines to obtain the data values ​​of the expert weight evaluation items (such as pipe age = 25 years, burial depth = 3m, soil corrosivity = medium).

[0034] S1053. Calculate the health score weight value corresponding to the health score based on the expert weight value;

[0035] S1054. The comprehensive health score is obtained by weighting the health score, the health score weight value, the expert weight evaluation item data value, and the expert weight value. (e.g., health score × health score weight value) + Σ(standardized score of expert weight evaluation item data value × expert weight value).

[0036] The calculation process is illustrated with an example:

[0037] For example, if the current output health score is 95, the tube age accounts for 30%, the actual tube age is X years, the burial depth accounts for 20%, and the actual burial depth is Y meters, then the weight of the health score is 1-30%-20%=50%, and the final comprehensive health score is 95×50%+X×30%+Y×20%. Before the calculation, the values ​​X and Y need to be standardized to convert them into values ​​between 0 and 100.

[0038] Furthermore, prior to step S106, the procedure also includes:

[0039] S1055. Determine whether the multi-source data of urban underground pipelines triggers the scoring adjustment conditions. If so, update the comprehensive health score according to the preset scoring adjustment rules.

[0040] This process primarily involves adjusting the overall health score by supplementing it with specific rules, thereby obtaining a more intuitive risk level assessment. The score adjustment conditions and rules are pre-defined. For example, if the current multi-source data on urban underground pipelines triggers the adjustment condition of "pipe age > 30 years and strong soil corrosivity," then according to the corresponding adjustment rule, the overall health score will be reduced by 20% as the new overall health score. Alternatively, if it triggers the adjustment condition of "vibration frequency > 5Hz and duration > 10 minutes," then according to the corresponding adjustment rule, the overall health score will be reduced by 15% as the new overall health score. If no adjustment condition is triggered, the overall health score will not be updated.

[0041] The adjustment rules can be made through a fuzzy reasoning system (BL-DFIS), which is a process of converting a comprehensive health score into an interpretable risk level.

[0042] Furthermore, the health level of urban underground pipelines is divided into four levels. The risk of each level is inversely proportional to the corresponding comprehensive health score. For example, it can be divided as follows:

[0043] Level I: Overall health score > 90 points, corresponding to "good condition", extremely low risk, normal maintenance (such as regular inspection) is sufficient;

[0044] Level II: Overall health score of 70-90 points, corresponding to "medium risk", which is low risk and requires enhanced monitoring (such as increasing sensor sampling frequency);

[0045] Level III: Overall health score of 50-70 points, corresponding to "relatively unsafe", with a high risk, requiring local repair (such as repairing corroded parts of the pipe wall);

[0046] Level IV: Overall health score <50 points, corresponding to "high risk", extremely high risk, requiring emergency repair or replacement (such as replacing the entire aging pipeline).

[0047] Furthermore, the health assessment model is a width learning model, whose structure includes an input layer, a feature mapping layer, an enhancement node layer, and an output layer;

[0048] Step S104 specifically includes:

[0049] S1041. Input the pipeline feature data matrix into the input layer, and then the feature mapping layer outputs the feature nodes.

[0050] S1042. Input the feature nodes into the enhancement node layer to obtain the enhancement nodes;

[0051] S1043. Concatenate the feature nodes and enhancement nodes to obtain the extended feature matrix;

[0052] S1044. Calculate and output the health score based on the output weight β and extended feature matrix of the applicable health assessment model.

[0053] Furthermore, during the training of the health assessment model based on historical data, the following formula is used to calculate the output weight β of the applicable health assessment model;

[0054] β=(A T A+CI) -1 A T Y

[0055] Where A is the extended feature matrix, Y is the label of the health score (0-100 health score based on historical data annotation), C is the preset regularization parameter, and I is the identity matrix.

[0056] The output weight β obtained through model training is the initial output weight β applicable to the health assessment model. This output weight β will be updated through the incremental learning process.

