Robot for pipeline leakage maintenance and control method applied by robot

Abstract

The invention discloses a robot for pipeline leakage maintenance and a control method applied by the robot, and relates to the technical field of intelligent robots, and the robot for pipeline leakage maintenance comprises the steps that a robot control mechanism obtains pipeline detection information and maintenance task requirements, and plans an advancing path and an operation mode of the maintenance robot; predicting a first predicted moving speed and a second predicted operation speed of the maintenance robot in the operation maintenance area according to the advancing path and the operation mode; generating a control parameter table for adjusting the operation efficiency of the maintenance robot; a pipeline environment reference range representing a pipeline environment change threshold value is obtained, and a preset robot operation control model is optimized according to the pipeline environment reference range and the control parameter table; controlling a maintenance robot to execute pipeline maintenance operation based on the robot operation control model; the maintenance robot operation control model adjusts the operation state of the maintenance robot based on the real-time pipeline environment data; the pipeline leakage maintenance efficiency is improved.

Description

Technical Field

[0001] The present application relates to the technical field of intelligent robots, and particularly relates to a robot for pipeline drip leakage repair and a control method for its application. Background Technique

[0002] In the operation of cities and industrial production, the pipeline system undertakes the important task of transporting fluid media (such as water, oil, gas, etc.). However, due to various reasons such as pipeline aging, corrosion, and external force damage, pipeline drip leakage occurs from time to time. For example, existing sewage pipe galleries are usually relatively long and are formed by connecting multiple pipes end to end through connectors. During long-term operation, drip leakage may occur at the connection position between two pipes. Due to the long length of the sewage pipe gallery, the number of pipes is large and relatively dense, making it difficult to inspect and repair.

[0003] Most of the existing inspection and repair methods are through manual inspection and repair, relying on the inspection and experience judgment of staff. The staff needs to regularly inspect the pipelines. After discovering the drip leakage point, manual repair is carried out. The manual repair method is time-consuming and laborious and difficult to detect hidden or tiny drip leakage points. Therefore, there is a defect that the existing pipeline drip leakage repair efficiency is relatively low and needs to be improved. Summary of the Invention

[0004] In order to improve the pipeline drip leakage repair efficiency, the present application provides a robot for pipeline drip leakage repair and a control method for its application.

[0005] In a first aspect, the invention object of the present application is achieved by adopting the following technical solutions: A robot for pipeline drip leakage repair includes a repair robot and a robot control mechanism. The robot control mechanism acquires pipeline detection information and repair task requirements, and plans the traveling path and operation mode of the repair robot according to the pipeline detection information and the repair task requirements. According to the traveling path and operation mode, predict the first predicted moving speed and the second predicted operation speed of the repair robot in the operation and repair area. According to the repair task requirements, the first predicted moving speed, and the second predicted operation speed, generate a control parameter table for adjusting the operation efficiency of the repair robot. Acquire a pipeline environment reference range representing the pipeline environment change threshold, and optimize a preset robot operation control model according to the pipeline environment reference range and the control parameter table. Based on the robot operation control model, control the repair robot to perform pipeline repair operations; acquire real-time pipeline environment data, and input the real-time pipeline environment data into the repair robot operation control model to adjust the operation state of the repair robot.

[0006] By adopting the above technical solutions, an intelligent pipeline repair robot with high repair efficiency and capable of completing repair tasks with high quality is provided. Specifically, by obtaining pipeline detection information (including but not limited to pipeline material, diameter, curvature, and pipeline detection images) and repair task requirements (including but not limited to repair type, tools required for repair, and repair point location), the robot control mechanism can intelligently plan the travel path and operation mode of the repair robot, which is beneficial to improving the repair efficiency. At the same time, by predicting the first predicted movement speed and the second predicted operation speed of the repair robot in the sewage pipe gallery, the working structure of the repair robot is adjusted according to the actual situation, and a control parameter table for adjusting the working rhythm of the robot and the operation efficiency of the repair robot is generated, so that the repair robot can complete high-quality repair tasks within a suitable time. By obtaining real-time pipeline environment data and making timely adjustment responses, adjustments can be quickly made in case of unexpected situations to ensure the smooth progress of the repair work. Different from the prior art in which manual inspections are used to discover repair points and then repairs are carried out, this application significantly improves the pipeline leak repair efficiency and reliability through the prediction of the movement speed of the repair robot, the optimization of operation efficiency, strong environmental adaptability, and a real-time dynamic response mechanism.

