Method for rapid screening of structural defects of sewer network and decision support for repair

Abstract

This invention provides a rapid screening and repair decision support method for structural defects in drainage pipe networks. The method includes: constructing the pipe network topology and deploying multi-sensor IoT terminals to collect liquid level and flow velocity data; cleaning the data based on a microservice architecture and calculating groundwater infiltration; constructing a multi-factor risk model combining environmental and attribute parameters to calculate a comprehensive risk index; matching repair processes using a decision tree engine and generating a bill of quantities and investment estimate through an automatic cost estimation module; and finally performing a post-repair effect evaluation based on statistical significance testing. This invention, through a closed-loop system of "monitoring-modeling-decision-evaluation," solves the problems of high cost, inability to quantify infiltration, and strong subjectivity in traditional detection methods, achieving precise and economical management of drainage pipe networks.

Description

Technical Field

[0001] This invention relates to the field of drainage pipe network screening and repair, and in particular to a decision support method for rapid screening and repair of structural defects in drainage pipe networks. Background Technology

[0002] In areas with high groundwater levels, the groundwater level is consistently higher than the drainage network elevation. Structural defects such as loose pipe joints and pipe wall cracks allow groundwater to continuously infiltrate into the sewage network. This not only reduces the network's transport capacity but also severely dilutes the concentration of incoming wastewater (e.g., COD concentration), affecting the wastewater treatment plant's operational efficiency and compliance with discharge standards. Currently, screening for network defects mainly relies on traditional closed-circuit television (CCTV) or quadrant vision (QV) inspections. However, these physical inspection methods are costly, inefficient, and require pre-treatment (cleaning, dewatering) of the pipes, making it difficult to conduct comprehensive surveys of large-scale networks. More importantly, CCTV can only detect visible physical defects and cannot quantitatively assess the actual amount of groundwater infiltration, making it difficult for management departments to scientifically determine remediation priorities based on the severity of infiltration.

[0003] In existing technologies, such as CN105698013B, a method for determining the location and infiltration volume of river water and groundwater into sewage pipes is disclosed. This type of technology attempts to determine the location and infiltration volume of groundwater by deploying flow meters and level gauges, and uses monitoring data to infer the condition of the pipe network. However, the above-mentioned existing technologies still have many limitations in practical applications: First, at the data processing level, existing technologies mostly use simple threshold judgment or direct calculation, lacking refined data cleaning algorithms (such as the interquartile range method for noise reduction) for complex pipe network flow patterns, and are easily affected by sensor drift or instantaneous water flow fluctuations, leading to distorted calculation results; Second, at the system architecture level, existing monitoring systems are mostly monolithic architectures, lacking containerization technology (such as Docker) The microservice deployment capabilities of Swarm and Overlay virtual networks are insufficient to handle the concurrent processing and storage needs of massive amounts of high-frequency time-series data. Furthermore, at the decision support level, existing technologies typically stop at "problem discovery" or simply "infiltration calculation," failing to further integrate multi-source data such as soil permeability coefficients and pipe aging years to construct comprehensive risk assessment models. They also lack automated remediation process matching mechanisms based on decision trees, and cannot automatically generate engineering quantity lists and investment estimates, resulting in a gap between investigation results and engineering remediation implementation. Finally, in terms of remediation effectiveness evaluation, existing technologies largely rely on human experience and lack quantitative evaluation standards based on statistical significance tests (such as paired-samples t-tests), making it difficult to objectively verify the actual effectiveness of remediation projects. Therefore, there is an urgent need for a decision support method for rapid screening and remediation of structural defects in drainage pipe networks that integrates "precise monitoring, intelligent modeling, automated decision-making, and scientific evaluation." Summary of the Invention

[0004] The main objective of this invention is to provide a rapid screening and repair decision support method for structural defects in drainage pipe networks, which solves the technical problems in the investigation of structural defects in drainage pipe networks in areas with high groundwater levels, such as over-reliance on high-cost manual inspection (e.g., CCTV), inability to accurately quantify groundwater infiltration, and lack of automated closed-loop decision support from monitoring data to project cost budget.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a rapid screening and repair decision support method for structural defects in drainage pipe networks, the method comprising: S1. Construct a topology diagram of the drainage pipe network, identify key nodes based on the topology diagram as monitoring points for suspected defective pipe sections, configure multi-sensor fusion IoT monitoring terminals at the monitoring points, and collect pipe section liquid level sequence data and flow velocity sequence data. S2. The data processing center, deployed based on a microservice architecture, receives data uploaded by IoT monitoring terminals, calls anomaly detection algorithms to clean the data, and calculates the groundwater infiltration rate of the pipe section. S3. By combining the geographical environment parameters and physical attribute parameters of the pipeline segment, a multi-factor coupled risk assessment model is established, and the comprehensive risk index of the pipeline segment is calculated using a linear weighted algorithm. S4. Input the comprehensive risk index into the decision tree-based repair process matching engine to generate targeted repair technology solutions, and use the automatic cost calculation module to generate a bill of quantities and investment estimate. S5. After the repair project is completed, a post-repair effect evaluation process based on statistical significance testing is implemented to verify the degree of restoration of the pipeline network's operational status.

[0006] In the preferred embodiment, step S1, which involves configuring an IoT monitoring terminal at the node of a suspected defective pipe section, includes: Inspection wells at the junctions of main pipes with a catchment area greater than 0.5 square kilometers in the pipeline network topology were selected as primary monitoring nodes; The terminal branch manholes that have historically reported overflow events were selected as secondary monitoring nodes; A Doppler ultrasonic velocimeter is installed at the bottom elevation of the pipe inside the selected inspection well. The probe of the Doppler ultrasonic velocimeter faces the direction of water flow, and the center line of the probe is parallel to the pipe axis. It is used to measure the average flow velocity of the cross section. Install a submersible hydrostatic level gauge 8-12 cm above the bottom of the selected inspection well. The cable of the submersible hydrostatic level gauge is fixed along the well wall and extends to the data transmission unit at the wellhead. The sampling frequency is set to once every 5-8 minutes to simultaneously collect flow rate and liquid level data, which are then transmitted back via 4G / 5G network.

[0007] In the preferred embodiment, the data processing center in step S2 is deployed using containerization technology, and the steps include: Build a Docker Swarm container orchestration environment in the server cluster and create a virtual internal network based on the overlay driver to isolate communication traffic between different microservices; Deploy the InfluxDB time-series database service container, mount the / var / lib / influxdb directory on the host machine to the data storage volume inside the container to achieve data persistence, and map the container's port 8086 to the host machine's port 8086; Deploy a PostgreSQL relational database service container, mount the / var / lib / postgresql / data directory on the host machine into the container, map port 5432, and use it to store pipeline topology data; Deploy a compute service container based on the Python Flask framework, configure the compute service container to automatically discover the IP address of the database container through the internal DNS service of DockerSwarm, and set the number of replicas of the compute service container to at least 3 to achieve load balancing and high availability; Configure the Nginx reverse proxy container as an API gateway to intercept external requests and distribute them to the virtual IP address of the compute service container.

[0008] In the preferred embodiment, step S2, which involves cleaning the data and calculating the groundwater infiltration rate, includes: Read the raw traffic sequence from the InfluxDB database container; Outliers in the original flow sequence are detected by the interquartile range method. The upper quartile is defined as Q3 and the lower quartile as Q1. The interquartile range (IQR) is calculated to be equal to Q3 minus Q1. Data points in the original flow sequence that are greater than Q3 plus 1.5 times IQR or less than Q1 minus 1.5 times IQR are identified as anomalous noise. These anomalous noise points are then replaced with the arithmetic mean of the five consecutive normal time points before and after them, resulting in a cleaned flow sequence. The formula for identifying outlier intervals is as follows: ; In the formula: This indicates the range of valid data values. This represents the value at the 25th percentile after the flow sequence values ​​are arranged in ascending order. This represents the value at the 75th position of the flow sequence; The flow rate is calculated using the flow velocity and liquid level after cleaning, and the average flow rate from 02:00 to 04:00 is extracted as the minimum flow rate for the night. And combined with the dry season flow correction coefficient and upstream transfer of basic sewage volume Calculate groundwater infiltration The calculation formula is: ; In the formula: For the calculated groundwater infiltration rate, The extracted minimum flow characteristic value at night, This is a dry season flow correction factor, with a value ranging from 0.9 to 0.95. The theoretical basis for the transfer of domestic sewage upstream of this pipeline section is the amount of sewage.

