Sewage pipe network alignment design method and system based on artificial intelligence

Abstract

The invention discloses a sewage pipe network alignment design method and system based on artificial intelligence, and relates to the technical field of sewage pipe network alignment design, a sensor is deployed through a data acquisition module to collect pipe network data in real time, the state of a sewage pipe network can be monitored in real time, and the limitation of dependence on manual inspection traditionally is overcome; and the sensor can obtain accurate data, so that a reliable basis is provided for subsequent analysis. The data preprocessing and cleaning module ensures the quality of the data, noise and abnormity in the data are eliminated through normalization and cleaning, the accuracy of the data is improved, and efficient operation of a machine learning model is ensured. Through a machine learning fault diagnosis module, the system can recognize a potential fault point PP in advance by learning normal and abnormal operation modes of a pipe network. According to the intelligent prediction method, early warning can be performed in advance, and delayed response after a fault occurs is avoided. The automatic diagnosis mechanism greatly improves the fault processing efficiency and reduces the dependence on manual judgment.

Description

Technical Field

[0001] The present invention relates to the technical field of sewage pipe network alignment design, and specifically to a sewage pipe network alignment design method and system based on artificial intelligence. Background Art

[0002] With the acceleration of the urbanization process, the urban sewage treatment system is facing increasingly complex challenges. Specifically, as an important part of urban infrastructure, the sewage pipe network is involved in the rational allocation of water resources and the efficient treatment of sewage. In this field, the design and management of the sewage pipe network are the core contents, especially the planning, optimization, and real-time monitoring of the pipe network.

[0003] Currently, traditional sewage pipe network design and management rely on manual experience and analysis methods based on historical data, and there are many limitations. For example, in the design stage of the sewage pipe network, it is often impossible to accurately predict potential problems such as blockage, leakage, or corrosion that may occur in the future. Traditional methods often rely on on-site inspections and manual monitoring, which not only increases labor costs but also may lead to delayed responses to faults due to personnel experience and judgment errors. Even when pipe network failures occur.

[0004] The above problems mainly stem from the lack of automatic analysis and intelligent diagnosis of real-time data in traditional sewage pipe network management. Due to the lack of a feedback mechanism based on real-time data in the pipe network design stage, it is difficult to accurately predict the occurrence of potential problems, and manual inspections rely on on-site data and personnel experience, resulting in difficulty in comprehensively grasping the health status of the pipe network. More seriously, after a failure occurs, it often relies on manual judgment and the response is delayed, leading to possible secondary damage to the pipe network system and even affecting the entire urban drainage system and environmental quality. Summary of the Invention

[0005] Aiming at the deficiencies of the prior art, the present invention provides a sewage pipe network alignment design method and system based on artificial intelligence, which solves the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention is realized through the following technical solutions: A sewage pipe network alignment design system based on artificial intelligence, including a data acquisition module, a data preprocessing and cleaning module, a machine learning fault diagnosis module, a fault location and analysis module, an emergency response and repair strategy generation module, and a feedback optimization and continuous learning module;

[0007] The data acquisition module collects real-time data of the pipe network through sensors deployed inside the pipe network nodes to obtain the original data set S;

[0008] The data preprocessing and cleaning module cleans and normalizes the collected original data set S to obtain the pipe network data set SG;

[0009] The machine learning fault diagnosis module builds a sewage pipe network fault diagnosis model based on the acquired pipe network data set SG, and learns the normal operation mode and abnormal operation mode of the pipe network to obtain potential fault points PP;

[0010] The fault location and analysis module spatially locates the potential fault point PP obtained through the GIS system and the pipe network topology structure, determines the severity and repair priority of the fault, and obtains the severity assessment result S(PP) and the repair priority score Pf(PP);

[0011] The emergency response and repair strategy generation module starts the preset emergency response plan to generate a repair plan based on the severity assessment result S(PP) and the repair priority score Pf(PP), and obtains the repair result score Pre(PP);

[0012] The feedback optimization and continuous learning module provides feedback and adjustments to the sewage network fault diagnosis model and repair plan based on the repair result score Pre(PP).

[0013] Preferably, the data acquisition module includes a data acquisition unit and a data integration unit;

[0014] The data acquisition unit collects real-time operation data through sensors installed at various nodes of the pipe network. The sensors include flow sensors, pressure sensors, temperature sensors, vibration sensors, water quality sensors, and corrosion sensors.

[0015] The flow data Qe inside the pipeline is collected through the flow sensor, the pressure data Pe inside the pipeline is collected through the pressure sensor, the temperature data Te inside the pipeline is collected through the temperature sensor, the vibration data Ve of the pipeline is collected through the vibration sensor, the pollutant concentration Ce in the sewage is collected through the water quality sensor, and the corrosion rate Re of the inner wall of the pipeline is collected through the corrosion sensor;

[0016] Among them, pollutant concentration Ce includes chemical oxygen demand, total phosphorus and total nitrogen;

[0017] The data integration unit integrates the collected flow data Qe, pressure data Pe, temperature data Te, vibration data Ve, pollutant concentration Ce and corrosion rate Re to obtain an original data set S.

[0018] Preferably, the data preprocessing and cleaning module includes a data cleaning unit and a data normalization unit;

[0019] The data cleaning unit performs anomaly detection, missing value processing and noise filtering on the collected original data set S to obtain a data cleaning set SC;

[0020] Among them, anomaly detection includes using an outlier detection algorithm to remove data that exceeds the normal value range;

[0021] Handling missing values includes using the mean filling method to fill in the missing values in the data;

[0022] Noise filtering removes high-frequency noise in the data through low-pass filtering technology;

[0023] The data normalization unit normalizes the data cleaning set SC, converts data with different dimensions to a unified range, and obtains the pipe network data set SG;

[0024] The pipe network data set SG is obtained through the following formula:

[0025]

[0026] In the formula, SCd represents the d-th item of data in the data cleaning set SC, SCdmin represents the valley value of the d-th item of data in the data cleaning set SC, and SCdmax represents the peak value of the d-th item of data in the data cleaning set SC.

[0027] Preferably, the machine learning fault diagnosis module includes a data feature extraction and fusion unit and a fault diagnosis and prediction unit;

[0028] The data feature extraction and fusion unit extracts features from the pipe network data set SG, including the flow rate change rate ΔQe, the pollutant concentration change rate ΔCe, and the corrosion change rate ΔRe, and fuses them to form the feature set F;

[0029] The flow rate change rate ΔQe is obtained through the following formula:

[0030] ΔQe = Qe(t) - Qe(t - 1);

[0031] In the formula, Qe(t) represents the flow rate data at time t, and Qe(t - 1) represents the flow rate data at time t - 1;

[0032] The pollutant concentration change rate ΔCe is obtained through the following formula:

[0033] ΔCe = Ce(t) - Ce(t - 1);

[0034] In the formula, Ce(t) represents the pollutant concentration at time t, and Ce(t - 1) represents the pollutant concentration at time t - 1; the corrosion change rate ΔRe is obtained through the following formula:

[0035] ΔRe = Re(t) - Re(t - 1);

[0036] In the formula, Re(t) represents the corrosion rate at time t, and Re(t - 1) represents the corrosion rate at time t - 1.