[0007] In a preferred example of this application: obtaining the pipeline detection information and repair task requirements, and planning the travel path and operation mode of the repair robot according to the pipeline detection information and the repair task requirements specifically includes: Obtaining pipeline detection information, where the pipeline detection information includes pipeline material, pipeline diameter, pipeline curvature, and the location of obstacles inside the pipeline, and the repair task requirements include repair type, repair location, and required tools; Planning the travel path of the repair robot according to the pipeline diameter, the pipeline curvature, and the location of obstacles inside the pipeline; determining the operation mode of the repair robot according to the repair type, the repair location, and the required tools.

[0008] By adopting the above technical solutions, this application integrates pipeline detection information (including pipeline material, diameter, curvature, and obstacle location) with repair task requirements (such as repair type, location, and required tools) to accurately plan the travel path and operation mode of the repair robot. Different from the problem of inaccurate robot operation caused by the lack of precise path planning in the prior art, the repair robot of this application realizes the improvement of repair efficiency and success rate.

[0009] In a preferred example of this application: predicting the first predicted movement speed and the second predicted operation speed of the repair robot in the pipeline according to the travel path and operation mode specifically includes: Predict the first predicted moving speed of the maintenance robot inside the pipeline according to the pipeline diameter, the pipeline curvature, and the traveling path; Predict the second predicted operation speed of the maintenance robot at the operation point according to the maintenance type, the required tools, and the operation mode.

[0010] By adopting the above technical solution, the problem of improper time management caused by the inability to accurately estimate the working speed of the maintenance robot in the past is effectively solved, and thus the time arrangement is optimized and the working efficiency of the maintenance robot is improved; the present application predicts the first predicted moving speed of the maintenance robot based on the pipeline diameter, curvature, and traveling path, and combines the maintenance type, required tools, and operation mode to predict the second predicted operation speed. Therefore, the action process can pre-evaluate the performance of the robot in different environments, so as to estimate the working time consumption of the maintenance robot and perform high-efficiency maintenance robot maintenance time management.

[0011] In a preferred example of the present application: generate a control parameter table for adjusting the operation efficiency of the maintenance robot according to the maintenance task requirements, the first predicted moving speed, and the second predicted operation speed, specifically including: Calculate the total time consumption of the maintenance task according to the maintenance position, the first predicted moving speed, and the second predicted operation speed; Obtain the target operation efficiency of the maintenance robot according to the total time consumption and the maintenance task requirements; Adjust the moving speed and operation speed parameters of the maintenance robot according to the target operation efficiency, the first predicted moving speed, and the second predicted operation speed, and generate a control parameter table for adjusting the operation efficiency of the maintenance robot.

[0012] By adopting the above technical solution, in order to flexibly adjust the working strategy of the robot according to the actual situation, the present application adopts a method of calculating the total time consumption according to the maintenance position, the first predicted moving speed, and the second predicted operation speed, and adjusts the moving speed and operation speed parameters of the maintenance robot based on this to generate a control parameter table, effectively solving the problem of resource waste or time delay caused by fixed parameter settings, and thus achieving the maximization of the operation efficiency of the maintenance robot.

[0013] In the second aspect, the invention object of the present application is achieved by adopting the following technical solution: A control method for a robot applied to pipeline drip repair, applied to a robot for pipeline drip repair as described above, the control method includes: Obtain pipeline detection information and maintenance task requirements; Plan the traveling path and operation mode of the maintenance robot according to the pipeline detection information and the maintenance task requirements; Predict the first predicted moving speed and the second predicted operation speed of the maintenance robot in the operation and maintenance area according to the planned travel path and operation mode; Generate a control parameter table for adjusting the operation efficiency of the maintenance robot according to the maintenance task requirements, the first predicted moving speed and the second predicted operation speed; Obtain a pipeline environment reference range representing the pipeline environment change threshold, and optimize the preset robot operation control model according to the pipeline environment reference range and the control parameter table; Control the maintenance robot to perform pipeline maintenance operations based on the optimized robot operation control model; During the execution of the maintenance operation, obtain real-time pipeline environment data; input the real-time pipeline environment data into the robot operation control model to dynamically adjust the operation state of the maintenance robot.