[0009] In the preferred embodiment, step S3, which involves establishing a multi-factor coupled risk assessment model and calculating the comprehensive risk index, includes: Selecting groundwater infiltration rate Soil permeability coefficient Pipeline service life and pipeline burial depth These are the four core risk assessment factors; Introduce the design full pipe flow rate respectively Maximum soil permeability coefficient in the region Pipeline design service life and the average depth of groundwater in the region As a normalization baseline parameter; The weight coefficients of the four evaluation factors were determined using the analytic hierarchy process (AHP). And the sum of the weights is 1; The comprehensive risk index of the pipeline section is obtained by linearly weighting and summing the ratios of each factor value to the benchmark parameter. The calculation formula is: ; In the formula: The comprehensive risk index for the pipeline segment is a dimensionless value between 0 and 1. Weights for infiltration factors; For environmental factor weights; Weights for aging factors; The weights are for the burial depth factor.

[0010] In the preferred embodiment, the deployment steps of the computer software and computer modules for the decision tree-based repair process matching engine in step S4 include: In the offline training environment, a CART classification and regression tree model is built based on historical pipeline defect data and expert repair case library, and the model structure is optimized by pruning algorithm to prevent overfitting. The trained decision tree model is serialized into a binary model file using Python's Joblib library; The Scikit-learn machine learning library is integrated into the compute service container, and binary model files are loaded into memory when the container starts to build an inference engine object; By exposing a RESTful API interface through the Flask framework, the system receives the comprehensive risk index and pipe segment attribute JSON data calculated and output in step S3, calls the inference engine object for real-time prediction, and outputs the repair process category code.

[0011] In the preferred embodiment, the logical judgment steps executed by the repair process matching engine in step S4 include: Decision tree root node determines comprehensive risk index ,like Entering a high-risk (Category A) branch; if Enter the risk branch of category B; if Enter the low-risk C-category branch; Under the Class A high-risk branch, read the pipe segment deformation parameters. If the deformation is less than 10%, output the UV-cured CIPP overall repair plan; if the deformation is greater than or equal to 10% and less than 30%, output the mechanical spiral winding repair plan; if the deformation is greater than or equal to 30%, output the excavation and replacement plan. Under the risk branch in category B, read the number of defect points per 100 meters. If there are more than 5, switch to the category A branch for processing; if there are 5 or less, output a stainless steel double expansion ring point repair scheme.

[0012] In the preferred embodiment, the cost aggregation step performed by the automatic cost calculation module based on the depth-first search (DFS) algorithm in step S4 includes: Initialize an empty stack Stack and a collection Visited for recording visited nodes, and push the starting node of the network topology onto Stack. While the Stack is not empty, execute the loop: pop the top element from the stack as the current processing segment. ; like If it is not marked in Visited, then read the length of the pipe segment. burial depth and repair process unit price Introducing the concept of burial depth The difficulty of construction is determined by road traffic conditions. ; Calculate the direct repair cost of the current pipe segment. And add it to the total cost variable TotalCost; Will Mark as Visited and find all related to Push directly connected but unaccessed downstream pipe segments into the Stack; Repeat the above loop until Stack is empty, output the final accumulated TotalCost as the direct cost of the project, and calculate the measures cost and regulatory fees accordingly.

[0013] In the preferred embodiment, step S5, which involves performing a post-treatment effect evaluation process based on statistical significance testing, includes: The sample set consists of minimum nighttime flow data for seven consecutive days following the completion of the restoration project. A sample set was constructed by retrieving the minimum nighttime flow data for seven consecutive days under the same weather conditions before the repair. ; Constructing the null hypothesis Alternative hypothesis ; Calculate the average of the differences between the two groups of samples. and the standard deviation of the difference ; Calculate the statistic using the paired samples t-test formula. The value, the formula is: ; In the formula: To test the statistic, It is the arithmetic mean of the differences between the two sets of samples. The standard deviation of the sample difference. The number of paired samples, This represents the overall mean.

[0014] In the preferred scheme, step S5 is based on the statistical measure. The steps for determining the repair effect include: According to degrees of freedom Consult the standard t-distribution critical value table to obtain the one-tailed critical value at a significance level of 0.05. ; Comparison of calculations Value and ; like Furthermore, the corresponding p-value is less than 0.05, therefore the null hypothesis is rejected. If the repair work is deemed effective, the system automatically updates the pipe section status to healthy and archives it. like Unable to reject the null hypothesis If the repair is deemed ineffective, the system automatically generates a secondary inspection work order, indicating that there may be undetected branch pipe misconnections or structural leaks.

[0015] This invention provides a rapid screening and repair decision support method for structural defects in drainage pipe networks, which has significant technical advantages compared to existing technologies. First, this invention adopts a containerized microservice architecture based on Docker Swarm, isolating InfluxDB, PostgreSQL, and computing services through an overlay virtual network, achieving high availability and elastic scaling of the system, and stably supporting high-frequency monitoring data streams from large-scale pipe networks. Second, this invention uses the interquartile range (IQR) method to deeply clean the original flow velocity sequence, effectively eliminating abnormal noise, and combines it with the minimum nighttime flow method to accurately calculate groundwater infiltration, overcoming the problem of poor data anti-interference capability of traditional methods, and achieving low-cost quantitative assessment of pipe network health status without excavation.

[0016] Furthermore, this invention constructs a multi-factor coupled risk assessment model incorporating infiltration rate, soil permeability coefficient, pipe age, and burial depth. It also introduces a remediation process matching engine based on CART decision trees, which can automatically recommend targeted remediation solutions such as UV curing, spiral winding, or excavation and replacement based on the risk level and specific defect parameters (e.g., deformation degree) of the pipe section. This avoids "blind overhauls" or "missed repairs" in traditional remediation methods, significantly improving the efficiency of fiscal fund utilization. In addition, the system incorporates an automatic cost calculation module based on the depth-first search (DFS) algorithm, capable of traversing the topology path and aggregating costs in real time, directly outputting a bill of quantities and investment estimates, greatly shortening the pre-project decision-making cycle. Finally, this invention introduces a post-remediation effect evaluation process based on paired-samples t-tests, objectively verifying the significance of flow changes before and after remediation through statistical methods. This provides scientific and rigorous data support for project acceptance and achieves closed-loop management of the entire pipeline network lifecycle. Attached Figure Description

[0017] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is a flowchart of the decision support method for rapid screening and repair of structural defects in drainage pipe networks according to the present invention; Figure 2 This is a structural diagram of the IoT monitoring terminal installation of the present invention.

[0018] In the attached diagram: 1. Data transmission unit; 2. Fixing buckle; 3. Submersible hydrostatic level gauge; 4. Doppler ultrasonic flow meter. Detailed Implementation

[0019] Example 1 like Figure 1-2 As shown, a rapid screening and repair decision support method for structural defects in drainage pipe networks is proposed. The method includes: S1. Construct a topology diagram of the drainage pipe network, identify key nodes based on the topology diagram as monitoring points for suspected defective pipe sections, configure multi-sensor fusion IoT monitoring terminals at the monitoring points, and collect pipe section liquid level sequence data and flow velocity sequence data. S2. The data processing center, deployed based on a microservice architecture, receives data uploaded by IoT monitoring terminals, calls anomaly detection algorithms to clean the data, and calculates the groundwater infiltration rate of the pipe section. S3. By combining the geographical environment parameters and physical attribute parameters of the pipeline segment, a multi-factor coupled risk assessment model is established, and the comprehensive risk index of the pipeline segment is calculated using a linear weighted algorithm. S4. Input the comprehensive risk index into the decision tree-based repair process matching engine to generate targeted repair technology solutions, and use the automatic cost calculation module to generate a bill of quantities and investment estimate. S5. After the repair project is completed, a post-repair effect evaluation process based on statistical significance testing is implemented to verify the degree of restoration of the pipeline network's operational status.