[0037] Preferably, the fault diagnosis and prediction unit analyzes the feature set F through a support vector machine model, and constructs a sewage pipeline network fault diagnosis model in combination with the pipeline network data set SG;

[0038] First, use the pipeline network data set SG, the feature set F, and the manually labeled fault cases to construct a training set, including samples in normal state and fault state, and train the sewage pipeline network fault diagnosis model;

[0039] Among them, the sample represents a set of data in the feature set F and the corresponding fault type label, including blockage fault, leakage fault, and crack fault, and obtain the fault label set Y;

[0040] The fault label set Y is obtained through the following formula:

[0041] Y = {y 1 , y 2 ,..., y i};

[0042] In the formula, y i represents the label of the i-th sample; indicating that the fault type of this sample includes blockage fault, leakage fault, and crack fault;

[0043] Train the feature set F through the sewage pipeline network fault diagnosis model, learn the normal mode and abnormal mode of the pipeline network, and obtain the classification prediction result f(F) of the feature set F;

[0044] The classification prediction result f(F) is obtained through the following formula:

[0045] f(F) = ω T *F + b;

[0046] In the formula, ω represents the weight vector of the support vector machine, b represents the bias term, and ω T represents the result after the weight vector ω is transposed; the transpose operation changes a column vector into a row vector, or a row vector into a column vector. During the training process, the model will adjust ω and b so that different types of faults can be accurately distinguished;

[0047] Use the trained sewage pipeline network fault diagnosis model to perform fault detection on the classification prediction result f(F), and calculate and obtain the fault type yP predicted by the model;

[0048] The fault type yP predicted by the model is obtained through the following formula:

[0049] yP = argmax(f(F));

[0050] Wherein, argmax represents the maximum value in the set function, and f(F) represents the classification prediction result; the output value of the sewage pipe network fault diagnosis model, corresponding to the confidence of each fault type, selects the classification with the highest confidence as the prediction result;

[0051] Locate the position where the fault type yP appears through the sewage pipe network fault diagnosis model to obtain the potential fault point PP;

[0052] The potential fault point PP is obtained through the following formula:

[0053]

[0054] Wherein, p represents the set of all fault point positions, argmin represents the minimum value in the set function, Fj represents the feature of the j-th data point in the feature set F, Fp represents the feature of the predicted fault point position, and m represents the total number of data points.

[0055] Preferably, the fault location and analysis module includes a spatial location unit and a fault severity assessment unit;

[0056] The spatial location unit combines the geographical spatial information of the pipe network with the sensor data through the GIS system to perform spatial location on the potential fault point PP;

[0057] Each potential fault point PP corresponds to a spatial coordinate in the pipe network. By comparing with the sensor data through the GIS system, the actual geographical location D(PP) = xPP(x, y) is obtained;

[0058] Wherein, xPP represents the spatial coordinate of the potential fault point, and x and y represent the positions in the geographical coordinate system;

[0059] By analyzing the topological structure of the pipe network, determine the connection relationship between the potential fault point PP and other components of the pipe network including valves and pumping stations; judge whether the potential fault point PP is located on the main pipeline or branch, thereby affecting the flow rate and maintenance plan;

[0060] The formula for analyzing the topological relationship of the pipe network is as follows:

[0061] T(PP) = {connected(PP, C1), connected(PP, C2),...};

[0062] Wherein, T(PP) represents the topological relationship of the potential fault point PP; includes the connection information of this point with other nodes (pipes, valves, pumping stations); connected(PP, C1) represents the pipe network node directly connected to the potential fault point PP, and C1 represents the pipe network node.

[0063] Preferably, the fault severity assessment unit evaluates the health condition of the pipeline through the corrosion rate Re, the pipeline material coefficient ML, and the pipeline age PA, obtains the severity assessment result S(PP), and performs normalization processing;

[0064] The severity assessment result S(PP) is obtained through the following formula:

[0065]

[0066] In the formula, Remax represents the maximum value of the corrosion rate, PAmax represents the maximum value of the pipeline age, and α 1 、α 2 and α 3 respectively represent the preset weight values of the corrosion rate Re, the pipeline age PA, and the pipeline material coefficient ML;

[0067] According to the flow data Qe and the pollutant concentration Ce in the pipeline, the fault impact is evaluated to obtain the fault impact index I(PP); when a high-flow pipeline or a pipeline with a relatively high pollutant concentration fails, it will have a more serious impact on the system;

[0068] The fault impact index I(PP) is obtained through the following formula:

[0069]

[0070] By comprehensively considering the severity assessment result S(PP) and the fault impact index I(PP) of the pipeline, the repair priority score Pf(PP) is calculated, and the potential fault points PP are sorted according to the repair priority to obtain the priority data sorting set Pfix(SO); the pipelines with a greater impact on the system should be repaired first;

[0071] The repair priority score Pf(PP) is obtained through the following formula:

[0072]

[0073] In the formula, Dis(PP) represents the distance between the potential fault point PP and the main pipeline;

[0074] All the obtained potential fault points PP and the corresponding repair priority scores Pf(PP) are stored in the priority data set Pfix, Pfix = {(PP, Pfix(PP))|PP ∈ p};

[0075] In the formula, p represents the set of all fault point positions;

[0076] Sort the prioritized dataset Pfix from high to low according to the repair priority score Pf(PP) of each potential failure point PP, and obtain the prioritized data sorted set Pfix(SO) = sort(Pfix, key = Pf(PP), reverse = True);

[0077] Where sort represents sorting from high to low.

[0078] Preferably, the emergency response and repair strategy includes a repair plan generation unit and a repair evaluation and result scoring unit;

[0079] The repair plan generation unit generates a repair plan RPlan(PP) according to the severity evaluation result S(PP) and the repair priority score Pf(PP) of the potential failure point PP, including the repair method, required resources, repair time Time, and repair cost Cos;

[0080] The repair methods include cleaning the pipeline, replacing the pipeline, and injecting sealant;

[0081] The resources include maintenance personnel, tools, and materials;

[0082] The repair plan RPlan(PP) is obtained through the following formula:

[0083] RPlan(PP) = g(S(PP), Pf(PP), yP));

[0084] Where g() represents the repair plan generation function; select the most suitable repair method according to the severity, repair priority, and failure type of the failure point, and yP represents the failure type; including blockage, leakage, and crack, different types of failures may require different repair methods;

[0085] The repair evaluation and result scoring unit calculates and obtains the repair result score Pre(PP) according to the repair time Tim and repair cost Cos in the repair plan RPlan(PP);

[0086] The repair result score Pre(PP) is obtained through the following formula:

[0087]

[0088] The repair result score Pre(PP) is used to evaluate the effectiveness of the repair plan; the higher the value, the more effective the repair plan, that is, under the comprehensive consideration of time and cost during the repair process, the better the repair effect of the failure point.

[0089] Preferably, the feedback optimization and continuous learning module includes a feedback data analysis unit and a model and plan optimization unit;

[0090] The feedback data analysis unit compares the obtained repair result score Pre(PP) with the actual repair score Ptu(PP) to obtain the repair deviation ΔP(PP).

