A method and system for intelligent pipeline flushing control based on pressure changes

Abstract

This invention provides a method and system for intelligent pipeline flushing control based on pressure changes. The target pipeline is divided into several sub-pipes according to variations in length, diameter, or areas prone to blockage. Pressure sensors are installed at both ends of each sub-pipe to collect real-time pressure data. Each sub-pipe is equipped with a nozzle with a solenoid valve at one end, which opens or closes according to commands, adjusting the water pressure and flow rate. A deep learning-based graph neural network model is used, taking the pressure data from each pipe segment as input, to learn the pressure transmission relationships and blockage impact patterns between segments. The system receives pressure data collected by the pressure sensors in real-time, analyzes the trained graph neural network model, generates nozzle flushing control commands, and sends them to the corresponding nozzles. Specifically, segmented monitoring and intelligent control can accurately address pipeline blockage problems.

Description

Technical Field

[0001] This invention belongs to the field of intelligent pipeline flushing control technology, specifically relating to an intelligent pipeline flushing control method and system based on pressure changes. Background Technology

[0002] Activated carbon and other media addition devices are widely used in various industrial production, water treatment and environmental protection fields.

[0003] During long-term operation, existing media addition devices are prone to residual reagent crystals and impurities in the pipeline, which can lead to pipeline blockage and poor flow, affecting the accuracy and stability of media addition, thereby reducing production efficiency and increasing maintenance costs.

[0004] Traditional pipeline flushing methods are mostly manual, which is cumbersome and the flushing effect is difficult to guarantee, failing to meet the needs of automated production. At the same time, some automated flushing devices are complex in structure and expensive, lacking wide applicability. Summary of the Invention

[0005] Based on this, the present invention provides a method and system for intelligent pipeline flushing control based on pressure changes, which aims to achieve the function of automatic pipeline flushing by monitoring pipeline pressure and intelligently controlling electric valves, thereby solving the problems of easy clogging and inconvenient flushing in existing media adding devices, and improving the operating efficiency and stability of media adding devices.

[0006] A first aspect of this invention provides a pipeline intelligent flushing control method based on pressure changes, the method comprising:

[0007] The target pipeline is divided into several sub-pipes according to changes in length, pipe diameter, or areas prone to blockage. Pressure sensors are installed at both ends of each sub-pipe to collect real-time pressure data. A nozzle with a solenoid valve is installed at one end of each sub-pipe to open or close according to instructions and to adjust the water pressure and flow rate.

[0008] A graph neural network model based on deep learning is adopted, with pressure data of each pipeline segment as input, and the pressure transmission relationship and blockage impact law between each pipeline segment are learned through training.

[0009] It receives pressure data collected by pressure sensors in real time, calls a trained graph neural network model for analysis, generates nozzle flushing control commands, and sends them to the corresponding nozzles.

[0010] Furthermore, the step of employing a deep learning-based graph neural network model, using pressure data from each pipe segment as input, and training to learn the pressure transmission relationship and blockage impact patterns between pipe segments includes:

[0011] During the normal operation of the pipeline, pressure sensor data of each section of the pipeline is continuously collected to record pressure changes under different operating conditions, which serve as a sample of normal conditions.

[0012] Different levels of blockages were placed in different sub-pipes, and pressure data of each pipe section was collected when blockages occurred to construct an abnormal state sample.

[0013] The collected data is preprocessed, including data cleaning and normalization, to obtain preprocessed data.

[0014] Based on graph neural networks, each sub-pipe segment is regarded as a node in the graph, and the connection between pipes is regarded as an edge. The mutual influence of pressure changes in each pipe segment is learned through the information transmission between nodes.

[0015] With the goal of predicting the degree of blockage in each section of the pipeline, mean squared error is used as the loss function, and stochastic gradient descent algorithm is used to train the model to obtain a trained graph neural network model.

[0016] Furthermore, the step of receiving pressure data collected by the pressure sensor in real time, analyzing it by calling the trained graph neural network model, generating a nozzle flushing control command, and sending it to the corresponding nozzle includes:

[0017] It receives pressure data collected by pressure sensors in real time, inputs it into a trained graph neural network model, and outputs the prediction results.