[0057] Furthermore, the urban underground pipeline health management method based on width learning also includes:

[0058] S108. After updating the multi-source data of urban underground pipelines (such as sensor data updates or pipeline maintenance data updates), calculate the incremental weights of the enhanced node layer based on the updated multi-source data of urban underground pipelines; update the output weights β using the incremental weights. In this method, only the weights of the enhanced layer are updated, without retraining the entire model.

[0059] like Figure 2 As shown, this embodiment of the invention also provides a city underground pipeline health management system based on width learning, including:

[0060] The model building module 21 is used to build a health assessment model based on width learning, and train the health assessment model according to historical data to obtain an applicable health assessment model.

[0061] The grid division module 22 is used to divide the urban underground pipelines into grids, so that the size of each grid conforms to the preset size.

[0062] The data acquisition module 23 is used to acquire multi-source data of urban underground pipelines in the target grid, and to obtain a pipeline feature data matrix based on the multi-source data of urban underground pipelines.

[0063] The health score generation module 24 is used to input the pipeline feature data matrix into the applicable health assessment model to obtain a health score corresponding to the pipeline feature data matrix.

[0064] The health level assessment module 25 is used to adaptively adjust the health score to obtain a comprehensive health score; and to determine the health level of the urban underground pipelines in the target grid based on the comprehensive health score.

[0065] The health management module 26 is used to perform corresponding health management according to the stated health level.

[0066] The above method will be described in detail below through a specific embodiment:

[0067] Step 1: Urban grid division and data collection:

[0068] Based on the distribution characteristics of urban underground pipelines, the city is divided into multiple grid units (e.g., each grid is 500m × 500m), and the grid size is dynamically adjusted according to the density of urban underground pipelines (e.g., for high-density areas, it can be reduced to 250m × 250m to improve monitoring accuracy).

[0069] Multiple types of sensors are deployed in each grid cell (e.g., 10 pressure sensors per kilometer, 15 temperature sensors per kilometer, and one vibration sensor every 200m). At the same time, fiber optic monitoring equipment is deployed at key nodes to collect real-time operational data (i.e., sensor data) of urban underground pipelines through communication technologies such as Zigbee / 5G, and then upload it to the cloud via the 5G network.

[0070] By integrating multi-source data (such as geophysical data, GIS data, operation and maintenance records, etc.) of urban underground pipelines through a three-dimensional buffer matching algorithm (plane radius ±30cm, elevation radius ±15cm), a comprehensive pipeline dataset within the grid unit is formed.

[0071] Step 2: Data Preprocessing and Feature Extraction

[0072] Missing value processing (K-nearest neighbor interpolation) and noise filtering (wavelet denoising or median filtering) are performed on the sensor data.

[0073] Align multi-source data (GPS timestamp synchronization) and standardize (normalize) it.

[0074] Extracting multi-dimensional features of pipelines within grid cells from the integrated data:

[0075]

[0076] Table 1. Feature Hints

[0077] The features are processed using the Z-score normalization method, with the following formula:

[0078]

[0079] Where μ is the characteristic mean and σ is the characteristic standard deviation.

[0080] Step 3: Construction of the width-based learning model (health assessment model based on width-based learning):

[0081] The health assessment model based on width learning includes an input layer, a feature mapping layer, an augmentation node layer, and an output layer.

[0082] Input layer: Receives standardized feature data (dimension m).

[0083] Feature mapping layer design:

[0084] Feature nodes are generated using random Fourier feature mapping to capture nonlinear relationships such as pipe aging and environmental corrosion.

[0085] The number of feature nodes is set to 200, and the activation function is ReLU.

[0086] Enhanced node layer design:

[0087] Enhanced nodes are generated through very sparse mapping, reducing redundant information.

[0088] The number of augmentation nodes was set to 400 (ratio of augmentation nodes to feature nodes 1:2), and the activation function was Sigmoid.

[0089] Output layer design:

[0090] The outputs of the feature layer and the enhancement layer are concatenated to form an extended feature matrix. A health score is then generated directly through pseudo-inverse calculation, eliminating the need for backpropagation.

[0091] Parameter optimization: The RIME (Frost Ice Optimization) algorithm is used to optimize key BLS parameters, including the number of feature nodes, the number of augmentation nodes, and the regularization coefficient C (typically set to 0.001-0.01). The optimization objective is to minimize the prediction error (e.g., MSE).