[0091] The repair deviation ΔP(PP) is obtained through the following formula:

[0092] ΔP(PP)=Pre(PP)-Ptu(PP);

[0093] The model and scheme optimization unit integrates the obtained repair deviation ΔP(PP), feature set F, pipe network data set SG, manually labeled fault cases, and the set p of fault point locations as a new training set to retrain the sewage pipe network fault diagnosis model;

[0094] The repair plan RPlan(PP) is changed through the repair deviation ΔP(PP) to obtain a new repair plan nRPlan(PP);

[0095] The new repair plan nRPlan(PP) is obtained through the following formula:

[0096] nRPlan(PP)=g(S(PP),Pf(PP),yP,ΔP(PP))).

[0097] An artificial intelligence-based sewage pipe network alignment design method includes the following steps:

[0098] Step 1: The data acquisition module collects the real-time data of the pipe network through the sensors deployed inside the pipe network nodes to obtain the original data set S;

[0099] Step 2: The data preprocessing and cleaning module cleans and normalizes the collected original data set S to obtain the pipe network data set SG;

[0100] Step 3: The machine learning fault diagnosis module constructs a sewage pipe network fault diagnosis model based on the obtained pipe network data set SG, and learns the normal operation mode and abnormal operation mode of the pipe network to obtain potential fault points PP;

[0101] Step 4: The fault location and analysis module spatially locates the obtained potential fault points PP through the GIS system and the pipe network topology structure, determines the severity and repair priority of the fault, and obtains the severity assessment result S(PP) and the repair priority score Pf(PP);

[0102] Step 5: The emergency response and repair strategy generation module starts a preset emergency response plan to generate a repair plan according to the severity assessment result S(PP) and the repair priority score Pf(PP), and obtains the repair result score Pre(PP);

[0103] Step 6: The feedback optimization and continuous learning module provides feedback and adjustment to the sewage pipe network fault diagnosis model and repair plan based on the repair result score Pre(PP).

[0104] The present invention provides an alignment design method and system for sewage pipe networks based on artificial intelligence, having the following beneficial effects:

[0105] (1) During system operation, through various sensors in the data acquisition module, the system can monitor various operation indicators of the pipe network in real time. Compared with traditional manual inspection and single data acquisition methods, this multi-dimensional real-time data acquisition can provide more comprehensive and accurate pipe network status information. This refined monitoring helps to detect potential faults at an early stage, reduce the errors and lag of relying on manual inspections, and improve the accuracy of fault prediction.

[0106] (2) Through the data feature extraction and fusion unit, the system can extract multi-dimensional features from the pipe network dataset, including the flow rate change rate ΔQe, the pollutant concentration change rate ΔCe, and the corrosion change rate ΔRe, and fuse these features to construct a feature set F. This method enables the system to comprehensively evaluate the operation status of the pipe network from multiple aspects, increasing the accuracy of fault prediction. Compared with traditional single feature extraction methods, the method of fusing multi-dimensional features can capture more complex pipe network behaviors, improving the fine-grainedness and accuracy of fault diagnosis.

[0107] The fault diagnosis and prediction unit analyzes the fused feature set F through a support vector machine model and can automatically identify the normal operation mode and abnormal mode of the pipe network. By using manually labeled fault cases to construct a training set, combined with the pipe network dataset SG and the feature set F, the sewage pipe network fault diagnosis model can gradually improve its fault detection and prediction capabilities through continuous training and optimization. Compared with traditional diagnosis methods based on manual experience, the trained model can learn in multiple pipe network states, thus automatically adapting to different fault types, including blockages, leaks, and cracks, and accurately identifying them.

[0108] (3) Through the spatial positioning unit combined with the GIS system, the system can combine the geographical spatial information of the pipe network with real-time sensor data to achieve precise spatial positioning of potential fault points PP. Through this technology, each potential fault point in the pipe network can be accurately calibrated, greatly improving the intelligent level of pipe network management. Through the analysis of the pipe network topology relationship, the system can clarify the connection relationship between potential fault points and other components in the pipe network, including valves and pump stations. This analysis helps the system understand the possible propagation paths and influence ranges of faults, so as to more efficiently allocate resources and formulate repair plans during repair.

[0109] (4) By analyzing the repair time Tim and repair cost Cos in the repair plan RPlan(PP), the repair result score Pre(PP) is calculated. This scoring mechanism can help judge the effectiveness of the repair plan and the cost - effectiveness of its implementation. Through this systematic scoring method, it can be ensured that the repair activities can not only meet the timeliness of pipe network restoration but also control the repair cost to the greatest extent. This systematic evaluation method is more objective and efficient than the traditional method relying on empirical judgment. The feedback data analysis unit compares the actual repair result with the expected result. By calculating the repair deviation ΔP(PP), the system can identify possible problems or deviations in the repair process. This feedback mechanism enables the system to continuously learn and optimize, ensuring the consistency between the repair strategy and the actual situation. Description of the Drawings

[0110] Figure 1 It is a schematic diagram of the block - flow diagram of the sewage pipe network alignment design system based on artificial intelligence of the present invention;

[0111] Figure 2 It is a schematic diagram of the steps of the sewage pipe network alignment design method based on artificial intelligence of the present invention;

[0112] Figure 3 It is an area chart of the pipeline health status assessment of the present invention;

[0113] Figure 4 It is a line chart of the severity assessment result S(PP) varying with the corrosion rate Re of the present invention. Detailed Embodiment

[0114] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0115] Embodiment 1

[0116] The present invention provides a sewage pipe network alignment design system based on artificial intelligence. Please refer to Figure 1 , which includes a data acquisition module, a data pre - processing and cleaning module, a machine - learning fault diagnosis module, a fault location and analysis module, an emergency response and repair strategy generation module, and a feedback optimization and continuous learning module;

[0117] The data acquisition module collects real - time data of the pipe network through sensors deployed inside the pipe network nodes to obtain the original data set S;

[0118] The data preprocessing and cleaning module cleans and normalizes the collected original dataset S to obtain the pipeline network dataset SG;

[0119] The machine learning fault diagnosis module constructs a sewage pipeline network fault diagnosis model based on the obtained pipeline network dataset SG, learns the normal operation mode and abnormal operation mode of the pipeline network, and obtains potential fault points PP;

[0120] The fault location and analysis module spatially locates the obtained potential fault points PP through the GIS system and the pipeline network topology structure, determines the severity and repair priority of the fault, and obtains the severity evaluation result S(PP) and the repair priority score Pf(PP);

[0121] The emergency response and repair strategy generation module starts the preset emergency response plan to generate a repair plan according to the severity evaluation result S(PP) and the repair priority score Pf(PP), and obtains the repair result score Pre(PP);

[0122] The feedback optimization and continuous learning module feeds back and adjusts the sewage pipeline network fault diagnosis model and the repair plan according to the repair result score Pre(PP).

