Replaceable slope drainage pipe head with self-sensing function and monitoring method

Abstract

This invention discloses a replaceable slope drainage pipe head with self-sensing function and a monitoring method. The replaceable slope drainage pipe head with self-sensing function consists of a drainage pipe head structure and a cloud management platform. The drainage pipe head structure includes a permanently buried part and a replaceable insertion part. The replaceable insertion part integrates a flow sensor, a pressure sensor, a turbidity sensor, and a micro-vibration device. The cloud management platform embeds a pre-trained neural network model to perform fusion analysis on real-time collected flow, pressure, and turbidity data, and outputs in parallel siltation degree indicators, siltation state classifications, and control command parameters. When a mild siltation is determined, the micro-vibration device is automatically activated for active dredging. When a severe siltation is determined, a replacement warning is generated. This invention achieves real-time sensing and predictive maintenance of the drainage pipe status, can quantitatively diagnose the degree of siltation and actively intervene, and significantly reduces operation and maintenance costs through modular replaceable design, ensuring long-term slope stability.

Description

Technical Field

[0001] This invention relates to the field of slope engineering drainage technology, and in particular to a replaceable slope drainage pipe head with self-sensing function and a monitoring method. Background Technology

[0002] Groundwater activity is one of the core factors affecting slope stability. Increased pore water pressure significantly reduces the shear strength of soil and rock, thereby inducing geological disasters such as landslides and collapses. Therefore, constructing and maintaining an efficient slope drainage system to promptly drain groundwater and reduce seepage pressure is one of the fundamental engineering measures to ensure the long-term stability of slopes. As a key component of this system, the long-term reliable operation of slope drainage pipes is crucial.

[0003] However, traditional slope drainage pipes generally face a serious risk of clogging and failure in actual engineering projects. Groundwater seepage continuously carries silt, fine particulate matter, microorganisms and their metabolic products, as well as chemical deposits. These substances continuously deposit and adhere to the inside of the drainage pipe, especially on the pipe walls and filter layers in the inlet section, causing the cross-sectional area to continuously shrink, the water flow resistance to increase exponentially, and the drainage efficiency to decline exponentially, which may eventually lead to the complete loss of drainage function. Once the drainage function fails, water pressure inside the slope will re-accumulate, directly threatening the overall stability of the project and risking the loss of the huge safety investment and project benefits in the early stage.

[0004] Currently, the operation and maintenance management of drainage pipe blockages remains in a crude mode of passive inspection, experience-based judgment, and post-event handling, exhibiting several interconnected technical shortcomings: 1) Severely delayed status perception: Operation and maintenance work heavily relies on periodic manual inspections and passive investigations, failing to achieve early real-time detection of blockage occurrence and development. Even when investigations are conducted, they often require equipment such as pipe endoscopes, which are cumbersome and costly, making it difficult to implement routine online monitoring of large-scale drainage pipe networks; 2) Lack of quantitative basis for blockage diagnosis: Existing technologies mostly rely on the final phenomenon of a significant decrease or complete interruption of flow to determine the blockage status, failing to assess the evolution process, severity, and water quality of the blockage. 3) The system lacks the ability to accurately quantify and classify micro-changes in its properties, resulting in a lack of effective data support for subsequent maintenance decisions; 4) The system lacks the ability to intervene initially: when drainage efficiency declines, the existing drainage pipes themselves cannot take any proactive measures to curb the development of blockage, and can only wait for it to deteriorate to the point where manual intervention is necessary, thus completely missing the best window of opportunity to deal with the blockage at the initial stage with the lowest cost; 5) The maintenance methods are inefficient and uneconomical: for completely blocked pipes, the usual remedial methods are destructive and costly, such as overall excavation and replacement or high-pressure water jet dredging, which are typical reactive maintenance methods that cannot achieve precise maintenance on demand, resulting in high operation and maintenance costs throughout the entire life cycle.

[0005] In summary, traditional slope drainage pipes have systemic shortcomings in real-time status perception, accurate siltation diagnosis, proactive intervention for anomalies, and intelligent maintenance decision-making. Engineering practice urgently needs an intelligent, modular, and innovative solution that integrates self-sensing of status, self-diagnosis of siltation, self-treatment of minor siltation, and intelligent early warning of replacement. This solution aims to fundamentally shift the operation and maintenance model from passively responding to faults to proactive predictive maintenance, thereby significantly optimizing the economic efficiency of slope engineering throughout its entire lifecycle while enhancing its safety and resilience. Summary of the Invention

[0006] To achieve the above-mentioned objectives, this invention provides a replaceable slope drainage pipe head with self-sensing function, which is composed of two main parts: a drainage pipe head structure and a cloud management platform. The drainage pipe head structure has a detachable connection function and a data acquisition function. The cloud management platform, as the back-end hub, is responsible for data aggregation, intelligent analysis, decision generation, and the issuance of operation and maintenance instructions. The drainage pipe head structure and the cloud management platform are connected via a wireless communication link to jointly achieve real-time monitoring and predictive maintenance of the drainage operation status.