[0014] By adopting the above technical solutions, the present application integrates the pipeline detection information based on the actual pipeline to be repaired and the method of maintenance task requirements, real-time plans the travel path and operation mode of the maintenance robot, and predicts the first predicted moving speed and the second predicted operation speed of the maintenance robot in the operation area, so the action process can accurately set the best working parameters for the robot.

[0015] In a preferred example of the present application: the pipeline detection information includes low-clarity damaged pipeline images. After obtaining the pipeline detection information, the method further includes: Determine the forward repair relationship between multiple segments of the first low-clarity damaged pipeline images and the first intact state pipeline model, and determine the reverse verification relationship between the first intact state pipeline model and multiple segments of the first repaired pipeline images; Construct a maintenance robot operation control model based on the image recognition network corresponding to the forward repair relationship and the mechanical operation model corresponding to the reverse verification relationship, wherein the image recognition network includes a repair accuracy loss function, and the machine operation model includes an operation accuracy loss function; Obtain a training data set for which the forward repair relationship and the reverse verification relationship have been determined; train the maintenance robot operation control model based on the training data set, wherein the repair accuracy loss function and the operation accuracy loss function are jointly optimized during training; Input multiple segments of the second low-clarity damaged pipeline images to be repaired into the trained maintenance robot operation control model to obtain repaired high-clarity pipeline images.

[0016] By adopting the above technical solution, the forward repair relationship and reverse verification relationship from multiple segments of the first low-definition damaged pipeline images to the first intact state pipeline model are determined, and based on this, a maintenance robot operation control model including a repair accuracy loss function and an operation accuracy loss function is constructed to accurately simulate and optimize the maintenance process, thereby improving the repair quality of the pipeline.

[0017] In a preferred example of this application: training the maintenance robot operation control model based on the training data set includes: Based on the training data set, determine the training high-definition repaired pipeline images, the training low-definition damaged pipeline images, and the training intact state pipeline model; wherein, the training low-definition damaged pipeline images and the training intact state pipeline model satisfy the forward repair relationship, and the training intact state pipeline model and the training high-definition repaired pipeline images satisfy the reverse verification relationship; Input multiple segments of the training low-definition damaged pipeline images into the maintenance robot operation control model. The image recognition network outputs a test recovery model based on the multiple segments of the training low-definition damaged pipeline images, and the mechanical operation model outputs multiple segments of the test repaired pipeline images based on the test repair model; Based on the training intact state pipeline model and the test repair model, determine the repair accuracy loss function; Based on the training high-definition repaired pipeline images and the test repaired pipeline images, determine the operation accuracy loss function; Based on the repair accuracy loss function and the operation accuracy loss function, determine the total maintenance loss function; Optimize the total maintenance loss function to train the maintenance robot operation control model until the training of the maintenance robot operation control model is completed.

[0018] By adopting the above technical solution, the model generalization ability of the maintenance robot operation control model and the problem-solving ability to handle complex problems are improved. This application adopts a method of determining the training high-definition repaired pipeline images, the training low-definition damaged pipeline images, and the training intact state pipeline model based on the training data set, and training the maintenance robot operation control model by jointly optimizing the repair accuracy loss function and the operation accuracy loss function, so that the maintenance robot operation control model has good adaptability and accuracy to various pipeline damage situations.

[0019] In a preferred example of this application: the mechanical operation model outputs multiple segments of the test repaired pipeline images based on the test repair model, including: Input the test repair model into the mechanical operation model to obtain multiple images of the pipeline after the test repair, where the mechanical operation model is a mechanical model that simulates the repair process based on the robot operation path and the application parameters of the repair tool.

[0020] By adopting the above technical solution, a method of inputting the test repair model into the mechanical model constructed based on the robot operation path and the application parameters of the repair tool to generate the image of the pipeline after the test repair is used to dynamically adjust the repair strategy according to the actual damage situation of the pipeline. Furthermore, the repair robot can flexibly respond to different damage types and optimize the repair results.