[0018] Based on the prediction results, the location of the anomaly and its range of influence are determined. Based on the location of the anomaly and its range of influence, and in conjunction with the pressure changes in adjacent pipelines, the required flushing pressure and flushing time for the nozzle are determined.

[0019] Based on the flushing pressure and the flushing time, the corresponding nozzles are controlled to flush the pipeline. During the flushing process, pressure changes are continuously monitored. If the pressure returns to normal, the flushing stops; if the pressure does not return to normal, the flushing intensity is increased or adjacent pipeline nozzles are activated for auxiliary flushing.

[0020] Furthermore, in the step of receiving pressure data collected by the pressure sensor in real time, calling the trained graph neural network model for analysis, generating a nozzle flushing control command, and sending it to the corresponding nozzle, the fuzzy logic model is combined with the graph neural network model. The graph neural network model is used to predict the pipe blockage state, and the fuzzy logic is used to generate flushing control commands based on the prediction results and real-time pressure changes.

[0021] Furthermore, in the fuzzy logic model, the pressure change rate and pressure difference of each pipeline segment are used as inputs to the fuzzy logic model; the nozzle flushing pressure and flushing time are used as outputs to the fuzzy logic model; based on practical experience and expert knowledge, fuzzy rules for the fuzzy logic model are formulated; according to the fuzzy rules, fuzzy reasoning is performed on the input variables to obtain fuzzy output results, and then the fuzzy output results are transformed into specific flushing pressure and flushing time through defuzzification to guide the nozzle to perform flushing operations.

[0022] Furthermore, the graph neural network model uses a graph convolutional network as its basic architecture, which consists of 3-5 graph convolutional network layers. Each graph convolutional network layer updates its own features by aggregating information from neighboring nodes. A ReLU activation function is used after each convolutional network layer. The number of neurons in the output layer is the same as the number of pipe segments, and each neuron corresponds to the predicted value of the blockage degree of a pipe segment.

[0023] Furthermore, in the step of predicting the degree of blockage in each section of the pipeline, using mean squared error as the loss function, and employing stochastic gradient descent to train the model and obtain the trained graph neural network model, the loss function formula is:

[0024] ;

[0025] Where n is the number of samples. Labels indicating the actual degree of congestion. This represents the level of congestion predicted by the model.

[0026] A second aspect of this invention provides a pipeline intelligent flushing control system based on pressure variation, used to implement the pipeline intelligent flushing control method based on pressure variation provided in the first aspect of this invention. The system includes:

[0027] The segmentation module is used to divide the target pipeline into several sub-pipes according to changes in length, pipe diameter, or areas prone to blockage. Pressure sensors are installed at both ends of each sub-pipe to collect pressure data in real time. Each sub-pipe is equipped with a nozzle with a solenoid valve at one end to open or close according to instructions and to adjust the water pressure and flow rate.

[0028] The training module is used to learn the pressure transmission relationship and blockage impact patterns between pipe sections by using a deep learning-based graph neural network model with pressure data of each pipe section as input.

[0029] The analysis module receives pressure data collected by the pressure sensor in real time, calls the trained graph neural network model for analysis, generates nozzle flushing control commands, and sends them to the corresponding nozzles.

[0030] A third aspect of the present invention provides a computer-readable storage medium, comprising:

[0031] The readable storage medium stores one or more programs that, when executed by a processor, implement the intelligent pipeline flushing control method based on pressure variation as described in the first aspect.

[0032] A fourth aspect of the present invention provides an electronic device, the electronic device including a memory and a processor, wherein:

[0033] The memory is used to store computer programs;

[0034] When the processor executes the computer program stored in the memory, it implements the intelligent pipeline flushing control method based on pressure changes as described in the first aspect.

[0035] This invention provides a method and system for intelligent pipeline flushing control based on pressure changes. The target pipeline is divided into several sub-pipes according to variations in length, diameter, or areas prone to blockage. Pressure sensors are installed at both ends of each sub-pipe to collect real-time pressure data. Each sub-pipe has a nozzle equipped with a solenoid valve at one end to open or close according to commands, adjusting the water pressure and flow rate. A deep learning-based graph neural network model is used, taking the pressure data from each pipe segment as input, to learn the pressure transmission relationships and blockage impact patterns between segments. The system receives pressure data collected by the pressure sensors in real-time, analyzes the trained graph neural network model, generates nozzle flushing control commands, and sends them to the corresponding nozzles. Specifically, segmented monitoring and intelligent control can accurately address pipeline blockage problems. Attached Figure Description

[0036] Figure 1 A schematic diagram of the device for adding media;

[0037] Figure 2 This is a flowchart illustrating the implementation of a pipeline intelligent flushing control method based on pressure changes, as provided in Embodiment 1 of the present invention.