[0123] In this embodiment, by deploying sensors through the data acquisition module to collect pipeline network data in real time, the state of the sewage pipeline network can be monitored in real time, overcoming the limitations of traditional manual inspections. The sensors can obtain accurate data, providing a reliable basis for subsequent analysis. At the same time, the data preprocessing and cleaning module ensures the quality of the data. Through normalization and cleaning, the noise and anomalies in the data are eliminated, improving the accuracy of the data and ensuring the efficient operation of the machine learning model. Through the machine learning fault diagnosis module, the system can identify potential fault points PP in advance by learning the normal and abnormal operation modes of the pipeline network. Different from the traditional fault diagnosis methods that rely on manual inspections and experience, this intelligent prediction method can give early warnings and avoid the delayed response after a fault occurs. This real-time and automatic diagnosis mechanism can greatly improve the efficiency of fault handling and reduce the dependence on manual judgment.

[0124] By combining the fault location and analysis module with the GIS system and the pipe network topology, potential fault points can be accurately located in space. This module can not only determine the specific location of the fault, but also analyze the severity of the fault and determine the priority of repair according to the layout of the pipe network and the flow of sewage. The emergency response and repair strategy generation module is based on real-time fault diagnosis and location results. It can automatically generate repair plans and evaluate the repair effects according to the severity of the fault and the repair priority. This module reduces the interference of human factors on the repair plan through automated solution generation, and improves the accuracy and response speed of the repair plan. At the same time, the repair result scoring is used to evaluate the repair effect, providing a basis for subsequent improvements and adjustments.

[0125] Example 2

[0126] This embodiment is explained in Example 1, please refer to Figure 1 ,Specifically: the data acquisition module includes a data acquisition unit and a data integration unit;

[0127] The data acquisition unit collects real-time operation data through sensors installed at various nodes of the pipe network. The sensors include flow sensors, pressure sensors, temperature sensors, vibration sensors, water quality sensors, and corrosion sensors.

[0128] The flow data Qe inside the pipeline is collected through the flow sensor, the pressure data Pe inside the pipeline is collected through the pressure sensor, the temperature data Te inside the pipeline is collected through the temperature sensor, the vibration data Ve of the pipeline is collected through the vibration sensor, the pollutant concentration Ce in the sewage is collected through the water quality sensor, and the corrosion rate Re of the inner wall of the pipeline is collected through the corrosion sensor;

[0129] Among them, pollutant concentration Ce includes chemical oxygen demand, total phosphorus and total nitrogen;

[0130] The data integration unit integrates the collected flow data Qe, pressure data Pe, temperature data Te, vibration data Ve, pollutant concentration Ce and corrosion rate Re to obtain an original data set S.

[0131] The data preprocessing and cleaning module includes a data cleaning unit and a data normalization unit;

[0132] The data cleaning unit performs anomaly detection, missing value processing and noise filtering on the collected original data set S to obtain a data cleaning set SC;

[0133] Among them, anomaly detection includes using an outlier detection algorithm to remove data that exceeds the normal value range;

[0134] Dealing with missing values ​​includes using mean imputation method to fill in missing values ​​in the data;

[0135] Noise filtering removes high-frequency noise in the data through low-pass filtering technology;

[0136] The data normalization unit normalizes the data cleaning set SC, converts data with different dimensions to a unified range, and obtains the pipe network data set SG;

[0137] The pipe network data set SG is obtained through the following formula:

[0138]

[0139] In the formula, SCd represents the d-th item of data in the data cleaning set SC, SCdmin represents the valley value of the d-th item of data in the data cleaning set SC, and SCdmax represents the peak value of the d-th item of data in the data cleaning set SC.

[0140] In this embodiment, through various sensors in the data acquisition module, the system can monitor various operating indicators of the pipe network in real time. Compared with traditional manual inspection and single data acquisition methods, this multi-dimensional real-time data acquisition can provide more comprehensive and accurate pipe network status information. This refined monitoring helps to detect potential faults at an early stage, reduce the errors and lag of relying on manual inspections, and improve the accuracy of fault prediction.

[0141] The data integration unit obtains the original data set S by integrating data from different sensors, making the data source more unified and easier to process. Through the data cleaning unit, the system can automatically detect and eliminate abnormal data, fill in missing values, and remove noise. This process can ensure the high quality and integrity of the data, ensure the accuracy of subsequent analysis and machine learning models, and avoid omissions and subjective errors that may occur during traditional manual data cleaning. The data cleaning and normalization unit effectively improves the quality of the data by automatically processing abnormal data, missing values, and noise, and removes high-frequency noise through low-pass filtering, avoiding the influence of external disturbances on the data during the analysis process. In addition, the data normalization technology converts data with different dimensions into a unified scale, ensuring the consistency of data from different sources, making subsequent analysis and model training more accurate and efficient.

[0142] Through the acquisition, cleaning, and normalization of data, the system can provide high-quality input data for subsequent fault diagnosis models. The model will be able to identify potential fault points PP in the pipe network with broader data support, improve the accuracy of fault prediction, and reduce false alarms and missed alarms. The intelligence level of this system can greatly improve the reliability of the entire pipe network, especially when facing complex and changing pipe network environments, with more prominent performance.

[0143] Embodiment 3

[0144] This embodiment is explained in Embodiment 2. Please refer to Figure 1 , specifically: The machine learning fault diagnosis module includes a data feature extraction and fusion unit and a fault diagnosis and prediction unit;

[0145] The data feature extraction and fusion unit extracts features from the pipe network data set SG, including the flow rate change rate ΔQe, the pollutant concentration change rate ΔCe, and the corrosion change rate ΔRe, and fuses them to form a feature set F;

[0146] The flow rate change rate ΔQe is obtained through the following formula:

[0147] ΔQe = Qe(t) - Qe(t - 1);

[0148] In the formula, Qe(t) represents the flow rate data at time t, and Qe(t - 1) represents the flow rate data at time t - 1;

[0149] The pollutant concentration change rate ΔCe is obtained through the following formula:

[0150] ΔCe = Ce(t) - Ce(t - 1);

[0151] In the formula, Ce(t) represents the pollutant concentration at time t, and Ce(t - 1) represents the pollutant concentration at time t - 1; The corrosion change rate ΔRe is obtained through the following formula:

[0152] ΔRe = Re(t) - Re(t - 1);

[0153] In the formula, Re(t) represents the corrosion rate at time t, and Re(t - 1) represents the corrosion rate at time t - 1.