[0021] In a preferred example of the present application: the step of inputting the test repair model into the mechanical operation model to obtain multiple images of the pipeline after the test repair includes: Input the damage type and location information corresponding to the training low - definition damaged pipeline image input to the image recognition network, and the test repair model output by the image recognition network into the mechanical operation model to obtain multiple images of the pipeline after the test repair, where the mechanical operation model adjusts the robot operation path according to the damage type, location information, and the application parameters of the repair tool.

[0022] By adopting the above technical solution, a method of inputting the damage type and location information corresponding to the training low - definition damaged pipeline image input to the image recognition network and the test repair model output by the image recognition network into the mechanical operation model together. Therefore, the operation process can dynamically adjust the operation path of the robot based on the specific damage situation (including damage type and location) and the application parameters of the repair tool. By accurately positioning the damaged location of the pipeline, it is beneficial to improve the operation efficiency of the repair robot.

[0023] In a preferred example of the present application: the first predicted moving speed v m and the second predicted operation speed ω j are calculated by the following formula: v m = 200 / D eq + 50×(1 - α / 1 + α) (1) where D eq is the equivalent diameter of the pipeline; α is the obstacle density index, and α = N / S r where N is the number of obstacles, and S r is the specified unit area based on the travel path; T j is the tool torque requirement of the repair tool (unit: N·m); D j is the pipeline diameter at the operation point; S j is the contact area.

[0024] By adopting the above technical solution, factors such as the equivalent diameter of the pipeline, the obstacle density index, the tool torque requirement of the maintenance tool, the pipeline diameter and the contact area at the operation point are considered to accurately calculate the speed parameters of the maintenance robot at different stages according to the actual environmental conditions and task requirements. To better predict the time required for the entire maintenance task, which is beneficial to reasonably arrange the work plan.

[0025] In summary, the present application includes at least one of the following beneficial technical effects: 1. Adopt a method that integrates pipeline detection information and maintenance task requirements, plan the traveling path and operation mode of the maintenance robot according to this information, and predict its first predicted moving speed and second predicted operation speed in the operation area, which can accurately set the best working parameters for the robot; effectively solve the problem of low maintenance efficiency caused by the lack of detailed planning in the prior art; 2. Determine the forward repair relationship and reverse verification relationship between multiple segments of first low - clarity damaged pipeline images and the first intact state pipeline model, and based on this, construct a maintenance robot operation control model including a repair accuracy loss function and an operation accuracy loss function to accurately simulate and optimize the maintenance process. Description of the Drawings

[0026] Figure 1 is a schematic diagram of the application scenario of a robot for pipeline drip repair in an embodiment of the present application; Figure 2 is a schematic sectional view of two maintenance schemes of a robot for pipeline drip repair when a pipeline drip leak fault occurs in an embodiment of the present application; Figure 3 is a schematic sectional view of two maintenance schemes of a robot for pipeline drip repair when a pipe crack fault occurs in an embodiment of the present application; Figure 4 is a schematic sectional view of the glue injection plugging scheme of a robot for pipeline drip repair when a pipeline is semi - detached in an embodiment of the present application; Figure 5 is a flowchart of a control method of a robot applied to pipeline drip repair in an embodiment of the present application; Figure 6 is another flowchart of a control method of a robot applied to pipeline drip repair in an embodiment of the present application. Detailed Embodiments

[0027] The following further describes the present application in detail with reference to the drawings.

[0028] In one embodiment, as Figure 1As shown, the present application discloses a robot for pipeline drip repair. The robot for pipeline drip repair includes a repair robot and a robot control mechanism. The robot control mechanism obtains pipeline detection information and repair task requirements, and plans the traveling path and operation mode of the repair robot according to the pipeline detection information and repair task requirements; according to the traveling path and operation mode, predicts the first predicted moving speed and the second predicted operation speed of the repair robot in the operation and repair area; generates a control parameter table for adjusting the operation efficiency of the repair robot according to the repair task requirements, the first predicted moving speed and the second predicted operation speed; obtains a pipeline environment reference range representing the pipeline environment change threshold, and optimizes a preset robot operation control model according to the pipeline environment reference range and the control parameter table; controls the repair robot to perform pipeline repair operations based on the robot operation control model; obtains real-time pipeline environment data, and inputs the real-time pipeline environment data into the repair robot operation control model to adjust the operation state of the repair robot.