[0020] In the preferred embodiment, step S1, which involves configuring an IoT monitoring terminal at the node of a suspected defective pipe section, includes: Inspection wells at the junctions of main pipes with a catchment area greater than 0.5 square kilometers in the pipeline network topology were selected as primary monitoring nodes; The terminal branch manholes that have historically reported overflow events were selected as secondary monitoring nodes; A Doppler ultrasonic velocimeter 4 is installed at the bottom elevation of the pipe inside the selected inspection well. The probe of the Doppler ultrasonic velocimeter 4 faces the direction of water flow, and the center line of the probe is parallel to the pipe axis. It is used to measure the average flow velocity of the cross section. Install a submersible hydrostatic level gauge 3 8-12 cm above the bottom of the selected inspection well. The cable of the submersible hydrostatic level gauge 3 is fixed along the well wall and extends to the data transmission unit 1 at the wellhead. The sampling frequency is set to once every 5-8 minutes to simultaneously collect flow rate and liquid level data, which are then transmitted back via 4G / 5G network.

[0021] In this embodiment, the specific implementation process of step S1 involves the scientific selection of monitoring nodes and the standardized installation of IoT monitoring terminals. During the node selection phase, the system first traverses the topology data of the drainage network and calculates the catchment area of ​​each manhole node. The system selects manholes at the junctions of main pipes with a catchment area exceeding 0.5 square kilometers, defining them as primary monitoring nodes to monitor the operational status of the main stream. Simultaneously, the system retrieves historical maintenance data to identify end-branch manholes with past records of overflow alarms or complaints, defining them as secondary monitoring nodes to cover high-risk areas.

[0022] After selecting the nodes, engineers deployed sensors within the chosen inspection wells. For flow velocity measurement, a Doppler ultrasonic velocimeter was chosen as the core sensor. During installation, the velocimeter probe was fixed at the bottom elevation of the inner wall of the pipe, ensuring the sensing surface of the probe faced the direction of the incoming water flow. Simultaneously, a leveling instrument was used to adjust the probe position, ensuring the probe's geometric centerline was strictly parallel to the pipe's axial centerline. This ensured accurate capture of the velocity of suspended particles in the water flow, allowing for the calculation of the average flow velocity across the cross-section.

[0023] For liquid level monitoring, a submersible hydrostatic level gauge is selected. During installation, the level gauge probe is fixed at a vertical height of 8 to 12 centimeters from the bottom of the pipe. This installation height is designed to prevent the probe from being buried by silt at the bottom, while ensuring that the probe remains submerged in water during low water level operation. The signal transmission cable of the level gauge is laid upward along the inner wall of the inspection well using stainless steel clips and connected to an explosion-proof data transmission unit located below the wellhead.

[0024] The operating parameters of the IoT monitoring terminal are configured via remote commands. The system sets the data sampling frequency to trigger once every 5 to 8 minutes. At each sampling moment, the terminal simultaneously activates the Doppler ultrasonic flow meter and the submersible hydrostatic level gauge, acquiring the current flow velocity and level values. The collected raw data is packaged and sent to the cloud server via the built-in 4G or 5G communication module. Based on the collected flow velocity and level data, the system calculates the instantaneous flow rate in the pipeline using fluid dynamics formulas, as follows: ; In the formula, Represents the instantaneous flow rate within the pipeline, measured in cubic meters per second; This represents the cross-sectional average velocity measured by a Doppler ultrasonic velocimeter, in meters per second. The cross-sectional area, expressed in square meters, is calculated based on the liquid level measured by an immersion hydrostatic level gauge and the pipe's geometric dimensions. This calculation process ensures a high degree of accuracy and representativeness in the data foundation for subsequent infiltration analysis.

[0025] In the preferred embodiment, the data processing center in step S2 is deployed using containerization technology, and the steps include: Build a Docker Swarm container orchestration environment in the server cluster and create a virtual internal network based on the overlay driver to isolate communication traffic between different microservices; Deploy the InfluxDB time-series database service container, mount the / var / lib / influxdb directory on the host machine to the data storage volume inside the container to achieve data persistence, and map the container's port 8086 to the host machine's port 8086; Deploy a PostgreSQL relational database service container, mount the / var / lib / postgresql / data directory on the host machine into the container, map port 5432, and use it to store pipeline topology data; Deploy a compute service container based on the Python Flask framework, configure the compute service container to automatically discover the IP address of the database container through the internal DNS service of DockerSwarm, and set the number of replicas of the compute service container to at least 3 to achieve load balancing and high availability; Configure the Nginx reverse proxy container as an API gateway to intercept external requests and distribute them to the virtual IP address of the compute service container.

[0026] In this embodiment, the data processing center deployment process described in step S2 employs advanced containerization technology and a microservice architecture to ensure high system availability and data security. First, a Docker Swarm container orchestration environment is built at the physical server cluster level. This environment abstracts multiple physical servers into a unified virtual resource pool and enables cross-host container communication by creating a virtual internal network based on an overlay driver. This virtual internal network is logically independent of the external physical network, effectively isolating communication traffic between different microservice components and preventing sensitive pipeline monitoring data from being illegally stolen or interfered with by the external network during transmission.

[0027] On this infrastructure, the system deploys time-series database and relational database service containers respectively. For massive, high-frequency liquid level and flow rate monitoring data, the system deploys an InfluxDB time-series database service container. To prevent data loss due to container restarts or migrations, the system mounts the host machine's file path ` / var / lib / influxdb` to the container's internal data storage volume, achieving persistent data storage. Simultaneously, the system maps the container's port 8086 to the host machine's port 8086, using it as a dedicated channel for data writing and querying. For static relational data such as pipeline network topology and node attributes, the system deploys a PostgreSQL relational database service container. Similarly, the host machine's ` / var / lib / postgresql / data` directory is mounted into the container and mapped to port 5432, ensuring secure storage and rapid indexing of basic pipeline network geographic information data.

[0028] The core computing logic is handled by a compute service container developed based on the Python Flask framework. This container is responsible for performing data cleaning, infiltration calculation, and risk assessment model computation tasks. During deployment, the compute service container is configured to utilize Docker Swarm's built-in internal DNS service. This feature allows the compute service to automatically resolve its corresponding dynamic IP address from the database service's service name, eliminating the need to hardcode the database address in the code and thus improving system flexibility. To achieve load balancing and eliminate single points of failure, the system sets the number of compute service container replicas to at least three. This means that at least three compute service instances run simultaneously in the cluster, and if one instance fails, the others can immediately take over the tasks. Docker Swarm automatically distributes external request traffic evenly across these three instances. The load calculation formula for a single instance is as follows: ; In the formula, This represents the number of requests a single compute service container instance needs to process per second. This represents the total request rate received by the system. This represents the number of replicas configured for the computing service container; in this embodiment, this value is greater than or equal to 3. This mechanism ensures that the CPU and memory resources of the computing center can be utilized efficiently in high-concurrency data upload scenarios.

[0029] The system's entry point is configured with an Nginx reverse proxy container as the API gateway. This container listens for external HTTP or HTTPS requests and acts as a traffic scheduler. It intercepts all access requests from IoT monitoring terminals or user clients and forwards these requests to the virtual IP addresses of the computing service containers in the Overlay network according to load balancing policies. This reverse proxy layer not only hides the real network topology of the backend microservices but also provides unified interface authentication and traffic control functions, further enhancing the robustness of the entire decision support system.

[0030] In the preferred embodiment, step S2, which involves cleaning the data and calculating the groundwater infiltration rate, includes: Read the raw traffic sequence from the InfluxDB database container; Outliers in the original flow sequence are detected by the interquartile range method. The upper quartile is defined as Q3 and the lower quartile as Q1. The interquartile range (IQR) is calculated to be equal to Q3 minus Q1. Data points in the original flow sequence that are greater than Q3 plus 1.5 times IQR or less than Q1 minus 1.5 times IQR are identified as anomalous noise. These anomalous noise points are then replaced with the arithmetic mean of the five consecutive normal time points before and after them, resulting in a cleaned flow sequence. The formula for identifying outlier intervals is as follows: ; In the formula: This indicates the range of valid data values. This represents the value at the 25th percentile after the flow sequence values ​​are arranged in ascending order. This represents the value at the 75th position of the flow sequence; The flow rate is calculated using the flow velocity and liquid level after cleaning, and the average flow rate from 02:00 to 04:00 is extracted as the minimum flow rate for the night. And combined with the dry season flow correction coefficient and upstream transfer of basic sewage volume Calculate groundwater infiltration The calculation formula is: ; In the formula: For the calculated groundwater infiltration rate, The extracted minimum flow characteristic value at night, This is a dry season flow correction factor, with a value ranging from 0.9 to 0.95. The theoretical basis for the transfer of domestic sewage upstream of this pipeline section is the amount of sewage.