[0007] The drainage pipe head structure includes a permanently buried part and a replaceable insertion part; the permanently buried part is buried in the slope body at one time during construction, and includes an outer pipe, a fixed base, a cone top cover and an inner pipe fixing section; The outer pipe is pre-embedded in the drainage hole of the slope, serving as a duct framework permanently fixed in the slope body, and also playing the role of the first coarse filter. The fixed base is fixed to the port of the outer tube that is exposed on the slope. It has internal threads machined on its inner side and regularly distributed drainage holes on its side wall for draining water and as a mechanical interface for connecting the replaceable insertion part. The cone-shaped top cover is fixed to the port of the outer pipe that is deeply buried in the slope. Its shape is bullet-shaped, which facilitates the advancement during construction and installation and prevents the soil and rock from directly intruding into the pipe. The inner tube fixing section is inserted and fixed inside the outer tube, and one end is connected to the top cover of the cone. The replaceable insertion section is an independent detachable pipe section, including a replaceable inner pipe section, an inner pipe filter geotextile, a quick-connect pipe joint, a monitoring module, and a micro-vibration device; The two ends of the replaceable section of the inner tube are respectively connected to the fixed section of the inner tube and the fixed base; The inner tube filter geotextile is wrapped around the outside of the replaceable section and the fixed section of the inner tube, and is used to filter fine particles of mud and sand. The quick-connect pipe fitting is located at the connection between the replaceable section of the inner pipe and the fixed section of the inner pipe, and is used to achieve quick sealing connection and separation; the quick-connect pipe fitting adopts a conical sleeve tightening mechanism and adds an O-ring rubber sealing ring to the pipe end mating surface; The monitoring module is integrated into the replaceable section of the inner pipe and has data acquisition and transmission functions. It includes at least one flow sensor, one pressure sensor, and one turbidity sensor. The flow sensor monitors the instantaneous flow rate through the pipe in real time. C 1 The pressure sensor monitors the fluid pressure flowing through the pipe in real time. C 2 The turbidity sensor is used to monitor the turbidity of the discharged water in real time. C 3 The monitoring module has built-in power supply and wireless transmission functions, which can vectorize the collected real-time operating status data. C =( C 1 , C 2 , C 3 )Continuously send to the cloud management platform; The micro-vibration device is a miniature vibration motor encapsulated in a waterproof housing. It is directly fixed to the outer wall of the replaceable section of the inner tube by stainless steel clamps. After receiving the control command issued by the cloud management platform, it starts and generates mechanical vibration. The cloud management platform includes a data storage module, a data preprocessing module, and an intelligent decision-making module. The data storage module is responsible for receiving and storing data from various monitoring modules. The data preprocessing module preprocesses the data acquired by the monitoring modules to obtain a standardized real-time working status data matrix. Specifically, within a sliding time window... T Internal interception A The group consists of real-time operational status data vectors uploaded by the monitoring module. C This forms a real-time working status data matrix with dimensions [A,3], where the three features correspond to... C 1 , C 2 and C 3 Each real-time operating status data matrix undergoes data cleaning and standardization preprocessing to eliminate dimensional differences and improve model training stability. The intelligent decision-making module embeds a pre-trained neural network model, which can analyze the real-time operating status data matrix and output clogging degree indicators, clogging status classifications, and control command parameters.

[0008] Preferably, the intelligent decision-making module embeds a pre-trained neural network model, which is a deep learning-based multi-task learning model consisting of a feature extraction network and a multi-task output head. It can infer in parallel and output stagnation degree indicators, stagnation state classifications, and corresponding control command parameters. The input of the model is a standardized real-time working status data matrix. The feature extraction network architecture includes a one-dimensional convolutional neural network, a long short-term memory network, and a feature fusion layer. The one-dimensional convolutional neural network layer is responsible for automatically extracting local temporal features and spatial correlations from the real-time working state data matrix. It is configured to contain two layers of one-dimensional convolutional operations, each layer using 64 convolutional kernels with a width of 3, and using modified linear units as activation functions. The long short-term memory network layer receives the local temporal features extracted by the one-dimensional convolutional neural network layer, captures and remembers the forward and backward long-term dependencies of the data in the time dimension, and understands the temporal dynamic process of the occurrence and development of blockage. The feature fusion layer is a fully connected layer connected after the long short-term memory network layer, used to fuse and compress the learned high-level features to form a feature vector containing the essential information of blockage. The multi-task output head receives feature vectors from the feature fusion layer and has three parallel fully connected output heads: a clogging severity regression head, a clogging state classification head, and a control command generation head. The clogging severity regression head, implemented by a fully connected layer, outputs a continuous scalar value between 0 and 1, which represents the clogging severity index, characterizing the current severity of clogging in the pipeline; a larger value indicates more severe clogging. The clogging state classification head, consisting of a fully connected layer connected to a Softmax activation function, calculates the probability distribution of four preset clogging state categories: "normal," "mild clogging," "severe clogging," and "abnormal turbidity," and uses the category with the highest probability as the final clogging state classification result. The control command generation head is a multi-output regression head that is activated only when the clogging state classification head determines "mild clogging," and outputs the control command parameters as a two-dimensional vector. D =( D 1 , D 2 ),in D 1 The recommended vibration frequency for the micro-vibration device. D 2 The recommended continuous working time for the micro-vibration device; if the sludge status classification head determines the result as "normal", "severe sludge", or "abnormal turbidity", the output of the control command generation head will be ignored and a preset fixed command will be used instead, which corresponds to "no action", "generate replacement warning information", or "generate abnormal turbidity warning information" respectively.