[0038] Figure 3 This is a structural block diagram of a pipeline intelligent flushing control system based on pressure changes, provided in Embodiment 2 of the present invention.

[0039] Figure 4 This is a structural block diagram of an electronic device provided in Embodiment 3 of the present invention.

[0040] Reference numerals in the attached diagram: 1. Medium supply tank; 2. Medium pump; 3. Pressure sensor; 4. Electric valve; 5. DI water flushing branch pipe; 6. Main pipeline; 7. Equipment requiring medium addition; 8. Controller; 9. Signal line. Detailed Implementation

[0041] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Several embodiments of the invention are illustrated in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.

[0042] It should be noted that when a component is said to be "fixed to" another component, it can be directly on the other component or there may be an intervening component. When a component is said to be "connected to" another component, it can be directly connected to the other component or there may be an intervening component. The terms "vertical," "horizontal," "left," "right," and similar expressions used in this document are for illustrative purposes only.

[0043] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0044] Taking a basic media addition device as an example, which only has a short pipeline, the device mainly consists of a media supply tank, a pressure sensor, an electric valve, a controller, a media pump, a DI water flushing branch pipe, and a main pipeline. The media supply tank stores the prepared media solution and is connected to the equipment requiring media addition via a pipeline. The pressure sensor is installed at the end of the main pipeline near the media pump outlet, monitoring pressure changes in the pipeline in real time and converting the pressure signal into an electrical signal that is transmitted to the controller. It should be noted that the flushing branch pipe is connected to the main pipeline near the media pump outlet via a tee, and its opening and closing status is controlled by the controller. One end of the flushing branch pipe is connected to the main pipeline, and the other end is connected to the flushing DI water, which is deionized water.

[0045] More specifically, the controller has preset pressure ranges for normal media addition and flushing trigger pressure thresholds. Under normal media addition conditions, when the pressure value detected by the pressure sensor is within the normal pressure range, the electric valve on the flushing branch closes. When the pressure value detected by the pressure sensor exceeds the flushing trigger pressure threshold, indicating a possible blockage in the pipeline, the controller issues a command to open the electric valve on the flushing branch. DI water then enters the main pipeline through the flushing branch to flush the main pipeline, removing impurities and blockages. After cleaning, when the pressure value detected by the pressure sensor is within the normal pressure range, the electric valve on the flushing branch closes.

[0046] Please see Figure 1 This is a schematic diagram of the media adding device. The connection relationship of each component is as follows: The media supply tank 1 is securely installed in a suitable position to ensure its safe storage. The media supply tank 1 is connected to the inlet end of the media pump 2 through the main pipe 6. The equipment 7 that needs to add media is connected to the outlet end of the media pump 2 through the main pipe 6. The pressure sensor 3 is installed on the main pipe 6 near the outlet end of the media pump 2. The electric valve 4 is installed on the DI water flushing branch 5. The DI water flushing branch 5 is connected between the outlet end of the media pump 2 and the pressure sensor 3 through a tee. The pressure sensor 3 and the electric valve 4 are connected to the controller 8 through the signal line 9.

[0047] Understandably, the above only provides a simple pipeline flushing control scheme based on pressure changes. However, for long pipelines with different specifications, the pressure in different parts of the pipeline and the pressure when the pipeline is blocked are different. Simply setting up a pressure sensor and judging the pressure value collected by the pressure sensor cannot clear the blockage in the pipeline in time and ensure the pipeline is unobstructed. Therefore, it is necessary to provide a pipeline intelligent flushing control method based on pressure changes. The specific implementation method is as follows.

[0048] Example 1

[0049] Embodiment 1 of this invention provides a pipeline intelligent flushing control method based on pressure changes. Please refer to [link / reference]. Figure 2 This is a flowchart of a pipeline intelligent flushing control method based on pressure changes, specifically including steps S01 to S03.