[0029] In this embodiment, the repair robot includes a moving mechanism (such as a crawler or wheeled chassis), an operation tool module (such as a sealant injector, welding device, camera, etc.), a sensor module (such as lidar, infrared sensor, pressure sensor), and a communication module; the robot control mechanism consists of a central processor (CPU / GPU), a storage unit, a communication interface, and control algorithm software, and is responsible for receiving data, performing path planning, and generating control instructions.

[0030] Specifically, the robot control mechanism obtains pipeline detection information through a multi-modal sensor group (including a laser scanner, infrared thermal imaging, and industrial camera). The pipeline detection information includes pipeline material, pipeline diameter, pipeline curvature, and the position of obstacles inside the pipeline; the repair task requirements include repair type, repair position, and required tools; the robot control mechanism plans the traveling path of the repair robot according to the pipeline diameter, pipeline curvature, and the position of obstacles inside the pipeline; determines the operation mode of the repair robot according to the repair type, repair position, and required tools; the repair tools include a sealant gun, welding equipment, and grinding tools; among them, the optimization strategy of the traveling path includes the ways of preferential expansion area, safety constraint, and path smoothing. The preferential expansion area is centered on the repair point, and the expansion radius increases by 30%. The safety constraint condition is that the minimum distance from the pipeline wall is ≥ 3 mm (based on the real-time feedback of lidar, and the Sigmoid interpolation algorithm is used for path smoothing operation, and the improved A* algorithm is used for traveling path planning; the selection of the operation mode is based on a decision tree model according to the repair type, repair position, and required repair tools.

[0031] Exemplarily, the operation modes for different repair types and operation modes are as Figures 2 - 4As shown in the figure, the maintenance types include pipeline drip leakage and pipeline burst. The operation modes for pipeline drip leakage include glue injection plugging and tape winding plugging. Pipeline burst is further divided into pipe crack, pipeline semi-disconnection, and pipeline full-disconnection. For pipe crack, the operation modes can include tape winding plugging, joint glue injection, pipeline replacement with glue connection, and pipeline replacement with glue injection connection. For pipeline semi-disconnection, the operation modes include joint glue injection, pipeline replacement with glue connection, and pipeline replacement with glue injection connection. For pipeline full-disconnection, the operation modes include pipeline replacement with glue connection and pipeline replacement with glue injection connection.( Figure 2 It is a schematic sectional view of two maintenance solutions, namely glue injection plugging and tape winding plugging, when there is a pipeline drip leakage fault; Figure 3 It is a schematic sectional view of two maintenance solutions, namely glue injection plugging and tape winding plugging, when there is a pipe crack fault; Figure 4 It is a schematic sectional view of the glue injection plugging solution when there is a pipeline semi-disconnection), for other operation types and maintenance solutions, this application will not be described one by one, and the specific maintenance solutions can be increased or decreased according to actual needs.

[0032] Specifically, according to the traveling path and operation mode, the first predicted moving speed and the second predicted operation speed of the maintenance robot in the pipeline are predicted. Specifically, it includes predicting the first predicted moving speed of the maintenance robot in the pipeline according to the pipeline diameter, pipeline curvature, and traveling path; predicting the second predicted operation speed of the maintenance robot at the operation point according to the maintenance type, required tools, and operation mode. In this embodiment, the first predicted moving speed v 移动 = D 管道 / C 弯曲 × K 路径 × reference speed; the reference speed is set according to the maximum designed speed of the robot (such as the default value is 0.5 m / s); D 管道 is the pipeline diameter, C 弯曲 is the pipeline curvature, K 路径 is the traveling path. If the pipeline curvature exceeds the threshold (such as the radius of curvature < 5 meters) or C 弯曲 = 1.5, the speed is further reduced; the second predicted operation speed v 作业 = T 工具 / d 作业点 × A 接触 × reference operation speed, T 工具 is the torque requirement of the required tool, that is, the maximum torque of the tool (such as the torque of the welding gun is 15 N·m), d 作业点 is the pipeline diameter, A 接触 is the contact area between the tool and the pipeline (such as the contact area of the welding nozzle is 5 cm 2 ), and the reference operation speed is the default speed of the tool (such as the default speed of the welding gun is 0.2 m / s).