[0009] Preferably, the neural network model is trained using end-to-end supervised learning. The training dataset is obtained through a dedicated slope drainage pipe physical simulation test platform, which can simulate different slope seepage conditions and reproduce the entire process from "normal" to "slightly clogged" to "severely clogged," as well as the "turbidity anomaly" process, by controllably injecting silt water with different concentrations and particle sizes. During the experiment, real-time working status data vectors are synchronously collected through a monitoring module. C =( C 1 , C 2 , C 3 For each time period, the data samples collected are labeled with corresponding reference values ​​for the degree of sludge, sludge status category labels, and control command parameters for a micro-vibration device that has been experimentally verified as effective for clearing "mild sludge". D =( D 1 , D 2 ),in D 1 Recommended vibration frequency for micro-vibration devices D 2 To determine the recommended continuous operating time for the micro-vibration device, a high-quality labeled training dataset was constructed. The overall loss function of the neural network model was defined as the weighted sum of the regression loss of the clogging degree index, the cross-entropy loss of the clogging state classification, and the regression loss of the control command parameters. During the training process, the model parameters were jointly optimized using the backpropagation algorithm, enabling the model to learn a complete and accurate mapping relationship from the original multi-source sensor data to the final operation and maintenance decision. The trained model was finally deployed in the intelligent decision-making module of the cloud management platform.

[0010] A monitoring method, applied to the replaceable slope drainage pipe head with self-sensing function, includes the following steps: S1. Real-time data acquisition: The monitoring module integrated into the replaceable insertion section automatically collects the instantaneous flow rate through the pipe body according to the preset sampling frequency. C 1 Fluid pressure C 2 and the turbidity of the discharged water C 3 This constitutes a real-time working status data vector. C =( C 1 , C 2 , C 3The monitoring module continuously sends the collected time-series data to the data storage module of the cloud management platform through its built-in wireless transmission unit. S2. Data Preprocessing: The data preprocessing module of the cloud management platform operates within a preset sliding time window. T Inside, continuous interception A The group consists of real-time operational status data vectors uploaded by the monitoring module. C This forms a real-time operational status data matrix with dimensions [A,3]. The matrix is ​​then preprocessed, including the following steps: S201, Missing value handling: For individual data missing due to sensor momentary failure or transmission packet loss, linear interpolation is used to fill in the missing data. S202, Outlier Handling: Calculate the outlier value within the sliding time window. T mean of all data points m and standard deviation s According to statistics 3 s The principle is to identify and remove outlier data points that clearly do not conform to physical meaning or measurement logic; S203, Standardization Processing: For each data value X Perform Z-score normalization transformation, and the transformed data values X norm =( X - m ) / s This transforms the data into a distribution with a mean of 0 and a standard deviation of 1. S3, Cloud-based Intelligent Diagnosis and Decision-making: The pre-processed real-time working status data matrix is ​​input into a pre-trained neural network model for real-time analysis. The model performs parallel reasoning and outputs clogging degree indicators, clogging status classifications, and control command parameters. S4. Tiered Response Execution: Based on the sludge classification results generated in step S3, the cloud management platform executes a tiered response strategy. For sludge classification as "normal," the system continues monitoring without triggering any operations. For sludge classification as "mild sludge," control command parameters are sent to the corresponding micro-vibration device via wireless communication. The micro-vibration device is activated according to the control command parameters, generating mechanical vibration to loosen the sludge on the pipe wall. After vibration ends, the system returns to step S1, collects new data after vibration, and performs a new round of diagnosis, thus forming a closed loop of monitoring, unblocking, and effect verification. For sludge classification as "severe sludge," the intelligent decision module automatically generates a replacement warning message, which includes the identification code of the sludge-affected drainage pipe head equipment, its installation location, and the degree of sludge. This replacement warning message is immediately pushed to the monitoring interface of the cloud management platform and the bound mobile terminal of the maintenance personnel, providing a basis for initiating precise maintenance. For sludge classification as "abnormal turbidity," the intelligent decision module automatically generates a turbidity abnormality warning message, which includes the equipment identification code, its installation location, and the turbidity value. S5. On-site Maintenance and Component Replacement: When the blockage status is classified as "severe blockage" and an early warning is issued, maintenance personnel arrive at the site with a spare replaceable insert based on the replacement warning information. By operating the quick-connect pipe fitting, the severely blocked old replaceable insert is quickly disassembled, and a new replaceable insert integrating a brand-new or maintained monitoring module is installed. The disassembled old parts are taken back for thorough cleaning, inspection, and replacement of necessary components. After repair, they are reused as spare parts. After replacement, the monitoring module in the new part immediately starts working and uploads data. After receiving the new data, the cloud management platform completes system reset and then automatically starts a new monitoring cycle.