[0154] The fault diagnosis and prediction unit analyzes the feature set F through a support vector machine model and constructs a sewage pipe network fault diagnosis model in combination with the pipe network data set SG;

[0155] First, use the pipe network data set SG, the feature set F, and the manually labeled fault cases to construct a training set, including samples in the normal state and the fault state, and train the sewage pipe network fault diagnosis model;

[0156] Among them, the sample represents a set of data in the feature set F and the corresponding fault type label, including blockage faults, leakage faults, and crack faults, and obtains a fault label set Y;

[0157] The fault label set Y is obtained through the following formula:

[0158] Y = {y 1 , y 2 ,..., y i};

[0159] where y i represents the label of the i-th sample;

[0160] The feature set F is trained by the sewage pipe network fault diagnosis model to learn the normal mode and abnormal mode of the pipe network, and the classification prediction result f(F) of the feature set F is obtained;

[0161] The classification prediction result f(F) is obtained through the following formula:

[0162] f(F) = ω T *F + b;

[0163] where ω represents the weight vector of the support vector machine, b represents the bias term, and ω T represents the result after the weight vector ω is transposed;

[0164] The trained sewage pipe network fault diagnosis model is used to perform fault detection on the classification prediction result f(F), and the fault type yP predicted by the model is calculated and obtained;

[0165] The fault type yP predicted by the model is obtained through the following formula:

[0166] yP = argmax(f(F));

[0167] where argmax represents the maximum value in the set function, and f(F) represents the classification prediction result;

[0168] The sewage pipe network fault diagnosis model is used to locate the position where the fault type yP appears, and the potential fault point PP is obtained;

[0169] The potential fault point PP is obtained through the following formula:

[0170]

[0171] where p represents the set of all fault point positions, argmin represents the minimum value in the set function, Fj represents the feature of the j-th data point in the feature set F, Fp represents the feature of the predicted fault point position, and m represents the total number of data points.

[0172] In this embodiment, through the data feature extraction and fusion unit, the system can extract multi-dimensional features from the pipe network dataset, including the flow rate change rate ΔQe, the pollutant concentration change rate ΔCe, and the corrosion change rate ΔRe, and fuse these features to construct the feature set F. This method enables the system to comprehensively evaluate the operation status of the pipe network from multiple aspects, increasing the accuracy of fault prediction. Compared with the traditional single feature extraction method, the method of fusing multi-dimensional features can capture more complex pipe network behaviors, improving the fine-grainedness and accuracy of fault diagnosis.

[0173] The fault diagnosis and prediction unit analyzes the fused feature set F through a support vector machine model and can automatically identify the normal and abnormal operation modes of the pipe network. By using the manually labeled fault cases to construct a training set and combining the pipe network dataset SG and the feature set F, the sewage pipe network fault diagnosis model can gradually improve its fault detection and prediction capabilities through continuous training and optimization. Compared with the traditional diagnosis method based on manual experience, the trained model can learn under multiple pipe network states, thus automatically adapting to different types of faults, including blockages, leaks, and cracks, and accurately identifying them.

[0174] Based on the classification and prediction results of the model, the system can not only identify the type of fault, but also locate the fault point based on the data in the feature set F and generate potential fault points PP. This fault location method can provide a more accurate fault point location compared with the traditional manual inspection and simple automatic detection methods, greatly improving the efficiency of subsequent repair work. Real-time location of potential fault points can significantly reduce the comprehensive inspection of the pipe network, thus saving maintenance costs and time.

[0175] Due to the advantages of the support vector machine model, the system can quickly adapt to different pipe network operation modes and still maintain a high diagnosis accuracy when facing different sewage pipe network scenarios and complex situations. This enables the system to be applied not only to the existing pipe network environment, but also to be extended to new pipe network systems or complex urban sewage systems in the future, providing flexible and scalable fault diagnosis solutions.

[0176] Example 4

[0177] This example is an explanatory note based on Example 3. Please refer to Figure 1 , specifically: The fault location and analysis module includes a spatial location unit and a fault severity assessment unit;

[0178] The spatial location unit combines the geographical spatial information of the pipe network with the sensor data through the GIS system to spatially locate the potential fault points PP.

[0179] Each potential fault point PP corresponds to a spatial coordinate in the pipe network. By comparing with the sensor data through the GIS system, the actual geographical location D(PP) = xPP(x, y) is obtained;

[0180] In the formula, xPP represents the spatial coordinate of the potential fault point, and x and y represent the positions in the geographical coordinate system;

[0181] By analyzing the topological structure of the pipe network, the connection relationship between the potential fault points PP and other components of the pipe network, including valves and pump stations, is determined;

[0182] The formula for analyzing the pipe network topological relationship is as follows:

[0183] T(PP) = {connected(PP, C1), connected(PP, C2),...};

[0184] Wherein, T(PP) represents the topological relationship of the potential fault point PP; connected(PP, C1) represents the pipeline network node directly connected to the potential fault point PP, and C1 represents the pipeline network node.

[0185] The fault severity assessment unit evaluates the health status of the pipeline through the corrosion rate Re, the pipeline material coefficient ML, and the pipeline age PA, and obtains the severity assessment result S(PP);

[0186] The severity assessment result S(PP) is obtained through the following formula:

[0187]

[0188] Wherein, Remax represents the maximum value of the corrosion rate, PAmax represents the maximum value of the pipeline age, and α 1 , α 2 and α 3 respectively represent the preset weight values of the corrosion rate Re, the pipeline age PA, and the pipeline material coefficient ML;

[0189] Specific example:

[0190] The preset α 1 , α 2 and α 3 are 0.4, 0.3, and 0.3 respectively;

[0191] The maximum corrosion rate Rmax = 1.0, and the maximum pipeline age PAmax = 50;

[0192] The corrosion rate Re = 0.5, the pipeline age PA = 20, and the pipeline material coefficient ML = 1.2;

[0193] Calculate to obtain the severity assessment result S(PP):

[0194]

[0195] Table 1: Pipeline health status assessment

[0196] Corrosion rate Re Pipe age PA Pipe material coefficient ML Severity assessment result S(PP) Group 1 0.5 20 1.2 0.68 Group 2 0.8 30 1.5 0.95 Group 3 0.4 40 1.0 0.7 Group 4 0.6 25 1.3 0.78 Group 5 0.7 35 1.4 0.91

[0197] Table 2: Table of changes in the corrosion rate Re. Under the condition that the pipeline age PA and the pipeline material coefficient ML remain unchanged, the influence of the corrosion rate Re on the severity assessment result S(PP):

[0198] Corrosion rate Re Pipe age PA Pipe material coefficient ML Severity assessment result S(PP) Group 1 0.4 20 1.2 0.64 Group 2 0.5 20 1.2 0.68 Group 3 0.6 20 1.2 0.72 Group 4 0.7 20 1.2 0.76 Group 5 0.8 20 1.2 0.80

[0199] Evaluate the fault impact based on the flow rate data Qe and pollutant concentration Ce in the pipeline, and obtain the fault impact index I(PP);

[0200] The fault impact index I(PP) is obtained through the following formula:

[0201]

[0202] By comprehensively considering the severity assessment result S(PP) of the pipeline and the fault impact index I(PP), calculate the repair priority score Pf(PP), and sort the potential fault points PP according to the repair priority, and obtain the priority data sorting set Pfix(SO);

[0203] The repair priority score Pf(PP) is obtained through the following formula:

[0204]

[0205] In the formula, Dis(PP) represents the distance between the potential fault point PP and the main pipeline;

[0206] Store all the obtained potential fault points PP and the corresponding repair priority scores Pf(PP) in the priority data set Pfix, Pfix = {(PP, Pfix(PP))|PP ∈ p};

[0207] In the formula, p represents the set of all fault point positions;

[0208] According to the repair priority score Pf(PP) of each potential fault point PP, sort the priority data set Pfix from high to low, and obtain the priority data sorting set Pfix(SO) = sort(Pfix, key = Pf(PP), reverse = True);

[0209] In the formula, sort represents sorting from high to low.