[0033] Specifically, according to the maintenance task requirements, the first predicted moving speed, and the second predicted operation speed, a control parameter table for adjusting the operation efficiency of the maintenance robot is generated. Specifically, the total time required for the maintenance task is calculated based on the maintenance location, the first predicted moving speed, and the second predicted operation speed. In this embodiment, the calculation formula for the total time required is T 总 =L 路径 / v 移动 +N 作业点 / v 作业 , where L 路径 is the total path length required for the current maintenance task; according to the total time required and the maintenance task requirements, the target operation efficiency of the maintenance robot is obtained. The target operation efficiency refers to the completion time threshold set according to the maintenance task requirements (for example, an urgent task needs to be completed within 5 minutes). The calculation formula for the target operation efficiency E 目标 is: E 目标 =T 总 / T 需求 ×100%, where T 需求 is the preset expected completion time; according to the target operation efficiency, the first predicted moving speed, and the second predicted operation speed, the moving speed and operation speed parameters of the maintenance robot are adjusted to generate a control parameter table for adjusting the operation efficiency of the maintenance robot.

[0034] Each module in the above-mentioned robot for pipeline drip leakage repair can be implemented in whole or in part by software, hardware, and their combination; each of the above modules can be embedded in the processor of the computer device in the form of hardware or be independent of it, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0035] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, only the above division of each functional unit and module is used as an example. In actual applications, the above functions can be allocated to different functional units and modules according to needs, that is, the internal structure of the above-mentioned robot for pipeline drip leakage repair is divided into different functional units or modules to complete all or part of the functions described above.

[0036] In one embodiment, as Figure 5 shown, a control method for a robot applied to pipeline drip leakage repair is provided. The control method for the robot applied to pipeline drip leakage repair is applied to the robot for pipeline drip leakage repair. The data processing method applied to the multi-functional integrated service terminal specifically includes the following steps: S1: Obtain pipeline detection information and maintenance task requirements.

[0037] S2: According to the pipeline detection information and maintenance task requirements, plan the travel path and operation mode of the maintenance robot.

[0038] S3: Predict the first predicted moving speed and the second predicted operation speed of the maintenance robot in the operation and maintenance area according to the planned travel path and operation mode.

[0039] In this embodiment, the first predicted moving speed refers to the theoretical speed at which the robot travels along the planned path, which is a dynamic value considering the pipeline geometric characteristics and mechanical transmission characteristics; the second predicted operation speed refers to the rotation speed at which the robot performs maintenance actions at the operation point, which is related to the tool characteristics and contact state; the travel path refers to the optimal movement route of the robot from the current position to the maintenance point.

[0040] Further, in this embodiment, the first predicted moving speed v m and the second predicted operation speed ω j are calculated by the following formulas: v m = 200 / D eq + 50×(1 - α / 1 + α) (1) where D eq is the equivalent diameter of the pipeline; α is the obstacle density index, α = N / S r where N is the number of obstacles, and S r is the specified unit area based on the travel path; T j is the tool torque requirement of the maintenance tool (unit: N·m); D j is the pipeline diameter at the operation point; S j is the contact area.

[0041] S4: Generate a control parameter table for adjusting the operation efficiency of the maintenance robot according to the maintenance task requirements, the first predicted moving speed, and the second predicted operation speed.

[0042] In this embodiment, the control parameter table refers to an optimized configuration table containing the adjustment of the robot speed and tool parameters.

[0043] S5: Obtain the pipeline environment reference range representing the pipeline environment change threshold, and optimize the preset robot operation control model according to the pipeline environment reference range and the control parameter table.

[0044] Specifically, the pipeline environment reference range refers to the safety threshold of the pipeline environment parameters (such as temperature ≤ 50°C, pressure ≤ 10 MPa); the robot operation control model refers to the control algorithm based on historical data and real-time input. The robot operation control model of this application refers to the intelligent control model of the LSTM neural network.

[0045] S6: Control the maintenance robot to perform pipeline maintenance operations based on the optimized robot operation control model.