[0050] Step S01: Divide the target pipeline into several sections according to changes in length or diameter, or areas prone to blockage.

[0051] Each sub-pipeline is equipped with pressure sensors at both ends to collect real-time pressure data. Each sub-pipeline section has a nozzle with a solenoid valve at one end to open or close on command, and to adjust the water pressure and flow rate. It should be noted that the target pipeline can be manually divided into several sub-pipelines based on variations in length, pipe diameter, or areas prone to blockage.

[0052] Step S02: A graph neural network model based on deep learning is adopted, with the pressure data of each section of the pipeline as input, and the pressure transmission relationship and blockage effect law between each section of the pipeline are learned through training.

[0053] Specifically, during the normal operation of the pipeline, pressure sensor data from each section of the pipeline is continuously collected to record pressure changes under different operating conditions, serving as a sample of the normal state.

[0054] An artificial simulation of pipeline blockage can be conducted by setting different levels of blockage in different sub-pipelines, collecting pressure data of each section of the pipeline when blockage occurs, and constructing an abnormal state sample.

[0055] The collected data is preprocessed, including data cleaning and normalization, to obtain preprocessed data.

[0056] Based on graph neural networks, each sub-pipe segment is considered as a node in the graph, and the connections between pipes are considered as edges. By learning the mutual influence of pressure changes in each pipe segment through information transmission between nodes, it can be understood that each sub-pipe segment is abstracted as a node in the graph, and the node features are information such as real-time pressure data collected by the pressure sensor of that pipe segment and historical pressure change trends; the physical connections between pipes are defined as edges, and the weight of the edges can be set according to factors such as pipe diameter and length to reflect the pressure transmission efficiency between different pipes.

[0057] To predict the degree of blockage in each section of the pipeline, a mean squared error (MSE) loss function is used, and the stochastic gradient descent algorithm is employed for model training, resulting in a trained graph neural network (Graph Neural Network) model. It should be noted that the Graph Neural Network model uses a graph convolutional network (GCNN) as its basic architecture, with 3-5 GCNN layers. Each GCNN layer updates its own features by aggregating information from neighboring nodes. For example, the first GCNN layer receives raw node pressure data and learns local pressure change patterns through convolution operations. Subsequent layers further extract higher-level features to capture the complex relationships between different pipelines. The number of neurons in each layer can be adjusted according to the data dimension and computational resources. In this embodiment, it starts with 128 neurons and is optimized through experiments. A ReLU activation function is used after each GCNN layer, with the formula: Nonlinear factors are introduced to enhance the model's ability to express complex pressure change relationships; the number of neurons in the output layer is the same as the number of pipe segments, and each neuron corresponds to the predicted value of the blockage degree of a pipe segment. The output range is [0,1], and the closer the value is to 1, the higher the degree of blockage.

[0058] More specifically, during the training process, mean squared error is used as the loss function, and the formula for the loss function is:

[0059] ;

[0060] Where n is the number of samples. Labels indicating the actual degree of congestion. The MSE (Mean Sequence of Errors) effectively measures the average error between the model's predicted and actual values ​​to indicate the degree of congestion. Optimizing this function can improve the accuracy of model predictions.

[0061] Furthermore, the stochastic gradient descent (SGD) algorithm and its variants (such as the Adam optimizer) are selected for model training. In this embodiment of the invention, the Adam optimizer is used, with an initial learning rate of 0.001. The learning rate is dynamically adjusted based on the loss changes during training to prevent the model from getting trapped in local optima. Simultaneously, to avoid overfitting, an L2 regularization term can be added during training to constrain the model parameter sizes.

[0062] Furthermore, the dataset is divided into training, validation, and test sets in a 7:2:1 ratio. During training, a batch training method is used, with batch sizes set to 32-64. In each iteration, a batch of data is extracted from the training set to calculate the loss and update the model parameters. After each epoch (one iteration of the training set), the model performance is evaluated on the validation set. Based on the validation loss, the model parameters are adjusted or training is terminated early (early stopping) to prevent overfitting.

[0063] Step S03: Receive pressure data collected by the pressure sensor in real time, call the trained graph neural network model for analysis, generate nozzle flushing control commands, and send them to the corresponding nozzles.