[0011] In summary, compared with the prior art, the present invention has the following beneficial effects: 1) Transformation of the operation and maintenance mode of slope drainage pipes: Traditional methods rely on periodic manual inspections and post-event handling, which have serious perception lag and decision blind spots. However, this invention integrates real-time sensing, wireless transmission, cloud intelligent analysis and field execution terminals to build a complete closed-loop system for monitoring, decision and execution. This system can continuously perceive the working status of drainage pipes online and automatically execute hierarchical response strategies based on data analysis results, thereby completely transforming the operation and maintenance mode from passively responding to faults to proactive predictive maintenance, achieving the goal of intervention in the early stage of siltation. 2) Intelligent clogging diagnosis and decision-making capabilities: Existing technologies are difficult to quantitatively assess the degree of clogging, and usually rely on the final phenomenon of a significant decrease in flow rate for fuzzy judgment. However, this invention deploys multi-source sensors for flow rate, pressure and turbidity, and uses a pre-trained neural network model to fuse and analyze time-series data. It can output continuous quantitative indicators of clogging degree and clear state classification in parallel. This data-driven diagnosis method frees operation and maintenance decisions from dependence on personal experience and provides objective, accurate and traceable scientific basis, laying a solid foundation for implementing on-demand maintenance. 3) Preliminary active anti-clogging intervention function of drainage system: Traditional drainage pipes are completely passive when their performance deteriorates and cannot make any autonomous adjustments. The innovation of this invention is that a micro-vibration device is integrated into the pipe head and linked with the cloud intelligent decision module. When the system diagnoses mild clogging, it can automatically send a control command containing the optimal vibration frequency and working time to the designated pipe head, triggering the micro-vibration device to generate mechanical vibration to loosen the deposits on the pipe wall. This active unblocking attempt in the early stage of clogging can curb the development of clogging with extremely low energy consumption cost, extend the effective service life of components, and avoid the deterioration of severe clogging in most cases. 4) The modular and replaceable design significantly reduces operation and maintenance costs: In view of the actual situation that drainage pipe blockage is mainly concentrated on the slope pipe opening section, the present invention creatively designs the pipe body as a permanent buried part and a replaceable insertion part. When severe blockage occurs, only the integrated replaceable insertion part needs to be quickly replaced, without the need to excavate or replace the entire pipe buried underground. This design greatly reduces the amount of maintenance work, material consumption and damage to the slope structure, making on-site maintenance fast and economical. At the same time, the replaced parts can be recycled after cleaning and maintenance, further improving the economy and sustainability of the system. 5) Enhanced the long-term safety margin of slope engineering: By ensuring that the drainage network is under real-time monitoring and in efficient operation, the pore water pressure inside the slope can be continuously and reliably reduced, thereby fundamentally protecting the shear strength of the soil and rock mass; the system's early warning capability provides sufficient response time to deal with potential risks, curbing safety hazards in their infancy, and comprehensively enhancing the resilience and reliability of major slope engineering in the face of geological disasters. Attached Figure Description

[0012] Figure 1 This is a module connection diagram of a replaceable slope drainage pipe head with self-sensing function, as shown in an embodiment of the present invention. Figure 2 This is a schematic diagram of the drain pipe head structure shown in an embodiment of the present invention; Figure 3 This is a flowchart illustrating a monitoring method according to an embodiment of the present invention; Attached diagram labels: 1-Drainage pipe head structure, 11-Permanently buried part, 111-Outer pipe, 112-Fixed base, 113-Conical top cover, 114-Inner pipe fixed section, 12-Replaceable insertion part, 121-Replaceable inner pipe section, 122-Inner pipe filter geotextile, 123-Quick-connect pipe joint, 124-Monitoring module, 125-Micro-vibration device, 2-Cloud management platform, 21-Data storage module, 22-Data preprocessing module, 23-Intelligent decision-making module. Detailed Implementation

[0013] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments given herein are for illustration and explanation only and are not intended to limit the present invention.

[0014] This application discloses, as follows: Figure 1-2 The replaceable slope drainage pipe head with self-sensing function shown is composed of two main parts: drainage pipe head structure 1 and cloud management platform 2. The drainage pipe head structure 1 serves as the hardware and field terminal, responsible for data collection and initial command execution, and has a detachable connection function. The cloud management platform 2 serves as the back-end hub, responsible for data aggregation, intelligent analysis, decision generation, and the issuance of operation and maintenance instructions. The drainage pipe head structure 1 and the cloud management platform 2 are connected through a wireless communication link to jointly realize real-time monitoring and predictive maintenance of the drainage operation status.

[0015] The drainage pipe head structure 1 includes a permanently buried part 11 and a replaceable insertion part 12; the permanently buried part 11 is buried in the slope body at one time during construction, and includes an outer pipe 111, a fixed base 112, a cone top cover 113 and an inner pipe fixed section 114.

[0016] The outer pipe 111 is pre-embedded in the drainage hole of the slope, serving as a duct framework permanently fixed in the slope body, and also playing the role of the first coarse filter.

[0017] The fixed base 112 is fixed to the port of the outer tube that is exposed on the slope. It has internal threads machined on its inner side and regularly distributed drainage holes on its side wall for draining water and serving as a mechanical interface for connecting the replaceable insertion part 12.

[0018] The cone-shaped top cover 113 is fixed to the port of the outer pipe 111 that is deeply buried in the slope. Its shape is bullet-shaped, which facilitates the advancement during construction and installation, and can effectively prevent the soil and rock from directly intruding into the pipe.

[0019] The inner tube fixing section 114 is inserted and fixed inside the outer tube, and one end is connected to the cone top cover 113.

[0020] The replaceable insertion section 12 is an independent detachable pipe section, including an inner pipe replaceable section 121, an inner pipe filter geotextile 122, a quick-connect pipe joint 123, a monitoring module 124, and a micro-vibration device 125.

[0021] The two ends of the replaceable inner pipe section 121 are connected to the fixed inner pipe section 114 and the fixed base 112, respectively. In practice, the pipe body of the replaceable inner pipe section 121 is made of polyvinyl chloride and can be pulled out for cleaning or replacement periodically. According to engineering practice, the blockage of slope drainage pipes is mainly concentrated in the pipe opening section near the slope. Therefore, this design sets the length of the replaceable inner pipe section to 1m, specifically for this blockage-prone section, while the pipe inside the slope remains intact. This design can significantly reduce material costs while ensuring targeted maintenance.

[0022] The inner tube filter geotextile 122 is wrapped around the outside of the inner tube replaceable section 121 and the inner tube fixed section 114, serving as a fine filter layer to effectively prevent fine particles of mud and sand from entering the pipe.