[0046] S7: During the execution of the maintenance operation, obtain real-time pipeline environment data; input the real-time pipeline environment data into the robot operation control model to dynamically adjust the operation state of the maintenance robot.

[0047] In this embodiment, the real-time pipeline environment data refers to the environmental parameters continuously collected during the operation (such as temperature, pressure, and obstacle changes); for example, if it is detected that an obstacle moves in front of the path, the robot operation control model re-plans the path, reduces the moving speed, or adjusts the tool parameters, such as pausing welding to avoid the obstacle.

[0048] In one embodiment, as Figure 6 shown, the pipeline detection information includes low-clarity damaged pipeline images; after step S1, a control method for a robot applied to pipeline drip repair further includes: S11: Determine the forward repair relationship between multiple segments of the first low-clarity damaged pipeline images and the first intact state pipeline model, and determine the reverse verification relationship between the first intact state pipeline model and multiple segments of the first repaired pipeline images.

[0049] In this embodiment, the first low-clarity damaged pipeline images refer to pipeline defect images with low resolution or poor quality (such as blurred images caused by insufficient light or sensor limitations); the first intact state pipeline model is an ideal pipeline structure model generated by an algorithm, such as a pipeline geometric model without cracks or leaks.

[0050] Specifically, the forward repair relationship is a mapping function from low-clarity damaged images to the intact state pipeline model. The forward repair relationship trains an image recognition network with labeled data (such as the corresponding relationship between known damaged areas and the ideal model) to enable it to learn the mapping from low-quality images to the ideal model; the reverse verification relationship is a physical simulation process from the first intact state pipeline model to multiple segments of the first repaired pipeline images, used to verify the accuracy of the repair result; the reverse verification relationship is to construct a reverse model based on physical rules (such as material mechanics and fluid dynamics) to ensure that the generated repaired images meet the actual physical conditions.

[0051] S12: Based on the image recognition network corresponding to the forward repair relationship and the mechanical operation model corresponding to the reverse verification relationship, construct a maintenance robot operation control model, where the image recognition network includes a repair accuracy loss function, and the machine operation model includes an operation accuracy loss function.

[0052] In this embodiment, the image recognition network is a deep learning model, such as U-Net; connect the image recognition network in series with the mechanical operation model to form an end-to-end control model, that is, the maintenance robot operation control model.

[0053] Specifically, the calculation formula of the repair accuracy loss function is as follows: Among them, N corresponds to the number of image samples of the forward repair relationship, ‖I pred - I gt ‖2 is the pixel-level L2 norm, which is used to predict the pixel-level Euclidean distance between the image I pred and the real image I gt (here refers to the low-resolution damaged image); α is the loss weight coefficient, which is the balance factor between pixel loss and perceptual loss, and its value ranges from [0.3, 0.7]. Per is the feature matching loss based on a pre-trained convolutional network (such as VGG19). Specifically, extract the VGG feature maps of I pred and I gt , calculate the L2 norm difference of each layer of feature maps of the two images, and then obtain the weighted sum: that is L refers to the feature map layer, α l is the weight coefficient of the current layer, and γ1, γ2 are predefined weight coefficient values.

[0054] The calculation formula of the operation accuracy loss function is as follows: Among them, M is the number of samples of action sequences in the mechanical operation dataset that satisfy the reverse verification relationship; ‖F pred - F real ‖1 is the mechanical response L1 norm that satisfies, which is used to define the total absolute error between the predicted torque F pred and the actual torque F real . The torque considers the three-dimensional force vector of the force on the X-axis, Y-axis, and Z-axis, F pred = [F x , F y , F z T ; β is the loss weight coefficient, which is the balance factor between mechanical loss and smoothness penalty, and its value ranges from [0.5, 0.9]. For is the force smoothness penalty, which is used to define the smoothness index of the torque change between adjacent actions. The calculation formula of For is Among them, F avg is the average torque value, and T is the length of the action sequence.

[0055] S13: Obtain the training dataset for which the forward repair relationship and reverse verification relationship have been determined; based on the training dataset, train the operation control model of the maintenance robot. Among them, the repair accuracy loss function and the operation accuracy loss function are jointly optimized during training.