[0064] It should be noted that during the flushing process, no medium is supplied. The medium supply pipeline must be cut off and the medium in the pipeline emptied before flushing. The waste liquid generated during flushing is discharged through the drain port with a controllable switch reserved at the end of the pipeline.

[0065] In this embodiment of the invention, pressure data collected by a pressure sensor is received in real time and input into a trained graph neural network model to obtain a prediction result. The pressure sensor collects pressure data for each pipe section at a fixed frequency of once per second. This frequency can be adjusted according to the complexity of the pipe's operating conditions and response speed requirements. For example, for easily clogged or high-flow-rate pipes, it can be increased to five times per second. Furthermore, a preliminary threshold for pressure changes can be set. If the pressure in a pipe section increases by more than 1.5 times compared to the average value during normal operation, a preliminary warning is triggered. After the preliminary warning, the real-time pressure data of that pipe section and adjacent pipes are input into the trained neural network model, and the model outputs a predicted value for the degree of blockage. If the predicted value exceeds 0.7 (which can be adjusted according to actual conditions), the pipe section is determined to be blocked.

[0066] Based on the prediction results, the location of the anomaly and its impact range are determined. Specifically, based on the predicted blockage levels of each pipeline segment output by the neural network model, combined with pressure sensor data, the core location of the blockage is determined. If a pipeline segment has the highest predicted value and exceeds a threshold, it is identified as a blockage point. Furthermore, with the blockage point as the center, the pressure changes of adjacent pipelines in the upstream and downstream directions are analyzed. When the pressure change rate of adjacent pipelines is lower than a preset value, the expansion stops, and the pipelines within that range are identified as the affected area.

[0067] Subsequently, based on the location of the anomaly and the range of influence, combined with the pressure changes in adjacent pipes, the required flushing pressure and flushing time for the nozzle are determined. It should be noted that this calculation is based on the degree of blockage, pipe length, and pressure conditions in adjacent pipes. The base pressure is determined according to the formula based on the degree of blockage. The calculation is performed, where P0 is the reference pressure, k is a coefficient, and S is the degree of blockage. Compensation is then applied based on the upstream and downstream pressure difference: for every 1 MPa the upstream pressure exceeds the normal value, the pressure increases by 0.2 MPa; for every 1 MPa the downstream pressure falls below the normal value, the pressure decreases by 0.1 MPa. The formula is determined based on the degree of blockage and the pipe volume. Where β is the time coefficient and V is the pipe volume. Simultaneously, adjustments are made dynamically based on pressure recovery; if the pressure drop rate is lower than expected, the flushing time is extended proportionally.

[0068] Based on the flushing pressure and the flushing time, the corresponding nozzles are controlled to flush the pipeline. During the flushing process, pressure changes are continuously monitored. If the pressure returns to normal, the flushing stops; if the pressure does not return to normal, the flushing intensity is increased or adjacent pipeline nozzles are activated for auxiliary flushing.

[0069] In other embodiments of the present invention, in the scenario of pipe blockage, the relationship between pressure change and blockage degree is not a simple linear one, and there are various interfering factors (such as fluid flow fluctuations and ambient temperature changes). In order to generate more reasonable flushing control commands, specifically, the fuzzy logic model is combined with the graph neural network model. The graph neural network model is used to predict the pipe blockage state, and the fuzzy logic is used to generate flushing control commands based on the prediction results and real-time pressure changes.

[0070] It should be noted that in the fuzzy logic model, the pressure change rate and pressure difference of each pipeline segment are used as inputs. For example, the pressure change rate can be categorized as "low change rate," "medium change rate," and "high change rate"; the pressure difference can be categorized as "small difference," "medium difference," and "large difference." The nozzle flushing pressure and flushing time are used as outputs. For example, the flushing pressure can be categorized as "low pressure," "medium pressure," and "high pressure"; the flushing time can be categorized as "short time," "medium time," and "long time." Based on practical experience and expert knowledge, the fuzzy logic model is formulated... In this embodiment of the invention, if the pressure change rate is "high" and the pressure difference is "large", the flushing pressure is "high" and the flushing time is "long"; if the pressure change rate is "low" and the pressure difference is "small", the flushing pressure is "low" and the flushing time is "short". A complete rule base is formed through summarizing and optimizing a large number of practical cases. According to the fuzzy rules, fuzzy reasoning is performed on the input variables to obtain fuzzy output results. Then, the fuzzy output results are transformed into specific flushing pressure and flushing time through defuzzification to guide the nozzle to perform flushing operations. Defuzzification can be achieved through the centroid method.