[0023] The quick-connect pipe joint 123 is located at the connection between the replaceable section 121 of the inner pipe and the fixed section 114 of the inner pipe, and is used to achieve quick sealing connection and separation. The quick-connect pipe joint 123 adopts a conical sleeve tightening mechanism and adds an O-ring rubber sealing ring on the pipe end mating surface. When the locking nut is tightened, under the action of axial force, both the compression seal of the two pipe end faces to the O-ring is achieved, and the radial contraction of the inner conical sleeve clamps the pipe wall, thereby forming a double seal of axial end face and radial clamping, ensuring reliable sealing under slope drainage working pressure and micro-vibration environment. The O-ring sealing ring is a replaceable standard part. When reinstalling after maintenance disassembly, the original sealing performance can be restored by replacing the new O-ring.

[0024] The monitoring module 124 is integrated into the replaceable inner pipe section 121 and has data acquisition and transmission functions. It includes at least one flow sensor, one pressure sensor, and one turbidity sensor. The flow sensor monitors the instantaneous flow rate through the pipe in real time. C 1 The pressure sensor monitors the fluid pressure flowing through the pipe in real time. C 2 The turbidity sensor is used to monitor the turbidity of the discharged water in real time. C 3 The monitoring module 124 has built-in power supply and wireless transmission functions, and can vectorize the collected real-time operating status data. C =( C 1 , C 2 , C 3The data is continuously transmitted to the cloud management platform. In practice, the continuous working time of the monitoring module 124 is no less than the design maintenance cycle of the replaceable section of the inner tube. When the replaceable section is pulled out for maintenance or replacement, the monitoring module can be replaced as a whole or its battery can be replaced simultaneously, thus solving the power supply problem within the maintenance window. The monitoring module 124 adopts a cylindrical shell made of stainless steel, and the integrated packaging design ensures long-term reliability in pressurized and humid environments. The quantity sensor adopts an external clamp-on ultrasonic probe, fixed to the outside of the tube wall, to achieve non-invasive measurement. The pressure sensor contacts the fluid through a pressure measurement hole reserved in the tube wall, and the channel is sealed with a flexible diaphragm to transmit pressure. The turbidity sensor performs optical measurement through a transparent observation window in the tube wall. The observation window is made of sapphire glass and is fused and sealed with the outer shell. The monitoring module 124 has a built-in high-capacity lithium thionyl chloride battery, which has enough power to support continuous operation for more than one design maintenance cycle. The battery is placed in an independent sealed compartment and connected to the circuit through contact electrodes. When the replaceable insert 12 is pulled out for maintenance, the monitoring module 124 can be replaced as a whole or its battery can be replaced through a dedicated interface. The wireless transmission of the monitoring module 124 preferentially adopts low-power wide-area network communication technology. Since the replaceable insert 12 is ultimately located on the slope surface via a fixed base, its wireless signal transmitter is actually located at or near the ground surface after installation.

[0025] The micro-vibration device 125 is a miniature vibration motor encapsulated in a waterproof housing. It is directly fixed to the outer wall of the replaceable inner pipe section 121 by stainless steel clamps. After receiving the control command issued by the cloud management platform, it starts and generates mechanical vibration in the frequency range of 10Hz to 200Hz, which aims to loosen the silt attached to the pipe wall and achieve active unblocking of mild blockages.

[0026] The cloud management platform 2 includes a data storage module 21, a data preprocessing module 22, and an intelligent decision-making module 23. The data storage module 21 is responsible for receiving and storing data from various monitoring modules 124. The data preprocessing module 22 preprocesses the data acquired by the monitoring modules 124 to obtain a standardized real-time working status data matrix. Specifically, within a sliding time window... T Internal interception A The group consists of real-time operational status data vectors uploaded by the monitoring module. C This forms a real-time working status data matrix with dimensions [A,3], where the three features correspond to... C 1 , C 2 and C 3Each real-time working status data matrix undergoes data cleaning and standardization preprocessing to eliminate dimensional differences and improve model training stability. The intelligent decision-making module 23 embeds a pre-trained neural network model, which can analyze the real-time working status data matrix and output sludge degree indicators, sludge state classifications, and control commands.

[0027] In specific implementation, the data storage module 21 receives the real-time working status data vector uploaded by the monitoring module 124. C Then, a timestamp and a unique device identifier are added and stored. The data preprocessing module 22 queries the latest data from the data storage module 21 at preset intervals, performs sliding window truncation and standardization processing, and generates a real-time operating status data matrix. Subsequently, this matrix is ​​called and input into the neural network model within the intelligent decision-making module 23. After the model completes its forward computation, the intelligent decision-making module 23 analyzes its output clogging status classification and parameters: if the status is "mild clogging," the module automatically generates a data entry containing the target device identifier and vibration frequency. D 1 With working hours D 2 Control commands are sent to the micro-vibration device 125 at the corresponding drainage pipe head via the IoT communication link; if the status is "severe blockage", a replacement warning message is generated and pushed to the platform monitoring interface and maintenance terminal. All diagnostic records, issued commands and system status are written back to the data storage module 21 to form a traceable maintenance log.

[0028] In specific implementation, the intelligent decision-making module 23 embeds a pre-trained neural network model, which is a deep learning-based multi-task learning model consisting of a feature extraction network and a multi-task output head. It can infer in parallel and output the degree of stagnation, the stagnation status classification, and the corresponding control command parameters. The input of the model is a standardized real-time working status data matrix.