[0031] In this embodiment, step S2 performs a rigorous statistical processing procedure during the data cleaning and infiltration calculation stage to eliminate the interference of environmental noise on the monitoring data. The computing service container first retrieves the raw flow sequence data for a specified time period from the InfluxDB database container via the internal network. The system uses the interquartile range (IQR) method to detect outliers in the sequence. This method does not make specific assumptions about the data distribution pattern and therefore has strong robustness. The system first arranges the values ​​in the raw flow sequence in ascending order, determining the value at the 25th percentile as the lower quartile and the value at the 75th percentile as the upper quartile. Subsequently, the system calculates the IQR, which is equal to the difference between the upper quartile and the lower quartile, reflecting the dispersion of the middle 50% of the data.

[0032] Based on the calculated interquartile range, the system constructs a numerical value interval for valid data. Data points falling outside this interval are defined as abnormal noise. For each identified abnormal data point, the system does not directly delete it. Instead, it employs a smooth replacement strategy: it reads the five data points immediately before and after the timestamp of the abnormal point that are considered normal, calculates the arithmetic mean of these ten normal data points, and replaces the original abnormal value with this average, thus obtaining a smooth and continuous cleaned flow sequence. The formula for calculating the effective data numerical value interval is as follows: ; In the formula, This represents the range of valid data values ​​identified by the cleaning algorithm. Represents the lower quartile of the flow series; This represents the upper quartile of the flow series. The coefficient 1.5 in the formula is a commonly used standard multiple for identifying mild outliers in statistics, which can effectively cover approximately 99.3% of the data range in a normal distribution.

[0033] After data cleaning, the system synthesizes flow rate data using the processed velocity and level sequences and performs quantitative calculations of groundwater infiltration. The system extracts flow rate data from a specific time period between 02:00 and 04:00 daily, calculates the arithmetic mean of all flow rate data within this period, and defines it as the minimum nighttime flow characteristic value. This time period is chosen because both urban residential and commercial water consumption are at their lowest levels during this period, and the flow rate within the pipe network mainly consists of groundwater infiltration and a small amount of basic sewage flow. The system combines a dry season flow correction coefficient with the upstream transferred basic sewage volume and calculates the final groundwater infiltration rate through linear subtraction. The calculation formula is as follows: ; In the formula, This represents the calculated groundwater infiltration rate for the pipe section, expressed in cubic meters per day. This represents the extracted minimum nighttime flow characteristic value, in cubic meters per day; This represents the dry season flow correction factor, which is used to correct for flow deviations caused by very small amounts of water used by users at night. The value range is set from 0.9 to 0.95. This represents the theoretical volume of domestic sewage transferred from upstream of this pipeline section, a value determined by the operating records of the upstream pumping station or a population equivalent model. Through this calculation process, the system can accurately separate the groundwater infiltration component from the total flow.

[0034] In the preferred embodiment, step S3, which involves establishing a multi-factor coupled risk assessment model and calculating the comprehensive risk index, includes: Selecting groundwater infiltration rate Soil permeability coefficient Pipeline service life and pipeline burial depth These are the four core risk assessment factors; Introduce the design full pipe flow rate respectively Maximum soil permeability coefficient in the region Pipeline design service life and the average depth of groundwater in the region As a normalization baseline parameter; The weight coefficients of the four evaluation factors were determined using the analytic hierarchy process (AHP). And the sum of the weights is 1; The comprehensive risk index of the pipeline section is obtained by linearly weighting and summing the ratios of each factor value to the benchmark parameter. The calculation formula is: ; In the formula: The comprehensive risk index for the pipeline segment is a dimensionless value between 0 and 1. Weights for infiltration factors; For environmental factor weights; Weights for aging factors; The weights are for the burial depth factor.

[0035] In this embodiment, step S3 establishes a multi-factor coupled risk assessment model. This model aims to transform single-dimensional infiltration data into a comprehensive indicator that fully reflects the health status of the pipe section. The system selects four representative core risk assessment factors. The first factor is the groundwater infiltration volume calculated in step S2, which directly reflects the degree of structural leakage in the current pipe section. The second factor is the soil permeability coefficient, which reflects the ability of the soil medium surrounding the pipe section to transport groundwater; the higher the permeability coefficient, the greater the risk of infiltration. The third factor is the service life of the pipeline, which reflects the degree of physical aging and corrosion of the pipe material over time. The fourth factor is the pipeline burial depth, which determines the magnitude of the hydrostatic pressure of groundwater on the outside of the pipeline; the deeper the burial, the greater the external water pressure and the stronger the infiltration driving force.

[0036] To eliminate differences in dimensions and orders of magnitude among different evaluation factors, the system introduces normalized benchmark parameters to perform dimensionless processing on each factor. The benchmarks are as follows: the design full-pipe flow rate of the pipe section as the infiltration rate; the maximum soil permeability coefficient from the historical geological survey of the region as the soil permeability coefficient; the theoretical design service life corresponding to the pipe material as the service life; and the average burial depth of the groundwater level in the region as the pipe installation depth. When determining the relative importance of each factor, the system uses the analytic hierarchy process (AHP) to construct a judgment matrix. After consistency testing, the weight coefficients corresponding to the four evaluation factors are calculated, ensuring that the algebraic sum of these four weight coefficients is strictly equal to 1.

[0037] Based on the above parameters, the system uses a linear weighted summation algorithm to calculate the comprehensive risk index of the pipeline segment. The calculation formula is as follows: ; In the formula, The comprehensive risk index of the pipe section is calculated. This index is a dimensionless value between 0 and 1. The closer the value is to 1, the higher the risk of structural defects in the pipe section. This represents the groundwater infiltration rate calculated in step S2; This represents the full-pipe flow rate determined during the design phase of the pipe section; The measured soil permeability coefficient represents the location of the pipe section; This represents the historical maximum statistical value of the soil permeability coefficient in this region; This represents the length of time the pipe section has been in service since it was laid and passed acceptance, up to the present moment. This represents the design service life of the pipe section's material as specified in the national standard; Represents the vertical distance from the centerline of the pipeline to the ground; This represents the average depth of groundwater recorded by monitoring wells in the area; The weighting coefficients representing the infiltration factor are used to measure the contribution of infiltration volume to the total risk. The weighting coefficients representing environmental factors are used to measure the contribution of soil geological conditions to the total risk. The weighting coefficients representing aging factors are used to measure the contribution of management age to total risk. The weighting coefficient representing the burial depth factor is used to measure the contribution of external water pressure to the total risk.

[0038] In the preferred embodiment, the deployment steps of the computer software and computer modules for the decision tree-based repair process matching engine in step S4 include: In the offline training environment, a CART classification and regression tree model is built based on historical pipeline defect data and expert repair case library, and the model structure is optimized by pruning algorithm to prevent overfitting. The trained decision tree model is serialized into a binary model file using Python's Joblib library; The Scikit-learn machine learning library is integrated into the compute service container, and binary model files are loaded into memory when the container starts to build an inference engine object; By exposing a RESTful API interface through the Flask framework, the system receives the comprehensive risk index and pipe segment attribute JSON data calculated and output in step S3, calls the inference engine object for real-time prediction, and outputs the repair process category code.

[0039] In this embodiment, the software deployment process of the decision tree-based repair process matching engine described in step S4 is divided into an offline model building stage and an online inference service deployment stage. In the offline training environment, system engineers first collect historical defect records of the drainage pipe network and a repair case library reviewed and approved by senior maintenance experts to build a high-dimensional training dataset. The system selects the CART classification and regression tree algorithm as the core logic and constructs a decision tree model through recursive segmentation. In order to improve the generalization ability of the model and avoid overfitting to the training data, the system uses a cost complexity pruning algorithm to optimize the generated initial tree, removing the terminal branches that contribute less to the prediction accuracy, thereby obtaining a final model with a concise structure and strong robustness.