[0071] Finally, the flushing effect was evaluated by calculating the pressure recovery rate and pressure fluctuation range. The formula for calculating the pressure recovery rate is as follows: If the pressure recovery rate exceeds 90%, the flushing effect is considered good. The pressure fluctuation range refers to the pressure fluctuation within 10 minutes after flushing. If the pressure fluctuation range is within ±5% of the normal pressure, it is considered stable.

[0072] If the flushing effect does not meet expectations, the system will automatically analyze the cause, such as incomplete removal of blockages or improper flushing parameter settings, and readjust the flushing strategy, including changing the flushing mode, adjusting the pressure and time, and starting a secondary flush.

[0073] In summary, the present invention proposes a pipeline intelligent flushing control method based on pressure changes. This method divides the target pipeline into several sub-pipes according to changes in length, pipe diameter, or areas prone to blockage. Pressure sensors are installed at both ends of each sub-pipe to collect pressure data in real time. A nozzle with a solenoid valve is installed at one end of each sub-pipe to open or close according to instructions and adjust the water pressure and flow rate. A graph neural network model based on deep learning is used, with the pressure data of each pipeline segment as input, to learn the pressure transmission relationship and blockage influence law between pipeline segments through training. The pressure data collected by the pressure sensors is received in real time, and the trained graph neural network model is called for analysis to generate nozzle flushing control instructions, which are then sent to the corresponding nozzles. Specifically, segmented monitoring and intelligent control can accurately address pipeline blockage problems.

[0074] Example 2

[0075] Embodiment 2 of the present invention provides a pipeline intelligent flushing control system 200 based on pressure changes. Please refer to [link / reference]. Figure 3 The diagram shows the structure of a pressure-change-based intelligent pipeline flushing control system 200, which includes:

[0076] The segmentation module 21 is used to divide the target pipeline into several sub-pipes according to the length, pipe diameter variation or easy blockage risk area. Pressure sensors are installed at both ends of each sub-pipe to collect the pressure data in the pipeline in real time. A nozzle with a solenoid valve is installed at one end of each sub-pipe to open or close according to the command and adjust the water pressure and flow rate.

[0077] Training module 22 is used to employ a deep learning-based graph neural network model, taking pressure data from each pipe segment as input, and learning the pressure transmission relationship and blockage impact patterns between pipe segments through training. The graph neural network model uses a graph convolutional network as its basic architecture, with 3-5 layers. Each layer updates its features by aggregating information from neighboring nodes. A ReLU activation function is used after each convolutional network layer. The number of neurons in the output layer is the same as the number of pipe segments, and each neuron corresponds to a predicted blockage level for a specific pipe segment.

[0078] Analysis module 23 is used to receive pressure data collected by pressure sensors in real time, call a trained graph neural network model for analysis, generate nozzle flushing control commands, and send them to the corresponding nozzles. The graph neural network model is combined with the graph neural network model to predict pipe blockage status, while the fuzzy logic model generates flushing control commands based on the prediction results and real-time pressure changes. In the fuzzy logic model, the pressure change rate and pressure difference of each pipe segment are used as inputs; the nozzle flushing pressure and flushing time are used as outputs. Fuzzy rules for the fuzzy logic model are formulated based on practical experience and expert knowledge. Fuzzy inference is performed on the input variables according to the fuzzy rules to obtain fuzzy output results. Then, defuzzification is used to convert the fuzzy output results into specific flushing pressure and flushing time to guide the nozzles in performing flushing operations.

[0079] Furthermore, in some other embodiments of the present invention, the training module 22 includes:

[0080] The recording unit is used to continuously collect pressure sensor data from each section of the pipeline during normal operation and record pressure changes under different operating conditions as a normal state sample.

[0081] The data acquisition unit is used to collect pressure data of each section of the pipeline when blockages of different degrees are set in different sub-pipes, and to construct abnormal state samples.

[0082] The preprocessing unit is used to preprocess the collected data, including data cleaning and normalization, to obtain preprocessed data.