[0029] The feature extraction network architecture includes a one-dimensional convolutional neural network, a long short-term memory network, and a feature fusion layer. The one-dimensional convolutional neural network layer is responsible for automatically extracting local temporal features and spatial correlations from the real-time working state data matrix. It is configured to contain two layers of one-dimensional convolutional operations, each layer using 64 convolutional kernels with a width of 3, and using modified linear units as activation functions. The long short-term memory network layer receives the local temporal features extracted by the one-dimensional convolutional neural network layer, captures and remembers the forward and backward long-term dependencies of the data in the time dimension, and understands the temporal dynamic process of the occurrence and development of blockage. The feature fusion layer is a fully connected layer connected after the long short-term memory network layer, used to fuse and compress the learned high-level features to form a feature vector containing the essential information of blockage.

[0030] The multi-task output head receives feature vectors from the feature fusion layer and has three parallel fully connected output heads: a clogging severity regression head, a clogging state classification head, and a control command generation head. The clogging severity regression head, implemented by a fully connected layer, outputs a continuous scalar value between 0 and 1, which represents the clogging severity index, characterizing the current severity of clogging in the pipeline; a larger value indicates more severe clogging. The clogging state classification head, consisting of a fully connected layer connected to a Softmax activation function, calculates the probability distribution of four preset clogging state categories: "normal," "mild clogging," "severe clogging," and "abnormal turbidity," and uses the category with the highest probability as the final clogging state classification result. The control command generation head is a multi-output regression head that is activated only when the clogging state classification head determines "mild clogging," and outputs the control command parameters as a two-dimensional vector. D =( D 1 , D 2 ),in D 1 The recommended vibration frequency for the micro-vibration device. D 2 The recommended continuous working time for the micro-vibration device; if the sludge status classification head determines the result as "normal", "severe sludge", or "abnormal turbidity", the output of the control command generation head will be ignored and a preset fixed command will be used instead, which corresponds to "no action", "generate replacement warning information", or "generate abnormal turbidity warning information" respectively.

[0031] In practice, the neural network model is trained using end-to-end supervised learning. The training dataset is obtained through a dedicated slope drainage pipe physical simulation test platform. This platform can simulate different slope seepage conditions and reproduce the entire process from "normal" to "slightly clogged" to "severely clogged," as well as the "abnormal turbidity" process, by controllably injecting silt water with different concentrations and particle sizes. During the experiment, real-time working status data vectors are synchronously collected through a monitoring module. C =( C 1 , C 2 , C 3 Based on preset, quantifiable physical criteria for determining blockage, such as flow attenuation rate or the mass of sediment in the pipe, the experimenters labeled the data samples collected at each time period with corresponding reference values ​​for the degree of blockage, blockage status category labels, and control command parameters for a micro-vibration device that has been experimentally verified as effective for clearing "mild blockages". D =( D 1 ,D 2 ),in D 1 Recommended vibration frequency for micro-vibration devices D 2 To determine the recommended continuous operating time for the micro-vibration device, a high-quality labeled training dataset was constructed. The overall loss function of the neural network model was defined as the weighted sum of the regression loss of the clogging degree index, the cross-entropy loss of the clogging state classification, and the regression loss of the control command parameters. During the training process, the model parameters were jointly optimized through the backpropagation algorithm, enabling the model to learn a complete and accurate mapping relationship from the original multi-source sensor data to the final operation and maintenance decision. The trained model was finally deployed in the intelligent decision-making module 23 of the cloud management platform 2.

[0032] The second aspect of the present invention discloses as follows Figure 3 The monitoring method shown, applied to the replaceable slope drainage pipe head with self-sensing function, includes the following steps: S1. Real-time data acquisition: The monitoring module integrated into the replaceable insertion section automatically collects the instantaneous flow rate through the pipe body according to the preset sampling frequency. C 1 Fluid pressure C 2 and the turbidity of the discharged water C 3 This constitutes a real-time working status data vector. C =( C 1 , C 2 , C 3 The monitoring module continuously sends the collected time-series data to the data storage module of the cloud management platform through its built-in wireless transmission unit.

[0033] S2. Data Preprocessing: The data preprocessing module of the cloud management platform operates within a preset sliding time window. T Inside, continuous interception A The group consists of real-time operational status data vectors uploaded by the monitoring module. C This forms a real-time operational status data matrix with dimensions [A,3]. The matrix is ​​then preprocessed, including the following steps: S201, Missing value handling: For individual data missing due to sensor momentary failure or transmission packet loss, linear interpolation is used to fill in the missing data. S202, Outlier Handling: Calculate the outlier value within the sliding time window. T mean of all data points m and standard deviation sAccording to statistics 3 s The principle is to identify and remove outlier data points that clearly do not conform to physical meaning or measurement logic; S203, Standardization Processing: For each data value X Perform Z-score normalization transformation, and the transformed data values X norm =( X - m ) / s This transforms the data into a distribution with a mean of 0 and a standard deviation of 1.

[0034] S3, Cloud-based Intelligent Diagnosis and Decision-making: The pre-processed real-time working status data matrix is ​​input into a pre-trained neural network model for real-time analysis. The model performs parallel reasoning and outputs clogging degree indicators, clogging status classifications, and control command parameters.