[0040] After model training is complete, the system uses the Joblib utility library in the Python programming language to perform model serialization. This operation converts the trained decision tree object, along with its internal node splitting rules, threshold parameters, and leaf node weights, into a complete binary model file. This binary model file is small in size and loads quickly, and can be transmitted and stored independently of the training environment, providing a standardized encapsulation for subsequent deployment in containerized environments.

[0041] During the deployment phase of the online inference service, the system pre-integrates the Scikit-learn machine learning algorithm library into the base image of the compute service container. When the compute service container starts and initializes, the bootstrap script automatically reads the binary model file in the storage volume, deserializes it, and loads it into the server's random access memory, constructing a resident inference engine object. This loading mechanism ensures that the model only needs to be loaded once when the service starts, and can then be directly invoked in memory when processing concurrent requests, without repeated disk read operations, thus significantly reducing system response latency.

[0042] The system utilizes a standard RESTful application programming interface (API) developed based on the Flask web framework to expose prediction services. This API is designed to receive POST requests in JSON format. The request body contains the comprehensive risk index of the pipe segment calculated in step S3, as well as attribute data such as the segment's deformation degree and defect density. Upon receiving the request, the API immediately invokes the in-memory inference engine object, mapping the input JSON data into feature vectors and feeding them into the decision tree model for real-time prediction. The inference engine outputs the corresponding repair process category code based on the decision tree's path. This code will subsequently be used by the automatic cost calculation module to index specific engineering quota data.

[0043] In the preferred embodiment, the logical judgment steps executed by the repair process matching engine in step S4 include: Decision tree root node determines comprehensive risk index ,like Entering a high-risk (Category A) branch; if Enter the risk branch of category B; if Enter the low-risk C-category branch; Under the Class A high-risk branch, read the pipe segment deformation parameters. If the deformation is less than 10%, output the UV-cured CIPP overall repair plan; if the deformation is greater than or equal to 10% and less than 30%, output the mechanical spiral winding repair plan; if the deformation is greater than or equal to 30%, output the excavation and replacement plan. Under the risk branch in category B, read the number of defect points per 100 meters. If there are more than 5, switch to the category A branch for processing; if there are 5 or less, output a stainless steel double expansion ring point repair scheme.

[0044] In this embodiment, the logical judgment step executed by the repair process matching engine in step S4 uses a hierarchical decision tree algorithm to accurately classify the pipe segments and match them with corresponding remediation technologies. The root node of the decision tree first receives the comprehensive risk index output in step S3. As a primary discriminant variable, the system sets two key risk thresholds: 0.8 and 0.6. The engine compares the input comprehensive risk index with these two thresholds. When the comprehensive risk index is greater than or equal to 0.8, the system determines that the pipe section is in a high-risk state of structural failure and directs the process to a Class A high-risk branch. When the comprehensive risk index is less than 0.8 and greater than or equal to 0.6, the system determines that the pipeline segment is in a medium-risk state of functional degradation and directs the process to a Class B medium-risk branch. When the value is less than 0.6, the system determines that the pipe section is in a relatively safe low-risk state and directs the process to a Class C low-risk branch.

[0045] When the process enters the high-risk (Category A) branch, the inference engine executes the second-level logical judgment, at which point the judgment variable retrieved is the pipe segment's deformation parameter. This parameter reflects the degree of change in the geometric shape of the pipe's cross-section. The system matches three different structural repair processes based on the magnitude of the deformation value. If the pipe segment deformation value is less than 10%, it indicates that although the pipe has defects, its original structural shape is still good. The system outputs a UV-cured CIPP overall repair solution, using UV curing technology to form a high-strength resin lining inside the old pipe. If the pipe segment deformation value is greater than or equal to 10% but less than 30%, it indicates that the pipe has undergone significant elliptical deformation. The system outputs a mechanical spiral winding repair solution, using strip profiles to wind inside the old pipe to form a new structural pipe. If the pipe segment deformation value is greater than or equal to 30%, it indicates that the pipe structure has suffered severe collapse or lost its load-bearing capacity, and trenchless technology is no longer applicable. The system outputs an excavation and replacement solution, i.e., excavating the road surface and replacing it with new pipe material.

[0046] When the process enters the risk-prone branch (Category B), the inference engine retrieves the number of defect points per 100 meters as the second-level judgment variable. This variable measures the density of defect distribution in the pipe section. The system sets a threshold of 5 points. If the number of detected defect points per 100 meters is greater than 5, it indicates that although the risk of a single point has not reached an extremely high level, the defect distribution is too dense, and point-based repair is no longer economically or technically feasible. The system then forcibly transfers the handling logic for this pipe section to the high-risk branch (Category A), and re-selects the overall repair process based on the deformation parameters. If the number of detected defect points per 100 meters is less than or equal to 5, it indicates that the defects are scattered. The system outputs a stainless steel double-expansion ring point-based repair scheme, performing local reinforcement treatment on specific defect points. Through the above multi-level judgment logic, the system achieves automated and standardized mapping from risk assessment to engineering processes.

[0047] In the preferred embodiment, the cost aggregation step performed by the automatic cost calculation module based on the depth-first search (DFS) algorithm in step S4 includes: Initialize an empty stack Stack and a collection Visited for recording visited nodes, and push the starting node of the network topology onto Stack. While the Stack is not empty, execute the loop: pop the top element from the stack as the current processing segment. ; like If it is not marked in Visited, then read the length of the pipe segment. burial depth and repair process unit price Introducing the concept of burial depth The difficulty of construction is determined by road traffic conditions. ; Calculate the direct repair cost of the current pipe segment. And add it to the total cost variable TotalCost; Will Mark as Visited and find all related to Push directly connected but unaccessed downstream pipe segments into the Stack; Repeat the above loop until Stack is empty, output the final accumulated TotalCost as the direct cost of the project, and calculate the measures cost and regulatory fees accordingly.

[0048] In this embodiment, the automatic cost calculation module described in step S4 employs a cost aggregation strategy based on the depth-first search (DFS) algorithm to accurately calculate the estimated investment amount for large-scale pipeline repair projects. The system first abstracts the complex drainage network into a directed graph structure composed of nodes and edges, where edges represent pipe segments to be repaired and nodes represent inspection wells. The module initializes an empty stack (Stack) in memory as a traversal container and creates a set (Visited) to record visited nodes to prevent infinite loops and repeated calculations in ring-shaped networks. The system automatically identifies the starting node of the pipeline topology, typically the source pipe segment of the catchment area or a designated upstream starting point, and pushes this starting node onto the stack (Stack) to initiate the traversal process.

[0049] When the stack is not empty, the system enters a loop processing logic. In each loop, the system pops the top element of the stack as the current processing segment. The system immediately queries the Visited set; if the currently processed segment... If a pipe segment is not yet marked in the set, it means that its cost has not been calculated. At this point, the system calls the database interface to read the physical attribute data of the pipe segment, including its length. and pipeline burial depth Simultaneously, based on the repair process type matched in the previous steps, the system retrieves the corresponding repair process unit price from the engineering quota database. To reflect cost differences under different construction environments, the system introduces a construction difficulty coefficient. This coefficient is determined by the pipe burial depth. The coefficient value is determined by the road surface traffic conditions; for example, deeply buried pipelines or pipelines located under main roads will correspond to higher coefficient values.

[0050] Based on the parameters obtained above, the system performs direct repair cost calculation for a single pipe segment. The calculation process not only considers the product of the quantity and unit price, but also adjusts the cost using a difficulty coefficient. The calculation formula is as follows: ; In the formula, The direct repair cost for the current pipeline segment is expressed in RMB. This represents the length of the current pipe segment, in meters. This represents the unit price of the repair process corresponding to this pipe section, in yuan per meter; This represents the construction difficulty coefficient, which is a dimensionless value greater than or equal to 1. After calculation, the system will... The total cost is added to the globally defined project total cost variable, TotalCost.