[0092] Model training: Train the BLS model using historical data and calculate the output weights β.

[0093] β=(A T A+CI) -1 A T Y

[0094] Where A is the concatenation matrix of the feature layer and the enhancement layer, Y is the label (0-100) of the health score, and C is the regularization parameter.

[0095] Step 4: Health Assessment and Classification:

[0096] Convert the health score (0-1) output by BLS to a 0-100 scale;

[0097] Then, the health score is combined with expert weights (such as tube age accounting for 30% and burial depth accounting for 20%) to generate a comprehensive health score.

[0098] According to industry standards (such as "Maintenance and Assessment of Urban Drainage Pipelines" CJJ68), the health level is divided into four levels:

[0099] Level I (>90 points): Good condition, normal maintenance required.

[0100] Level II (70-90 points): Medium risk, monitoring needs to be strengthened.

[0101] Level III (50-70 points): Less safe, local repairs should be performed.

[0102] Level IV (<50 points): High risk, requiring urgent repair or replacement.

[0103] The interpretability of the model is enhanced by transforming continuous health scores into interpretable risk levels through a fuzzy inference system (BL-DFIS). For example, fuzzy rules are designed as follows:

[0104] If the pipe age is >30 years and the soil is highly corrosive, then the health status will decrease by 20%.

[0105] If the vibration frequency is >5Hz and the duration is >10 minutes, then health will decrease by 15%.

[0106] Step 5: Dynamic Model Updates and Collaboration

[0107] Incremental learning: When new sensor data or pipeline modification records are added, only the weights of the enhancement layer are updated; the entire model does not need to be retrained. The update formula is:

[0108] β = β + Δβ;

[0109] Where Δβ is the incremental weight difference.

[0110] Federated learning: Each grid cell trains its BLS model locally, while the FedNA+EWWA-FL method is used in the cloud to aggregate the weights of the enhancement layers, achieving collaborative training of multi-department data and privacy protection. The specific process is as follows:

[0111] 1) Send global model initialization parameters to each grid cell from the cloud.

[0112] 2) Each grid cell trains a BLS model based on local data to generate enhancement layer weights.

[0113] 3) The weights of the enhancement layer are aggregated at the element level in the cloud according to the EWWA-FL method, and the global model is optimized by combining the class weight strategy of FedNA.

[0114] The cloud distributes the optimized global model parameters to each grid cell to achieve collaborative updates.

[0115] Edge computing: A lightweight BLS model (50 feature nodes, 100 enhancement nodes) is deployed on a Zigbee gateway to perform preliminary local health calculations, and the results are uploaded to the cloud. The model occupies only 500MB of memory and has a training time of approximately 2 seconds, far shorter than traditional deep learning models.

[0116] Security measures: Blockchain technology is used to record model parameter versions and update logs to ensure traceability. Federated learning only exchanges encrypted weight differences (AES-256 encryption) to prevent the leakage of raw data.

[0117] The disclosed embodiments have been described above to enable any person skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be applied to other embodiments without departing from the spirit and scope of this disclosure. Therefore, this disclosure is not limited to the embodiments given herein, but is consistent with the broadest scope of the principles and novel features disclosed in this application.

[0118] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for health management of urban underground pipelines based on width learning, characterized in that, include: A health assessment model based on width learning is constructed, and the health assessment model is trained according to historical data to obtain an applicable health assessment model. The city's underground pipelines are divided into grids, with each grid's size conforming to a preset dimension; Obtain multi-source data of urban underground pipelines in the target grid, and obtain a pipeline feature data matrix based on the multi-source data of urban underground pipelines; The pipeline feature data matrix is ​​input into the applicable health assessment model to obtain a health score corresponding to the pipeline feature data matrix; The health score is adaptively adjusted to obtain a comprehensive health score. The health level of the urban underground pipelines in the target grid is determined based on the comprehensive health score. Health management should be carried out according to the stated health level.