[0035] S4. Tiered Response Execution: Based on the siltation status classification results generated in step S3, the cloud management platform executes a tiered response strategy. For siltation status classified as "normal," the system continues monitoring without triggering any operation. For siltation status classified as "mild siltation," control command parameters are sent to the corresponding micro-vibration device via wireless communication. The micro-vibration device is activated according to the control command parameters, generating mechanical vibration to loosen the silt on the pipe wall. After vibration ends, the system returns to step S1, collects new data after vibration, and performs a new round of diagnosis, thus forming a closed loop of monitoring, unblocking, and effect verification. For siltation status classified as "severe siltation," the intelligent decision module automatically generates a replacement warning message. The replacement warning message includes the identification code of the silted drainage pipe head equipment, its installation location, and the degree of siltation. This replacement warning message is immediately pushed to the monitoring interface of the cloud management platform and the bound mobile terminal of the maintenance personnel, providing a basis for initiating precise maintenance. For siltation status classified as "abnormal turbidity," the intelligent decision module automatically generates a turbidity abnormality warning message. The turbidity abnormality warning message includes the equipment identification code, its installation location, and the turbidity value.

[0036] S5. On-site Maintenance and Component Replacement: When the blockage status is classified as "severe blockage" and an early warning is issued, maintenance personnel arrive at the site with a spare replaceable insert based on the replacement warning information. By operating the quick-connect pipe fitting, the severely blocked old replaceable insert is quickly disassembled, and a new replaceable insert integrating a brand-new or maintained monitoring module is installed. The disassembled old parts are taken back for thorough cleaning, inspection, and replacement of necessary components. After repair, they are reused as spare parts. After replacement, the monitoring module in the new part immediately starts working and uploads data. After receiving the new data, the cloud management platform completes system reset and then automatically starts a new monitoring cycle.

[0037] The above describes one or more embodiments of the present invention in a relatively specific and detailed manner, but it 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 replaceable slope drainage pipe head with self-sensing function, characterized in that, It includes a drainage pipe head structure and a cloud management platform; the drainage pipe head structure includes a permanently buried part and a replaceable insertion part; the permanently buried part is buried in the slope body at one time during construction, and includes an outer pipe, a fixed base, a cone top cover and an inner pipe fixed section; The outer pipe is pre-embedded in the drainage hole of the slope, serving as a duct framework permanently fixed in the slope body, and also playing the role of the first coarse filter. The fixed base is fixed to the port of the outer pipe that is exposed on the slope, and its inner side is machined with internal threads, and its side wall is provided with regularly distributed drainage holes. The cone-shaped top cover is fixed to the port of the outer pipe that is deeply buried in the slope body, and its shape is bullet-shaped. The inner tube fixing section is inserted and fixed inside the outer tube, and one end is connected to the top cover of the cone. The replaceable insertion section is an independent detachable pipe section, including a replaceable inner pipe section, an inner pipe filter geotextile, a quick-connect pipe joint, a monitoring module, and a micro-vibration device; The two ends of the replaceable section of the inner tube are respectively connected to the fixed section of the inner tube and the fixed base; The inner tube filter geotextile is wrapped around the outside of the replaceable section and the fixed section of the inner tube, and is used to filter fine particles of mud and sand. The quick-connect pipe fitting is located at the connection between the replaceable section of the inner pipe and the fixed section of the inner pipe, and is used to achieve quick sealing connection and separation. The monitoring module is integrated into the replaceable section of the inner pipe and has data acquisition and transmission functions. It includes at least one flow sensor, one pressure sensor, and one turbidity sensor. The flow sensor monitors the instantaneous flow rate through the pipe in real time. C 1 The pressure sensor monitors the fluid pressure flowing through the pipe in real time. C 2 The turbidity sensor is used to monitor the turbidity of the discharged water in real time. C 3 ; The monitoring module has built-in power supply and wireless transmission functions, which can vectorize the collected real-time operating status data. C =( C 1 , C 2 , C 3 )Continuously send to the cloud management platform; The micro-vibration device is a miniature vibration motor encapsulated in a waterproof housing. It is directly fixed to the outer wall of the replaceable section of the inner tube by stainless steel clamps. After receiving the control command parameters sent by the cloud management platform, it starts and generates mechanical vibration. The cloud management platform includes a data storage module, a data preprocessing module, and an intelligent decision-making module. The data storage module is responsible for receiving and storing data from the monitoring module. The data preprocessing module preprocesses the data acquired by the monitoring module to obtain a standardized real-time operating status data matrix. The intelligent decision-making module embeds a pre-trained neural network model, which can analyze the real-time operating status data matrix and output siltation degree indicators, siltation status classifications, and control command parameters.

2. The replaceable slope drainage pipe head with self-sensing function according to claim 1, characterized in that, The pre-trained neural network model consists of a feature extraction network and a multi-task output head, which can infer in parallel and output clogging degree indicators, clogging status classifications and corresponding control command parameters; the input of the neural network model is a standardized real-time working status data matrix. The feature extraction network architecture includes a one-dimensional convolutional neural network (CNN), a long short-term memory (LSTM) network, and a feature fusion layer. The CNN layer is responsible for automatically extracting local temporal features and spatial correlations from the real-time working state data matrix. The CNN layer consists of two consecutive CNN layers, each using 64 convolutional kernels of size 3 and rectified linear units as the activation function. The LSTM layer receives the local temporal features extracted by the CNN layer, captures and remembers the forward and backward long-term dependencies of the data in the time dimension, and understands the temporal dynamic process of clogging occurrence and development. The feature fusion layer is a fully connected layer connected after the LSTM layer, used to fuse and compress the learned high-level features to form a feature vector containing essential information about clogging. The multi-task output head receives feature vectors from the feature fusion layer and has three parallel fully connected output heads: a clogging severity regression head, a clogging state classification head, and a control command generation head. The clogging severity regression head, implemented by a fully connected layer, outputs a continuous scalar value between 0 and 1, which represents the clogging severity index, characterizing the current severity of clogging in the pipeline; a larger value indicates more severe clogging. The clogging state classification head, consisting of a fully connected layer connected to a Softmax activation function, calculates the probability distribution of four preset clogging state categories: "normal," "mild clogging," "severe clogging," and "abnormal turbidity," and uses the category with the highest probability as the final clogging state classification result. The control command generation head is a multi-output regression head that is activated only when the clogging state classification head determines "mild clogging," and outputs the control command parameters as a two-dimensional vector. D =( D 1 , D 2 ),in D 1 The recommended vibration frequency for the micro-vibration device. D 2 The recommended continuous working time for the micro-vibration device; if the sludge status classification head determines the result as "normal", "severe sludge", or "abnormal turbidity", the output of the control command generation head will be ignored and a preset fixed command will be used instead, which corresponds to "no action" or "generate replacement warning information" or "generate abnormal turbidity warning information" respectively.