[0210] In this embodiment, through the spatial positioning unit combined with the GIS system, the system can combine the geographical spatial information of the pipe network with the real-time sensor data to achieve precise spatial positioning of the potential fault point PP. Through this technology, each potential fault point in the pipe network can be accurately calibrated, greatly improving the intelligent level of pipe network management. Through the analysis of the pipe network topology relationship, the system can clarify the connection relationship between the potential fault point and other components in the pipe network, including valves and pump stations. This analysis helps the system understand the possible propagation path and influence range of the fault occurrence, so as to allocate resources and formulate repair plans more efficiently during repair.

[0211] The fault severity assessment unit comprehensively evaluates the health status of the pipeline and the impact degree of faults by analyzing the corrosion rate Re of the pipeline, the pipeline material coefficient ML, and the pipeline age PA, in combination with flow data and pollutant concentration. Compared with the traditional single-factor analysis, this way of considering multiple factors can more comprehensively evaluate the potential risks of pipeline faults, help determine the areas that need to be repaired first, and ensure that limited resources are used in the most urgent places. Through the repair priority score Pf(PP) and the priority data sorting set Pfix, the system can perform dynamic priority sorting according to the severity and impact of each fault point. This method enables the system to give priority to repairing the fault points with the most serious impact or the most urgent situation, avoiding the unreasonable resource allocation or repair delay that may exist in the traditional method. The efficient priority sorting system ensures the timeliness and effectiveness of the pipeline network repair work.

[0212] The fault impact index I(PP) can accurately evaluate the scope and degree of the impact of faults on the system through the combined analysis of flow data and pollutant concentration. This accurate fault impact assessment enables the system to more reasonably allocate resources during the formulation of repair plans and ensure that key areas can be repaired in a timely manner, thus avoiding secondary damage caused by untimely handling of faults. Through continuous optimization and feedback mechanisms, the system can adjust the diagnostic model and repair strategy in real time according to the repair execution results. The continuously optimized system can better adapt to new fault types that may appear in the future and improve the long-term stability of the pipeline network operation. Over time, the system will gradually accumulate data and experience, become more intelligent and efficient, so as to maximize the effect and cost-effectiveness of pipeline network maintenance.

[0213] Embodiment 5

[0214] This embodiment is an explanatory description carried out in Embodiment 4, please refer to Figure 1 , specifically: The emergency response and repair strategy includes a repair plan generation unit and a repair evaluation and result scoring unit;

[0215] The repair plan generation unit generates a repair plan RPlan(PP) according to the severity assessment result S(PP) of the potential fault point PP and the repair priority score Pf(PP), including the repair method, the required resources, the time required for repair Time, and the repair cost Cos;

[0216] The repair methods include cleaning the pipeline, replacing the pipeline, and injecting sealant;

[0217] The resources include maintenance personnel, tools, and materials;

[0218] The repair plan RPlan(PP) is obtained through the following formula:

[0219] RPlan(PP) = g(S(PP), Pf(PP), yP));

[0220] Wherein, g() represents a repair plan generation function, and yP represents the fault type;

[0221] The repair evaluation and result scoring unit calculates and obtains the repair result score Pre(PP) according to the repair time Tim and repair cost Cos in the repair plan RPlan(PP);

[0222] The repair result score Pre(PP) is obtained through the following formula:

[0223]

[0224] The feedback optimization and continuous learning module includes a feedback data analysis unit and a model and plan optimization unit;

[0225] The feedback data analysis unit compares the obtained repair result score Pre(PP) with the actual repair score Ptu(PP) to obtain the repair deviation ΔP(PP);

[0226] The repair deviation ΔP(PP) is obtained through the following formula:

[0227] ΔP(PP) = Pre(PP) - Ptu(PP);

[0228] The model and plan optimization unit integrates the obtained repair deviation ΔP(PP), feature set F, pipe network data set SG, manually labeled fault cases, and the set p of fault point positions as a new training set to retrain the sewage pipe network fault diagnosis model;

[0229] The repair plan RPlan(PP) is changed through the repair deviation ΔP(PP) to obtain a new repair plan nRPlan(PP);

[0230] The new repair plan nRPlan(PP) is obtained through the following formula:

[0231] nRPlan(PP) = g(S(PP), Pf(PP), yP, ΔP(PP))).

[0232] In this embodiment, according to the severity evaluation result S(PP) of the potential fault point PP and the repair priority score Pf(PP), a repair plan RPlan(PP) is automatically generated. This plan not only includes appropriate repair methods, including pipe cleaning, pipe replacement, and sealant injection, but also considers the resources, time, and cost required for repair. This automated repair plan generation greatly reduces manual intervention, ensures the optimal allocation of resources, avoids human judgment errors, and improves the accuracy and efficiency of repair work.

[0233] The repair evaluation and result scoring unit analyzes the repair time Tim and repair cost Cos in the repair plan RPlan(PP) to calculate the repair result score Pre(PP). This scoring mechanism can help judge the effectiveness of the repair plan and the cost-effectiveness of its implementation. Through this systematic scoring method, it can ensure that the repair activities can not only meet the timeliness of pipe network restoration but also control the repair cost to the greatest extent. This systematic evaluation method is more objective and efficient than the traditional method that relies on empirical judgment. The feedback data analysis unit compares the actual repair result with the expected result. By calculating the repair deviation ΔP(PP), the system can identify potential problems or deviations in the repair process. This feedback mechanism enables the system to continuously learn and optimize, ensuring the consistency between the repair strategy and the actual situation.

[0234] Through the model and plan optimization unit, the system can optimize the fault diagnosis model and repair plan in real time based on the feedback analysis results. The repair plan RPlan(PP) will be adjusted according to the repair deviation ΔP(PP) to generate a new repair plan nRPlan(PP). This optimization mechanism ensures the adaptability and self-repair ability of the system during long-term operation, can continuously improve the efficiency of pipe network management, and avoids the inefficiency or resource waste caused by relying on fixed repair plans.

[0235] Through the automated repair plan generation, evaluation, and optimization process, the system can formulate more targeted repair strategies based on the real-time data of the pipe network and historical fault records. This intelligent decision-making mechanism can eliminate the interference of human factors on the selection of repair strategies, enabling the formulation of plans for each fault point based on accurate analysis data, thereby improving the operation efficiency and stability of the pipe network.

[0236] Example 6

[0237] For the method of sewage pipe network alignment design based on artificial intelligence, please refer to Figure 2 , specifically: including the following steps:

[0238] Step 1: The data acquisition module collects the real-time data of the pipe network through sensors deployed inside the pipe network nodes to obtain the original data set S;

[0239] Step 2: The data preprocessing and cleaning module cleans and normalizes the collected original data set S to obtain the pipe network data set SG;

[0240] Step 3: The machine learning fault diagnosis module constructs a sewage pipe network fault diagnosis model based on the obtained pipe network data set SG, and learns the normal operation mode and abnormal operation mode of the pipe network to obtain potential fault points PP;

[0241] Step 4: The fault location and analysis module uses the GIS system and the pipe network topology to spatially locate the obtained potential fault points PP, determine the severity of the fault and the repair priority, and obtain the severity assessment result S(PP) and the repair priority score Pf(PP);

[0242] Step 5: The emergency response and repair strategy generation module generates a repair plan by starting a preset emergency response plan according to the severity assessment result S(PP) and the repair priority score Pf(PP), and obtains the repair result score Pre(PP);

[0243] Step 6: The feedback optimization and continuous learning module provides feedback and adjustment to the sewage pipe network fault diagnosis model and the repair plan according to the repair result score Pre(PP).