[0051] After completing the cost calculation for the current pipe segment, the system will The node is marked and stored in the `Visited` set, indicating that the node has been processed. Then, the system traverses the topology database to find all nodes related to it. Downstream pipe segments that are directly physically connected but not yet present in the Visited set are identified. The system pushes these newly discovered downstream pipe segments onto the Stack, allowing for continued exploration along the pipe network branches in the next loop. The system repeats the pop, calculate, mark, and push operations until the Stack is empty, signifying that all connected pipe segments awaiting repair have been traversed. At this point, the system outputs the variable TotalCost as the direct cost of the entire project. Based on this, it further calculates the costs of measures, regulatory fees, and taxes according to nationally stipulated rates, ultimately generating a complete project investment estimate report.

[0052] In the preferred embodiment, step S5, which involves performing a post-treatment effect evaluation process based on statistical significance testing, includes: The sample set consists of minimum nighttime flow data for seven consecutive days following the completion of the restoration project. A sample set was constructed by retrieving the minimum nighttime flow data for seven consecutive days under the same weather conditions before the repair. ; Constructing the null hypothesis Alternative hypothesis ; Calculate the average of the differences between the two groups of samples. and the standard deviation of the difference ; Calculate the statistic using the paired samples t-test formula. The value, the formula is: ; In the formula: To test the statistic, It is the arithmetic mean of the differences between the two sets of samples. The standard deviation of the sample difference. The number of paired samples, This represents the overall mean.

[0053] In this embodiment, step S5 executes a post-repair effect evaluation process based on statistical significance testing. This process aims to objectively verify the effectiveness of the engineering treatment through rigorous data comparison, avoiding misjudgments based solely on single-day data fluctuations. The system first performs sample data collection and construction. After the repair project is fully completed and the equipment is confirmed to be operating stably, the system continuously collects nighttime minimum flow data for 7 days, defining this set of data as the post-repair sample set. To ensure the scientific validity of the comparison, the system simultaneously retrieved seven consecutive days of nighttime minimum flow data from the historical database, taken under the same weather conditions as the sample set before and after the repair work. This set of historical data was defined as the sample set before the repair. The same weather conditions here refer to no rainfall or rainfall within the same order of magnitude during the data collection period, in order to eliminate the interference of rainwater infiltration on the baseline flow.

[0054] After preparing the comparison samples, the system constructs a statistical hypothesis testing framework. First, the null hypothesis is set. This assumption represents the scenario where the remediation measures are ineffective or fail to achieve the expected reduction effect, i.e., the overall mean of the minimum nighttime flow after remediation. Greater than or equal to the overall mean of the minimum nighttime flow before the repair Correspondingly, alternative hypotheses are set. This hypothesis represents the scenario where the remediation measures are significantly effective, i.e., the overall mean of the minimum nighttime flow after remediation. Strictly less than the overall average of the minimum nighttime flow before the repair The system uses hypothesis testing to determine whether the null hypothesis should be rejected and the alternative hypothesis accepted at a given significance level.

[0055] To perform specific statistical calculations, the system first pairs the two sets of sample data, calculating the traffic difference for each pair of corresponding dates, thus obtaining a difference sequence containing seven values. Subsequently, the system calculates the arithmetic mean of this difference sequence, denoted as... Simultaneously, the system uses the standard deviation calculation formula to derive the sample standard deviation of this difference sequence, denoted as . Based on the statistical characteristic values ​​calculated above, the system uses the paired-samples t-test formula to calculate the test statistic. The value is calculated using the following formula: ; In the formula, This represents the calculated test statistic value, which is used to determine significance in subsequent table lookups. The arithmetic mean of the differences between the corresponding data of the two groups of samples before and after the repair reflects the average level of flow change; The standard deviation, representing the sample difference, reflects the degree of dispersion in the magnitude of flow variation; This represents the number of paired samples, which is 7 in this embodiment; Representing the population mean, this is used in hypothesis testing theory to describe the central tendency of a population. This formula measures whether an observed decrease in flow is statistically significant by standardizing the mean difference.

[0056] In the preferred scheme, step S5 is based on the statistical measure. The steps for determining the repair effect include: According to degrees of freedom Consult the standard t-distribution critical value table to obtain the one-tailed critical value at a significance level of 0.05. ; Comparison of calculations Value and ; like Furthermore, the corresponding p-value is less than 0.05, therefore the null hypothesis is rejected. If the repair work is deemed effective, the system automatically updates the pipe section status to healthy and archives it. like Unable to reject the null hypothesis If the repair is deemed ineffective, the system automatically generates a secondary inspection work order, indicating that there may be undetected branch pipe misconnections or structural leaks.

[0057] In this embodiment, step S5, based on statistical measures... The steps for determining the repair effect employ rigorous hypothesis testing and decision-making logic. The system first calculates the degrees of freedom. Equal to the number of paired samples Subtract 1. The system has a built-in standard t-distribution critical value table, based on the calculated degrees of freedom. And with a pre-set significance level of 0.05, retrieve the corresponding one-tailed critical value. A significance level of 0.05 means that the system tolerates a probability of no more than 5% of making a Type I error, which is the probability of incorrectly rejecting the null hypothesis when it is true.

[0058] After obtaining the critical value, the system will calculate the test statistic. Value and the critical value of negation Perform numerical comparisons. Due to the alternative hypothesis... This is a left-side test, meaning the expected flow rate after repair is significantly less than the flow rate before repair; therefore, the rejection region is located at the left tail of the t-distribution curve. If the calculated... Value strictly less than Furthermore, if the corresponding probability value (P-value) is less than 0.05, it indicates that the data falls within the rejection region, and the difference is statistically significant. In this case, the system rejects the null hypothesis. And accept the alternative hypothesis. The system determines that the repair work is effective and confirms that the structural defects of the pipe section have been fundamentally addressed. The system then automatically updates the health status field of the pipe section in the database to "healthy" and archives the repair project file.

[0059] Conversely, if the calculated Value greater than or equal to This indicates that the data did not fall within the rejection region, and the observed decrease in traffic may simply be due to random error; the difference is not statistically significant. In this case, the system cannot reject the null hypothesis. If the repair is deemed ineffective, it indicates that unresolved external water infiltration issues still exist within the pipe section. The system logic determines that the repair objective has not been achieved and automatically generates a secondary inspection work order, which is then pushed to the maintenance terminal. This work order prompts inspectors to focus on potential branch pipe misconnections, hidden structural leaks, or defects in repair construction quality, guiding subsequent remedial measures.

[0060] Example 2 Further explanation in conjunction with Example 1, such as Figure 1-2 As shown, this embodiment provides a decision support method for rapid screening and repair of structural defects in drainage pipe networks in areas with high groundwater levels. In the implementation process, a topological map of the drainage pipe network is first constructed based on a geographic information system (GIS). Graph theory algorithms are used to identify manholes at the junctions of main pipes with a catchment area greater than 0.5 square kilometers as primary monitoring nodes. Simultaneously, manholes at the end branches that have historically reported overflow events are selected as secondary monitoring nodes. Multi-sensor fusion IoT monitoring terminals are deployed within the selected monitoring manholes. Specifically, the probe of a Doppler ultrasonic flow meter is fixed at the bottom elevation of the pipe, ensuring the probe faces the water flow direction and its centerline is parallel to the pipe axis to accurately measure the average flow velocity of the cross-section. Simultaneously, a submersible hydrostatic level gauge is installed 8 to 12 centimeters above the bottom of the pipe, with its cable fixed along the manhole wall and extending to the data transmission unit at the manhole opening. The system is set to a sampling frequency of once every 5 to 8 minutes, simultaneously acquiring flow velocity and level sequence data, and transmitting the data back to the cloud server via a 4G or 5G communication network.