[0083] The definition unit is used to treat each sub-pipe as a node in the graph based on the graph neural network, and the connection relationship between pipes as an edge. The mutual influence of pressure changes in each pipe segment is learned through the information transmission between nodes.

[0084] The training unit, used to predict the degree of blockage in each section of the pipeline, employs mean squared error as the loss function and uses stochastic gradient descent to train the model, resulting in a trained graph neural network model. The loss function formula is as follows:

[0085] ;

[0086] Where n is the number of samples. Labels indicating the actual degree of congestion. This represents the level of congestion predicted by the model.

[0087] Furthermore, in some other embodiments of the present invention, the analysis module 23 includes:

[0088] The receiving unit is used to receive pressure data collected by the pressure sensor in real time, input the trained graph neural network model, and input the prediction results.

[0089] The determining unit is used to determine the location and range of influence of the anomaly based on the prediction results, and to determine the required flushing pressure and flushing time of the nozzle based on the location and range of influence of the anomaly and the pressure changes of adjacent pipelines.

[0090] The control unit is used to control the corresponding nozzle to flush the pipeline according to the flushing pressure and the flushing time. During the flushing process, the pressure change is continuously monitored. If the pressure returns to normal, the flushing is stopped. If the pressure does not return to normal, the flushing intensity is increased or the nozzle of the adjacent pipeline is activated to assist in flushing.

[0091] Example 3

[0092] Embodiment 3 of the present invention proposes an electronic device, please refer to [link / reference]. Figure 4 This is a structural block diagram of an electronic device, including a memory 20, a processor 10, and a computer program 30 stored in the memory and executable on the processor. When the processor 10 executes the computer program 30, it implements the intelligent pipeline flushing control method based on pressure changes as described above.

[0093] In some embodiments, the processor 10 may be a central processing unit (CPU), controller, microcontroller, microprocessor or other data processing chip, used to run program code stored in memory 20 or process data, such as executing access restriction programs.

[0094] The memory 20 includes at least one type of readable storage medium, such as flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 20 can be an internal storage unit of an electronic device, such as the hard disk of the electronic device. In other embodiments, the memory 20 can also be an external storage device of the electronic device, such as a plug-in hard disk, SmartMediaCard (SMC), SecureDigital (SD) card, FlashCard, etc., equipped on the electronic device. Furthermore, the memory 20 can include both internal and external storage units of the electronic device. The memory 20 can be used not only to store application software and various types of data of the electronic device, but also to temporarily store data that has been output or will be output.

[0095] This invention also proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned intelligent pipeline flushing control method based on pressure changes.

[0096] Those skilled in the art will understand that the logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can mean any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0097] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0098] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0099] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0100] The above embodiments merely illustrate several implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A method for intelligent pipeline flushing control based on pressure changes, characterized in that, The method includes: The target pipeline is divided into several sub-pipes according to changes in length, pipe diameter, or areas prone to blockage. Pressure sensors are installed at both ends of each sub-pipe to collect real-time pressure data. A nozzle with a solenoid valve is installed at one end of each sub-pipe to open or close according to instructions and to adjust the water pressure and flow rate. A graph neural network model based on deep learning is adopted, with pressure data of each pipeline segment as input, and the pressure transmission relationship and blockage impact law between each pipeline segment are learned through training. It receives pressure data collected by pressure sensors in real time, calls a trained graph neural network model for analysis, generates nozzle flushing control commands, and sends them to the corresponding nozzles; The steps of receiving pressure data collected by the pressure sensor in real time, analyzing it using a trained graph neural network model, generating nozzle flushing control commands, and sending them to the corresponding nozzles include: It receives pressure data collected by pressure sensors in real time, inputs it into a trained graph neural network model, and outputs the prediction results. Based on the prediction results, the location of the anomaly and its range of influence are determined. Based on the location of the anomaly and its range of influence, and in conjunction with the pressure changes in adjacent pipelines, the required flushing pressure and flushing time for the nozzle are determined. Based on the flushing pressure and the flushing time, the corresponding nozzles are controlled to flush the pipeline. During the flushing process, pressure changes are continuously monitored. If the pressure returns to normal, the flushing stops; if the pressure does not return to normal, the flushing intensity is increased or adjacent pipeline nozzles are activated for auxiliary flushing.