[0244] In this embodiment, the data preprocessing and cleaning module cleans and normalizes the collected original data set S to form a unified pipe network data set SG. By eliminating abnormal data, filling in missing values, removing noise, and normalizing, this module enables data from different sources to be analyzed under a unified standard, avoiding incorrect diagnoses caused by data inconsistency or noise problems. Data cleaning and normalization improve the accuracy of subsequent machine learning models, ensuring the quality and reliability of the data.

[0245] Combined with the GIS system and the pipe network topology, the fault location and analysis module can accurately locate the position of potential fault points, and determine the severity and repair priority of the fault points through the topological relationship of the pipe network. Compared with traditional methods that rely on manual judgment and simple location, this method based on geographic information system and topological analysis can more accurately identify the influence range and urgency of fault points, provide more scientific repair plan suggestions, and reduce human judgment errors.

[0246] The emergency response and repair strategy generation module automatically generates a repair plan according to the severity assessment result S(PP) and the repair priority score Pf(PP) of the fault point, and conducts a repair result score Pre(PP). The innovation of this module lies in its ability to intelligently generate the most suitable repair strategy for each fault point according to different fault types and repair priorities, while evaluating the repair effect, required time, and cost, avoiding omissions and unreasonable plans in the human decision-making process. Automated repair evaluation not only improves repair efficiency but also saves labor and time costs.

[0247] The feedback optimization and continuous learning module compares the repair result score Pre(PP) with the actual repair result, automatically calculates the repair deviation, and feeds it back to the fault diagnosis model and repair plan for optimization and adjustment. This continuous learning mechanism ensures that the system can continuously improve and optimize according to actual operation experience, enhancing the accuracy of future fault detection and repair.

[0248] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. The artificial intelligence-based sewage pipe network alignment design system is characterized by: It includes data acquisition module, data preprocessing and cleaning module, machine learning fault diagnosis module, fault location and analysis module, emergency response and repair strategy generation module and feedback optimization and continuous learning module; The data acquisition module collects real-time data of the pipeline network through sensors deployed inside the pipeline network nodes to obtain the original data set S; The data preprocessing and cleaning module cleans and normalizes the collected original data set S to obtain the pipe network data set SG; The machine learning fault diagnosis module builds a sewage pipe network fault diagnosis model based on the acquired pipe network data set SG, and learns the normal operation mode and abnormal operation mode of the pipe network to obtain potential fault points PP; The fault location and analysis module spatially locates the potential fault point PP obtained through the GIS system and the pipe network topology structure, determines the severity and repair priority of the fault, and obtains the severity assessment result S(PP) and the repair priority score Pf(PP); The emergency response and repair strategy generation module starts the preset emergency response plan to generate a repair plan based on the severity assessment result S(PP) and the repair priority score Pf(PP), and obtains the repair result score Pre(PP); The feedback optimization and continuous learning module provides feedback and adjustments to the sewage network fault diagnosis model and repair plan based on the repair result score Pre(PP).

2. The artificial intelligence-based sewage pipe network alignment design system according to claim 1, characterized in that: The data acquisition module includes a data acquisition unit and a data integration unit; The data acquisition unit collects real-time operation data through sensors installed at various nodes of the pipe network. The sensors include flow sensors, pressure sensors, temperature sensors, vibration sensors, water quality sensors, and corrosion sensors. The flow data Qe inside the pipeline is collected through the flow sensor, the pressure data Pe inside the pipeline is collected through the pressure sensor, the temperature data Te inside the pipeline is collected through the temperature sensor, the vibration data Ve of the pipeline is collected through the vibration sensor, the pollutant concentration Ce in the sewage is collected through the water quality sensor, and the corrosion rate Re of the inner wall of the pipeline is collected through the corrosion sensor; Among them, pollutant concentration Ce includes chemical oxygen demand, total phosphorus and total nitrogen; The data integration unit integrates the collected flow data Qe, pressure data Pe, temperature data Te, vibration data Ve, pollutant concentration Ce and corrosion rate Re to obtain an original data set S.

3. The artificial intelligence-based sewage pipe network alignment design system according to claim 1, characterized in that: The data preprocessing and cleaning module includes a data cleaning unit and a data normalization unit; The data cleaning unit performs anomaly detection, missing value processing and noise filtering on the collected original data set S to obtain a data cleaning set SC; Among them, anomaly detection includes using an outlier detection algorithm to remove data that exceeds the normal value range; Dealing with missing values ​​includes using mean imputation method to fill in missing values ​​in the data; Noise filtering removes high-frequency noise from the data through low-pass filtering technology; The data normalization unit normalizes the data cleaning set SC, converts data of different dimensions into a unified range, and obtains the pipe network data set SG; The pipe network data set SG is obtained by the following formula: Wherein, SCd represents the d-th item of data in the data cleaning set SC, SCdmin represents the valley value of the d-th item of data in the data cleaning set SC, and SCdmin represents the peak value of the d-th item of data in the data cleaning set SC.

4. The artificial intelligence-based sewage pipe network alignment design system according to claim 1, characterized in that: The machine learning fault diagnosis module includes a data feature extraction and fusion unit and a fault diagnosis and prediction unit; The data feature extraction and fusion unit extracts features from the pipe network data set SG, including the flow change rate ΔQe, the pollutant concentration change rate ΔCe and the corrosion change rate ΔRe, and fuses them to form a feature set F; The flow rate change rate ΔQe is obtained by the following formula: ΔQe=Qe(t)-Qe(t-1); In the formula, Qe(t) represents the flow data at time t, and Qe(t-1) represents the flow data at time t-1; The pollutant concentration change rate ΔCe is obtained by the following formula: ΔCe=Ce(t)-Ce(t-1); In the formula, Ce(t) represents the pollutant concentration at time t, and Ce(t-1) represents the pollutant concentration at time t-1; The corrosion change rate ΔRe is obtained by the following formula: ΔRe=Re(t)-Re(t-1); Where Re(t) represents the corrosion rate at time t, and Re(t-1) represents the corrosion rate at time t-1.