[0061] The data processing center is deployed based on a microservice architecture. A Docker Swarm container orchestration environment is built within the server cluster, and a virtual internal network based on an overlay driver is created to isolate communication traffic between different microservices. An InfluxDB time-series database service container is deployed to store high-frequency sampled monitoring data. The influxdb directory in the lib subdirectory of the host machine's file path is mounted to the data storage volume within the container for data persistence. The container's port 8086 is mapped to the host machine's port 8086. The system also deploys a PostgreSQL relational database service container to store the pipeline topology, and a computing service container based on the Python Flask framework to execute core algorithms. The computing service container automatically discovers the database service via internal DNS and sets the number of replicas to at least three for load balancing. An Nginx reverse proxy container acts as an API gateway to receive external requests. After the data enters the computing service, the system first uses the interquartile range (IQR) method to clean the raw traffic sequence, removing outliers caused by sensor drift or water flow disturbance. The system defines the upper quartile as the 75th percentile of the flow sequence values ​​arranged in ascending order, and the lower quartile as the 25th percentile. Next, the interquartile range is calculated, which is equal to the upper quartile minus the lower quartile. Data points falling outside the valid data range are identified as abnormal noise, and the arithmetic mean of the five consecutive normal time points before and after the abnormal point is used to fill the gap. The formula for calculating the valid data range is as follows: ; In the formula: This indicates the range of valid data values. This represents the value at the 25th position after the flow sequence values ​​are arranged in ascending order; This represents the value at the 75th percentile of the flow sequence. Based on this, the system calculates the flow rate using the post-cleaning flow velocity and liquid level, extracts the average flow rate from 02:00 to 04:00 AM as the minimum nighttime flow characteristic value, and combines this with a dry season flow correction coefficient and the upstream transferred wastewater volume to calculate groundwater infiltration. The calculation formula is as follows: ; In the formula: The calculated groundwater infiltration rate; The extracted minimum flow characteristic value at night; This is a dry season flow correction factor, with a value ranging from 0.9 to 0.95; The theoretical basis for the transfer of domestic sewage upstream of this pipeline section is the amount of sewage.

[0062] After calculating the infiltration rate, the system establishes a multi-factor coupled risk assessment model by combining the geographical environment parameters and physical attribute parameters of the pipeline segment. This model selects groundwater infiltration rate, soil permeability coefficient, pipeline service life, and pipeline burial depth as four core risk assessment factors, and introduces the design full-pipe flow rate, the maximum soil permeability coefficient in the region, the pipeline design service life, and the average groundwater depth in the region as normalized benchmark parameters. The system uses the analytic hierarchy process (AHP) to determine the weight coefficients corresponding to these four evaluation factors, ensuring that the sum of the four weight coefficients equals 1. Subsequently, the ratios of each factor value to the benchmark parameter are linearly weighted and summed to obtain the comprehensive risk index of the pipeline segment, calculated using the following formula: ; In the formula: The comprehensive risk index for the pipeline segment is a dimensionless value between 0 and 1. This is the design full-flow rate for this pipe section; The soil permeability coefficient of the area where the pipe section is located; This represents the highest soil permeability coefficient in the region's historical records. The pipeline's service life; The design service life corresponding to the pipe material; The depth of the pipeline burial; This represents the average depth of the groundwater level in the area. Weights for infiltration factors; For environmental factor weights; Weights for aging factors; The weights are for the burial depth factor.

[0063] After obtaining the comprehensive risk index, the system inputs it into a decision tree-based repair process matching engine built on the CART algorithm. This engine is built and serialized into a binary model file based on historical data in an offline training environment, and loaded into memory when the container starts. The root node of the decision tree first receives the comprehensive risk index as a primary discriminant variable. When the comprehensive risk index is greater than or equal to 0.8, the system determines that the pipe segment is in a high-risk state of structural failure and directs the process to a high-risk (A) branch. When the comprehensive risk index is less than 0.8 but greater than or equal to 0.6, the system determines that the pipe segment is in a medium-risk state of functional degradation and directs the process to a medium-risk (B) branch. When the comprehensive risk index is less than 0.6, the system determines that the pipe segment is in a relatively safe low-risk state and directs the process to a low-risk (C) branch. Under the high-risk (A) branch, the engine reads the pipe segment deformation parameters. If the deformation is less than 10%, it outputs a UV-cured CIPP overall repair scheme; if the deformation is greater than or equal to 10% but less than 30%, it outputs a mechanical spiral winding repair scheme; if the deformation is greater than or equal to 30%, it outputs an excavation and replacement scheme. Under the risk-prone branch of Class B, the engine reads the number of defect points per 100 meters. If the number of defect points detected per 100 meters is greater than 5, the system will forcibly switch the handling logic of the pipe section to the high-risk branch of Class A; if it is less than or equal to 5, the system will output a stainless steel double expansion ring point repair solution.

[0064] After determining the repair plan, the system calls the automatic cost calculation module and uses a depth-first search algorithm to traverse all pipe segments to be repaired. The system initializes an empty stack and a set to record visited nodes, pushing the starting node onto the stack. When the stack is not empty, the top element is popped as the currently processed pipe segment. If the pipe segment is not marked in the set, its length, burial depth, and repair process unit price are read, and a construction difficulty coefficient determined by the burial depth and road traffic conditions is introduced to calculate and accumulate the direct repair cost. The formula for calculating the cost of a single pipe segment is as follows: ; In the formula: This represents the direct repair cost of the current pipeline segment node; This represents the length of the current pipe segment; The unit price for the repair process; The construction difficulty coefficient is then used. The pipe segment is then marked as visited, and connected unvisited downstream pipe segments are pushed onto the stack until the stack is empty. Finally, the total project cost is output.

[0065] After the restoration project is completed, the system executes a post-restoration evaluation process based on statistical significance testing. The system collects minimum nighttime flow data for seven consecutive days after the restoration project completion to form a post-test sample set, and retrieves minimum nighttime flow data for seven consecutive days under the same weather conditions before the restoration to form a pre-test sample set. The system constructs a null hypothesis that the population mean after restoration is greater than or equal to the population mean before restoration, and an alternative hypothesis that the population mean after restoration is less than the population mean before restoration. By calculating the average and standard deviation of the differences between the two sample sets, the paired-samples t-test formula is used to calculate the statistic, as follows: ; In the formula: To test the statistic; It is the arithmetic mean of the differences between the two groups of samples; The standard deviation of the sample difference; The number of paired samples; This represents the population mean. The system queries the standard t-distribution table based on the degrees of freedom to obtain the one-tailed critical value at a significance level of 0.05. If the calculated statistic is less than the negative critical value and the probability value is less than 0.05, the system rejects the null hypothesis, determines the repair project is effective, and updates the pipe segment status to healthy; otherwise, it determines the repair effect is insignificant and automatically generates a secondary investigation work order.

[0066] The above embodiments are merely preferred technical solutions of the present invention and should not be considered as limitations on the present invention. The scope of protection of the present invention should be limited to the technical solutions described in the claims, including equivalent substitutions of the technical features described in the claims. That is, equivalent substitutions and improvements within this scope are also within the scope of protection of the present invention.

Claims

1. A rapid screening and repair decision support method for structural defects in drainage pipe networks, characterized by: the method include: S1. Construct a topology diagram of the drainage pipe network, identify key nodes based on the topology diagram as monitoring points for suspected defective pipe sections, configure multi-sensor fusion IoT monitoring terminals at the monitoring points, and collect pipe section liquid level sequence data and flow velocity sequence data. S2. The data processing center, deployed based on a microservice architecture, receives data uploaded by IoT monitoring terminals, calls anomaly detection algorithms to clean the data, and calculates the groundwater infiltration rate of the pipe section. S3. By combining the geographical environment parameters and physical attribute parameters of the pipeline segment, a multi-factor coupled risk assessment model is established, and the comprehensive risk index of the pipeline segment is calculated using a linear weighted algorithm. S4. Input the comprehensive risk index into the decision tree-based repair process matching engine to generate targeted repair technology solutions, and use the automatic cost calculation module to generate a bill of quantities and investment estimate. S5. After the repair project is completed, a post-repair effect evaluation process based on statistical significance testing is implemented to verify the degree of restoration of the pipeline network's operational status.