2. The intelligent pipeline flushing control method based on pressure changes according to claim 1, characterized in that, The steps of employing a deep learning-based graph neural network model, using pressure data from each pipeline segment as input, and training to learn the pressure transmission relationships and blockage impact patterns between pipeline segments include: During the normal operation of the pipeline, pressure sensor data of each section of the pipeline is continuously collected to record pressure changes under different operating conditions, which serve as a sample of normal conditions. Different levels of blockages were placed in different sub-pipes, and pressure data of each pipe section was collected when blockages occurred to construct an abnormal state sample. The collected data is preprocessed, including data cleaning and normalization, to obtain preprocessed data. Based on graph neural networks, each sub-pipe segment is regarded as a node in the graph, and the connection between pipes is regarded as an edge. The mutual influence of pressure changes in each pipe segment is learned through the information transmission between nodes. With the goal of predicting the degree of blockage in each section of the pipeline, mean squared error is used as the loss function, and stochastic gradient descent algorithm is used to train the model to obtain a trained graph neural network model.

3. The intelligent pipeline flushing control method based on pressure changes according to claim 2, characterized in that, In the step of receiving pressure data collected by the pressure sensor in real time, calling the trained graph neural network model for analysis, generating a nozzle flushing control command, and sending it to the corresponding nozzle, a fuzzy logic model is combined with the graph neural network model. The graph neural network model is used to predict the pipe blockage state, and the fuzzy logic is used to generate flushing control commands based on the prediction results and real-time pressure changes.

4. The intelligent pipeline flushing control method based on pressure changes according to claim 3, characterized in that, In the fuzzy logic model, the pressure change rate and pressure difference of each pipeline segment are used as inputs; the flushing pressure and flushing time of the nozzle are used as outputs; based on practical experience and expert knowledge, fuzzy rules for the fuzzy logic model are formulated; according to the fuzzy rules, fuzzy reasoning is performed on the input variables to obtain fuzzy output results; and then, through defuzzification, the fuzzy output results are transformed into specific flushing pressure and flushing time to guide the nozzle to perform flushing operations.

5. The intelligent pipeline flushing control method based on pressure changes according to claim 4, characterized in that, The graph neural network model uses a graph convolutional network as its basic architecture, which consists of 3-5 graph convolutional network layers. Each graph convolutional network layer updates its own features by aggregating information from neighboring nodes. A ReLU activation function is used after each convolutional network layer. The number of neurons in the output layer is the same as the number of pipeline segments, and each neuron corresponds to the predicted value of the blockage degree of a pipeline segment.

6. The intelligent pipeline flushing control method based on pressure variation according to claim 2, characterized in that, In the step of predicting the degree of blockage in each section of the pipeline, using mean squared error as the loss function, and training the model using stochastic gradient descent to obtain a trained graph neural network model, the loss function formula is as follows: ； Where n is the number of samples. Labels indicating the actual degree of congestion. This represents the level of congestion predicted by the model.

7. A pipeline intelligent flushing control system based on pressure changes, characterized in that, For implementing the intelligent pipeline flushing control method based on pressure variation as described in any one of claims 1-6, the system comprises: The segmentation module is used to divide the target pipeline into several sub-pipes according to changes in length, pipe diameter, or areas prone to blockage. Pressure sensors are installed at both ends of each sub-pipe to collect pressure data in real time. Each sub-pipe is equipped with a nozzle with a solenoid valve at one end to open or close according to instructions and to adjust the water pressure and flow rate. The training module is used to learn the pressure transmission relationship and blockage impact patterns between pipe sections by using a deep learning-based graph neural network model with pressure data of each pipe section as input. The analysis module receives pressure data collected by the pressure sensor in real time, calls the trained graph neural network model for analysis, generates nozzle flushing control commands, and sends them to the corresponding nozzles.

8. A computer-readable storage medium, characterized in that, include: The readable storage medium stores one or more programs that, when executed by a processor, implement the intelligent pipeline flushing control method based on pressure variation as described in any one of claims 1-6.

9. An electronic device, characterized in that, The electronic device includes a memory and a processor, wherein: The memory is used to store computer programs; When the processor executes the computer program stored in the memory, it implements the intelligent pipeline flushing control method based on pressure changes as described in any one of claims 1-6.

Images