5. The artificial intelligence-based sewage pipe network alignment design system according to claim 4, characterized in that: The fault diagnosis and prediction unit analyzes the feature set F through the support vector machine model and builds a sewage pipe network fault diagnosis model in combination with the pipe network data set SG; Firstly, the training set is constructed using the pipe network dataset SG, feature set F and manually annotated fault cases, including samples in normal state and fault state, to train the sewage pipe network fault diagnosis model; Among them, the sample represents a set of data in the feature set F and the corresponding fault type label, including blockage fault, leakage fault and crack fault, and obtains the fault label set Y; The fault label set Y is obtained by the following formula: Y={y1,y2,...,y i }; In the formula, y i Represents the label of the i-th sample; The feature set F is trained through the sewage pipe network fault diagnosis model to learn the normal mode and abnormal mode of the pipe network, and obtain the classification prediction result f(F) of the feature set F; The classification prediction result f(F) is obtained by the following formula: f(F)=ω T *F+b; In the formula, ω represents the weight vector of the support vector machine, b represents the bias term, and ω T Represents the result after transposing the weight vector ω; By using the trained sewage pipe network fault diagnosis model to perform fault detection on the classification prediction result f(F), the fault type yP predicted by the model is calculated; The fault type yP predicted by the model is obtained by the following formula: yP = argmax(f(F)); In the formula, argmax represents the maximum value in the aggregate function, and f(F) represents the classification prediction result; The fault diagnosis model of sewage pipe network is used to locate the location of fault type yP and obtain the potential fault point PP. The potential failure point PP is obtained by the following formula: In the formula, p represents the set of all fault point locations, argmin represents the minimum value in the set function, Fj represents the feature of the jth data point in the feature set F, Fp represents the predicted fault point location feature, and m represents the total number of data points.

6. The artificial intelligence-based sewage pipe network alignment design system according to claim 5, characterized in that: The fault location and analysis module includes a spatial location unit and a fault severity assessment unit; The spatial positioning unit combines the geographic spatial information of the pipe network with the sensor data through the GIS system to spatially locate the potential fault point PP; Each potential fault point PP corresponds to a spatial coordinate in the pipe network. By comparing the GIS system with the sensor data, the actual geographical location D(PP) = xPP(x, y) is obtained; Where xPP represents the spatial coordinates of the potential fault point, and x and y represent the location in the geographic coordinate system; By analyzing the topological structure of the pipe network, determine the connection relationship between the potential fault point PP and other components of the pipe network including valves and pump stations; The formula for analyzing the topological relationship of the pipe network is as follows: T(PP)={connected(PP,C1),connected(PP,C2),...}; Wherein, T(PP) represents the topological relationship of the potential fault point PP; connected(PP, C1) represents the pipeline network node directly connected to the potential fault point PP, and C1 represents the pipeline network node.

7. The artificial intelligence-based sewage pipe network alignment design system according to claim 6, characterized in that: The fault severity assessment unit assesses the health status of the pipeline through the corrosion rate Re, pipeline material coefficient ML and pipeline age PA, obtains the severity assessment result S(PP), and performs normalization processing; The severity assessment result S(PP) is obtained by the following formula: In the formula, Remax represents the maximum value of corrosion rate, PAmax represents the maximum value of pipeline life, α1, α2 and α3 represent the preset weight values ​​of corrosion rate Re, pipeline life PA and pipeline material coefficient ML respectively; The fault impact is evaluated based on the flow data Qe and pollutant concentration Ce in the pipeline to obtain the fault impact index I(PP); The fault impact index I(PP) is obtained by the following formula: By comprehensively considering the pipeline severity assessment result S(PP) and the fault impact index I(PP), the repair priority score Pf(PP) is calculated, and the repair priority of the potential fault points PP is sorted to obtain the priority data sorting set Pfix(SO); The repair priority score Pf(PP) is obtained by the following formula: Where Dis(PP) represents the distance between the potential fault point PP and the trunk pipeline; All the potential fault points PP and the corresponding repair priority scores Pf(PP) are stored in the priority data set Pfix, Pfix = {(PP, Pfix(PP))|PP∈p}; In the formula, p represents the set of all fault point locations; According to the repair priority score Pf(PP) of each potential fault point PP, the priority data set Pfix is ​​sorted from high to low to obtain the priority data sorting set Pfix(SO)=sort(Pfix, key=Pf(PP), reverse=True); In the formula, sort means sorting from high to low.

8. The artificial intelligence-based sewage pipe network alignment design system according to claim 7, characterized in that: The emergency response and restoration strategy includes the restoration plan generation unit and the restoration evaluation and result scoring unit; The repair plan generation unit generates a repair plan RPlan(PP) according to the severity assessment result S(PP) of the potential fault point PP and the repair priority score Pf(PP), including the repair method, required resources, repair time Time and repair cost Cos; Repair methods include cleaning the pipe, replacing the pipe, and injecting sealants; Resources include maintenance personnel, tools, and materials; The repair solution RPlan (PP) is obtained by the following formula: RPlan(PP)=g(S(PP),Pf(PP),yP)); In the formula, g() represents the repair solution generation function, yP represents the fault type; The repair evaluation and result scoring unit calculates and obtains the repair result score Pre(PP) according to the repair time Tim and the repair cost Cos in the repair plan RPlan(PP); The repair result score Pre(PP) is obtained by the following formula:

9. The artificial intelligence-based sewage pipe network alignment design system according to claim 8, characterized in that: The feedback optimization and continuous learning module includes the feedback data analysis unit and the model and solution optimization unit; The feedback data analysis unit compares the obtained repair result score Pre(PP) with the actual repair score Ptu(PP) to obtain the repair deviation ΔP(PP); The repair deviation ΔP(PP) is obtained by the following formula: ΔP(PP)=Pre(PP)-Ptu(PP); The model and solution optimization unit integrates the obtained repair deviation ΔP (PP), feature set F, pipe network data set SG, manually annotated fault cases and fault point location set p as a new training set to retrain the sewage pipe network fault diagnosis model; The repair plan RPlan(PP) is modified by the repair deviation ΔP(PP) to obtain a new repair plan nRPlan(PP); The new repair solution nRPlan(PP) is obtained by the following formula: nRPlan(PP)=g(S(PP),Pf(PP),yP,ΔP(PP))).

10. A sewage pipe network alignment design method based on artificial intelligence, applied to a sewage pipe network alignment design system based on artificial intelligence according to any one of claims 1 to 9, characterized in that: The following steps are involved: Step 1: The data acquisition module collects real-time data of the pipeline network through sensors deployed inside the pipeline network nodes to obtain the original data set S; Step 2: The data preprocessing and cleaning module cleans and normalizes the collected original data set S to obtain the pipe network data set SG; Step 3: The machine learning fault diagnosis module builds a sewage pipe network fault diagnosis model based on the acquired pipe network data set SG, and learns the normal operation mode and abnormal operation mode of the pipe network to obtain potential fault points PP; Step 4: The fault location and analysis module spatially locates the potential fault point PP obtained through the GIS system and the pipe network topology structure, determines the severity and repair priority of the fault, and obtains the severity assessment result S(PP) and the repair priority score Pf(PP); Step 5: The emergency response and repair strategy generation module starts the preset emergency response plan to generate a repair plan based on the severity assessment result S(PP) and the repair priority score Pf(PP), and obtains the repair result score Pre(PP); Step 6: The feedback optimization and continuous learning module provides feedback and adjustments to the sewage network fault diagnosis model and repair plan based on the repair result score Pre (PP).

Images