2. The decision support method for rapid screening and repair of structural defects in drainage pipe networks according to claim 1, characterized in that: Step S1, which involves configuring an IoT monitoring terminal at a node of a suspected defective pipe section, includes: Inspection wells at the junctions of main pipes with a catchment area greater than 0.5 square kilometers in the pipeline network topology were selected as primary monitoring nodes; The terminal branch manholes that have historically reported overflow events were selected as secondary monitoring nodes; A Doppler ultrasonic velocimeter (4) is installed at the bottom elevation of the selected inspection well. The probe of the Doppler ultrasonic velocimeter (4) faces the direction of water flow, and the center line of the probe is parallel to the axis of the pipe. It is used to measure the average flow velocity of the cross section. An immersion hydrostatic level gauge (3) is installed 8-12 cm above the bottom of the selected inspection well. The cable of the immersion hydrostatic level gauge (3) is fixed along the well wall and extends to the data transmission unit (1) at the wellhead. The sampling frequency is set to once every 5-8 minutes to simultaneously collect flow rate and liquid level data, which are then transmitted back via 4G / 5G network.

3. The decision support method for rapid screening and repair of structural defects in drainage pipe networks according to claim 1, characterized in that: In step S2, the data processing center is deployed using containerization technology. The steps include: Build a Docker Swarm container orchestration environment in the server cluster and create a virtual internal network based on the overlay driver to isolate communication traffic between different microservices; Deploy the InfluxDB time-series database service container, mount the / var / lib / influxdb directory on the host machine to the data storage volume inside the container to achieve data persistence, and map the container's port 8086 to the host machine's port 8086; Deploy a PostgreSQL relational database service container, mount the / var / lib / postgresql / data directory on the host machine into the container, map port 5432, and use it to store pipeline topology data; Deploy a compute service container based on the Python Flask framework, configure the compute service container to automatically discover the IP address of the database container through the internal DNS service of Docker Swarm, and set the number of replicas of the compute service container to at least 3 to achieve load balancing and high availability; Configure the Nginx reverse proxy container as an API gateway to intercept external requests and distribute them to the virtual IP address of the compute service container.

4. The decision support method for rapid screening and repair of structural defects in drainage pipe networks according to claim 3, characterized in that: Step S2, which involves cleaning the data and calculating the groundwater infiltration rate, includes: Read the raw traffic sequence from the InfluxDB database container; Outliers in the original flow sequence are detected by the interquartile range method. The upper quartile is defined as Q3 and the lower quartile as Q1. The interquartile range (IQR) is calculated to be equal to Q3 minus Q1. Data points in the original flow sequence that are greater than Q3 plus 1.5 times IQR or less than Q1 minus 1.5 times IQR are identified as anomalous noise. These anomalous noise points are then replaced with the arithmetic mean of the five consecutive normal time points before and after them, resulting in a cleaned flow sequence. The formula for identifying outlier intervals is as follows: ； In the formula: This indicates the range of valid data values. This represents the value at the 25th percentile after the flow sequence values ​​are arranged in ascending order. This represents the value at the 75th position of the flow sequence; The flow rate is calculated using the flow velocity and liquid level after cleaning, and the average flow rate from 02:00 to 04:00 is extracted as the minimum flow rate for the night. And combined with the dry season flow correction coefficient and upstream transfer of basic sewage volume Calculate groundwater infiltration The calculation formula is: ； In the formula: For the calculated groundwater infiltration rate, The extracted minimum flow characteristic value at night, This is a dry season flow correction factor, with a value ranging from 0.9 to 0.

95. The theoretical basis for the transfer of domestic sewage upstream of this pipeline section is the amount of sewage.

5. The decision support method for rapid screening and repair of structural defects in drainage pipe networks according to claim 1, characterized in that: Step S3 The steps for establishing a multi-factor coupled risk assessment model and calculating a comprehensive risk index include: Selecting groundwater infiltration rate Soil permeability coefficient Pipeline service life and pipeline burial depth These are the four core risk assessment factors; Introduce the design full pipe flow rate respectively Maximum soil permeability coefficient in the region Pipeline design service life and the average depth of groundwater in the region As a normalization baseline parameter; The weight coefficients of the four evaluation factors were determined using the analytic hierarchy process (AHP). And the sum of the weights is 1; The comprehensive risk index of the pipeline section is obtained by linearly weighting and summing the ratios of each factor value to the benchmark parameter. The calculation formula is: ； In the formula: The comprehensive risk index for the pipeline segment is a dimensionless value between 0 and 1. Weights for infiltration factors; For environmental factor weights; Weights for aging factors; The weights are for the burial depth factor.

6. The decision support method for rapid screening and repair of structural defects in drainage pipe networks according to claim 1, characterized in that: Step S4, the deployment steps of the computer software and computer modules for the decision tree-based repair process matching engine, include: In the offline training environment, a CART classification and regression tree model is built based on historical pipeline defect data and expert repair case library, and the model structure is optimized by pruning algorithm to prevent overfitting. The trained decision tree model is serialized into a binary model file using Python's Joblib library; The Scikit-learn machine learning library is integrated into the compute service container, and binary model files are loaded into memory when the container starts to build an inference engine object; By exposing a RESTful API interface through the Flask framework, the system receives the comprehensive risk index and pipe segment attribute JSON data calculated and output in step S3, calls the inference engine object for real-time prediction, and outputs the repair process category code.

7. The decision support method for rapid screening and repair of structural defects in drainage pipe networks according to claim 6, characterized in that: The logical judgment steps executed by the repair process matching engine in step S4 include: Decision tree root node determines comprehensive risk index ,like Entering a high-risk (Category A) branch; if Enter the risk branch of category B; if Enter the low-risk C-category branch; Under the Class A high-risk branch, read the pipe segment deformation parameters. If the deformation is less than 10%, output the UV-cured CIPP overall repair plan; if the deformation is greater than or equal to 10% and less than 30%, output the mechanical spiral winding repair plan; if the deformation is greater than or equal to 30%, output the excavation and replacement plan. Under the risk branch in category B, read the number of defect points per 100 meters. If there are more than 5, switch to the category A branch for processing; if there are 5 or less, output a stainless steel double expansion ring point repair scheme.

8. The decision support method for rapid screening and repair of structural defects in drainage pipe networks according to claim 1, characterized in that: Step S4, in which the automatic cost calculation module performs cost aggregation based on the depth-first search (DFS) algorithm, includes the following steps: Initialize an empty stack Stack and a collection Visited for recording visited nodes, and push the starting node of the network topology onto Stack. While the Stack is not empty, execute the loop: pop the top element from the stack as the current processing segment. ; like If it is not marked in Visited, then read the length of the pipe segment. burial depth and repair process unit price Introducing the concept of burial depth The difficulty of construction is determined by road traffic conditions. ; Calculate the direct repair cost of the current pipe segment. And add it to the total cost variable TotalCost; Will Mark as Visited and find all related to Push directly connected but unaccessed downstream pipe segments into the Stack; Repeat the above loop until Stack is empty, output the final accumulated TotalCost as the direct cost of the project, and calculate the measures cost and regulatory fees accordingly.

9. The decision support method for rapid screening and repair of structural defects in drainage pipe networks according to claim 1, characterized in that: Step S5, which involves performing a post-repair effectiveness evaluation process based on statistical significance testing, includes the following steps: The sample set consists of minimum nighttime flow data for seven consecutive days following the completion of the restoration project. A sample set was constructed by retrieving the minimum nighttime flow data for seven consecutive days under the same weather conditions before the repair. ; Constructing the null hypothesis Alternative hypothesis ; Calculate the average of the differences between the two groups of samples. and the standard deviation of the difference ; Calculate the statistic using the paired samples t-test formula. The value, the formula is: ； In the formula: To test the statistic, It is the arithmetic mean of the differences between the two sets of samples. The standard deviation of the sample difference. The number of paired samples, This represents the overall mean.

10. The decision support method for rapid screening and repair of structural defects in drainage pipe networks according to claim 9, characterized in that: In step S5, based on the statistics The steps for determining the repair effect include: According to degrees of freedom Consult the standard t-distribution critical value table to obtain the one-tailed critical value at a significance level of 0.

05. ; Comparison of calculations Value and ; like Furthermore, the corresponding p-value is less than 0.05, therefore the null hypothesis is rejected. If the repair work is deemed effective, the system automatically updates the pipe section status to healthy and archives it. like Unable to reject the null hypothesis If the repair is deemed ineffective, the system automatically generates a secondary inspection work order, indicating that there may be undetected branch pipe misconnections or structural leaks.

Images