2. The urban underground pipeline health management method based on width learning as described in claim 1, characterized in that, The multi-source data of the urban underground pipelines includes: sensor data, pipeline attribute data, geospatial information system (GIS) data, and maintenance data.

3. The urban underground pipeline health management method based on width learning as described in claim 2, characterized in that, The step of obtaining the pipeline feature data matrix based on the multi-source data of the urban underground pipelines specifically includes: The sensor data is subjected to anomaly processing, which includes missing value handling and noise filtering. Data alignment processing is performed on multi-source data of urban underground pipelines that has undergone anomaly data processing; From the multi-source data of urban underground pipelines that has undergone data alignment processing, feature data for each of the preset feature items is extracted; Standardize all feature data separately; The pipeline feature data matrix is ​​constructed by summarizing the standardized feature data.

4. The urban underground pipeline health management method based on width learning as described in claim 3, characterized in that, The adaptive adjustment of the health score to obtain a comprehensive health score specifically includes: The expert weight evaluation items are determined, and the expert weight values ​​corresponding to the expert weight evaluation items are determined. Each of the expert weight evaluation items is a type of data in the multi-source data of urban underground pipelines. The expert weight evaluation items are matched with the multi-source data of urban underground pipelines to obtain the data values ​​of the expert weight evaluation items. Calculate the health score weight value corresponding to the health score based on the expert weight value; A comprehensive health score is obtained by weighting the health score, the health score weight value, the expert weight evaluation item data value, and the expert weight value.

5. The urban underground pipeline health management method based on width learning as described in claim 4, characterized in that, Before determining the health level of the urban underground pipelines based on the comprehensive health score, the method further includes: Determine whether the multi-source data of the city's underground pipelines triggers the scoring adjustment conditions. If so, update the comprehensive health score according to the preset scoring adjustment rules.

6. The urban underground pipeline health management method based on width learning as described in claim 4, characterized in that, The health level of the city's underground pipelines is divided into four health levels, and the risk of each health level is inversely proportional to the corresponding comprehensive health score.

7. The urban underground pipeline health management method based on width learning as described in claim 1, characterized in that, The health assessment model includes an input layer, a feature mapping layer, an enhancement node layer, and an output layer; The step of inputting the pipeline feature data matrix into the applicable health assessment model to obtain a health score corresponding to the pipeline feature data matrix specifically includes: The pipeline feature data matrix is ​​input into the input layer, and then the feature mapping layer outputs feature nodes. The feature nodes are input into the enhancement node layer to obtain the enhancement nodes; The feature nodes and the enhancement nodes are concatenated to obtain the extended feature matrix; The health score is calculated and output based on the output weight β of the applicable health assessment model and the extended feature matrix.

8. The urban underground pipeline health management method based on width learning as described in claim 7, characterized in that, During the training of the health assessment model based on historical data, the output weight β of the applicable health assessment model is calculated using the following formula; β=(A T A+CI) -1 AND T Y Where A is the extended feature matrix, Y is the label of the health score, C is the preset regularization parameter, and I is the identity matrix.

9. The urban underground pipeline health management method based on width learning as described in claim 8, characterized in that, Also includes: After the urban underground pipeline multi-source data is updated, the incremental weight of the enhanced node layer is calculated based on the updated urban underground pipeline multi-source data. The output weight β is updated using the incremental weight.

10. A health management system for urban underground pipelines based on width learning, characterized in that, include: The model building module is used to build a health assessment model based on width learning and train the health assessment model according to historical data to obtain an applicable health assessment model. The grid division module is used to divide urban underground pipelines into grids, ensuring that the size of each grid conforms to a preset size. The data acquisition module is used to acquire multi-source data of urban underground pipelines in the target grid, and to obtain a pipeline feature data matrix based on the multi-source data of urban underground pipelines. A health score generation module is used to input the pipeline feature data matrix into the applicable health assessment model to obtain a health score corresponding to the pipeline feature data matrix. The health level assessment module is used to adaptively adjust the health score to obtain a comprehensive health score. The health level of the urban underground pipelines in the target grid is determined based on the comprehensive health score. The health management module is used to perform corresponding health management according to the stated health level.

Images