3. The replaceable slope drainage pipe head with self-sensing function according to claim 1, characterized in that, The neural network model was trained using end-to-end supervised learning. The training dataset was obtained through a dedicated slope drainage pipe physical simulation test platform. This platform can simulate different slope seepage conditions and reproduce the entire process from "normal" to "slightly clogged" to "severely clogged," as well as the "turbidity anomaly" process, by controllably injecting silt water with different concentrations and particle sizes. During the experiment, real-time working status data vectors were synchronously collected through a monitoring module. C =( C 1 , C 2 , C 3 For each time period, the data samples collected are labeled with corresponding reference values ​​for the degree of sludge, sludge status category labels, and control command parameters for a micro-vibration device that has been experimentally verified as effective for clearing "mild sludge". D =( D 1 , D 2 ),in D 1 Recommended vibration frequency for micro-vibration devices D 2 To determine the recommended continuous operating time for the micro-vibration device, a high-quality labeled training dataset was constructed. The overall loss function of the neural network model is defined as the weighted sum of the regression loss of the clogging degree index, the cross-entropy loss of the clogging state classification, and the regression loss of the control command parameters. During the training process, the model parameters are jointly optimized through the backpropagation algorithm, enabling the model to learn a complete and accurate mapping relationship from the original multi-source sensor data to the final operation and maintenance decision. The trained model is finally deployed in the intelligent decision-making module of the cloud management platform.

4. A monitoring method, characterized in that, The replaceable slope drainage pipe head with self-sensing function as described in any one of claims 1-3 includes the following steps: S1. Real-time data acquisition: The monitoring module integrated into the replaceable insertion section automatically collects the instantaneous flow rate through the pipe body according to the preset sampling frequency. C 1 Fluid pressure C 2 and the turbidity of the discharged water C 3 This constitutes a real-time working status data vector. C =( C 1 , C 2 , C 3 The monitoring module continuously sends the collected time-series data to the data storage module of the cloud management platform through its built-in wireless transmission unit. S2. Data Preprocessing: The data preprocessing module of the cloud management platform operates within a preset sliding time window. T Inside, continuous interception A The group consists of real-time operational status data vectors uploaded by the monitoring module. C This forms a real-time operational status data matrix with dimensions [A,3]. The matrix is ​​then preprocessed, including the following steps: S201. Missing value handling: For individual missing data, linear interpolation is used to fill in the missing values. S202, Outlier Handling: Calculate the outlier value within the sliding time window. T mean of all data points μ and standard deviation σ According to statistics 3 σ The principle is to identify and remove outlier data points; S203, Standardization Processing: For each data value X Perform Z-score normalization transformation, and the transformed data values X norm =( X - μ ) / σ ; S3, Cloud-based Intelligent Diagnosis and Decision-making: The pre-processed real-time working status data matrix is ​​input into a pre-trained neural network model for real-time analysis. The model performs parallel reasoning and outputs clogging degree indicators, clogging status classifications, and control command parameters. S4. Tiered Response Execution: Based on the sludge classification results generated in step S3, the cloud management platform executes a tiered response strategy. For sludge classification as "normal," the system continues monitoring without triggering any operation. For sludge classification as "mild sludge," control command parameters are sent to the corresponding micro-vibration device via wireless communication. The micro-vibration device starts according to the control command parameters, generating mechanical vibration to loosen the sludge on the pipe wall. After vibration ends, the system returns to step S1, collects new data after vibration, and performs a new round of diagnosis, thus forming a closed loop of monitoring, unblocking, and effect verification. For sludge classification as "severe sludge," the intelligent decision module automatically generates a replacement warning message, which includes the equipment identification code of the sludge-affected device, the installation geographical location, and the degree of sludge. For sludge classification as "abnormal turbidity," the intelligent decision module automatically generates a turbidity abnormality warning message, which includes the equipment identification code, the installation geographical location, and the turbidity value. S5. On-site Maintenance and Component Replacement: When the blockage status is classified as "severe blockage" and a replacement warning is issued, maintenance personnel arrive at the site with a spare replaceable insert based on the replacement warning information. By operating the quick-connect pipe fitting, the severely blocked old replaceable insert is quickly disassembled, and a new replaceable insert integrating a brand-new or maintained monitoring module is installed. The disassembled old parts are taken back for thorough cleaning, inspection, and replacement of necessary components. After repair, they are reused as spare parts. After replacement, the monitoring module in the new part immediately starts working and uploads data. The cloud management platform resets after receiving the new data and then starts a new monitoring cycle.

Images