[0056] ​In this embodiment, the Adam optimizer is used, and the learning rate is set to 0.001 to jointly optimize the network parameters of the repair accuracy loss function and the operation accuracy loss function.

[0057] Specifically, step S13 further includes: S131: Based on the training data set, determine the training high-definition repaired pipeline image, the training low-definition damaged pipeline image, and the training intact state pipeline model; wherein, the training low-definition damaged pipeline image and the training intact state pipeline model satisfy the forward repair relationship, and the training intact state pipeline model and the training high-definition repaired pipeline image satisfy the reverse verification relationship.

[0058] S132: Input multiple segments of the training low-definition damaged pipeline image into the maintenance robot operation control model. The image recognition network outputs a test repair model based on the multiple segments of the training low-definition damaged pipeline image, and the mechanical operation model outputs multiple segments of the test repaired pipeline image based on the test repair model.

[0059] Specifically, the damage type and location information corresponding to the training low-definition damaged pipeline image input into the image recognition network, and the test repair model output by the image recognition network are input into the mechanical operation model to obtain multiple segments of the test repaired pipeline image. Among them, the mechanical operation model adjusts the robot operation path according to the damage type, location information, and application parameters of the repair tool.

[0060] In this embodiment, the mechanical operation model includes performing mechanical simulation based on finite element analysis (FEA) or a dynamics model, combined with the robot kinematic model (such as joint angles, moving speeds) and tool mechanical parameters (such as heat conduction of a welding gun, adhesion force of a sealant) to simulate the repair process, that is, according to the geometric data of the test repair model, simulating the interaction between the tool and the pipeline (such as the heat affected zone during welding), and calculating the surface topography of the repaired pipeline (such as smoothness, stress distribution). The repaired image refers to converting the simulation result into a high-clarity image, and the surface of the repaired pipeline image has no cracks and meets the requirements of relevant repair process standards.

[0061] Among them, the test repair model is input into the mechanical operation model to obtain multiple segments of pipeline images after test repair, including: identifying the damage type and location information corresponding to the damaged pipeline images with low resolution in the training of the input image recognition network, and inputting the test repair model output by the image recognition network into the mechanical operation model to obtain multiple segments of pipeline images after test repair. Among them, the mechanical operation model adjusts the robot operation path according to the damage type, location information, and application parameters of the repair tool; for example, the damage type (such as crack width, corrosion depth) and location coordinates extracted from the training low-resolution image, and the repair result output by the image recognition network (such as the three-dimensional coordinates after crack closure); inputting the damage type (such as precise alignment required for cracks), location (such as coordinates X, Y, Z), and tool parameters (such as the radius of the welding gun) into the path planning algorithm to obtain the motion trajectory of the robot end effector (such as avoiding obstacles and aligning with the repair area); through the improved A* algorithm, combined with the contact angle between the tool and the pipeline (such as the welding gun needs to be perpendicular to the crack) and the maximum load of the tool (such as the torque does not exceed 15 N·m), to adjust the path (such as increasing the number of welding layers) according to the mechanical problems found in the simulation (such as stress concentration areas).

[0062] For example, the corresponding relationship between the damage type and the tool parameters includes: Crack → Welding gun (power 3000W, speed 0.2m / s).

[0063] Corrosion → Sealant spraying (pressure 5MPa, coverage rate 100%).

[0064] S133: Based on the trained intact pipeline model and the test repair model, determine the repair accuracy loss function.

[0065] In this embodiment, the repair accuracy loss function L im = MSE(test repair model, trained intact model) + λ im ×structural similarity loss), MSE refers to the mean square error, which measures the difference in geometric structure, and the structural similarity loss is used to evaluate the shape matching degree between the prediction model and the real model (such as curvature consistency). For example: If the difference in curvature between the prediction model and the real model is 0.1 and the MSE is 0.01, then L im = 0.01 + 0.2×s0.1 = 0.03.

[0066] S134: Based on the trained high-resolution pipeline images after repair and the test pipeline images after repair, determine the operation accuracy loss function.

[0067] In this embodiment, the operation accuracy loss function L me = SSIM(test repair image, trained high-resolution image) + λ me× Physical constraint loss), SSIM refers to the Structural Similarity Index, which evaluates the matching degree of image texture and structure (in this embodiment, it is set that SSIM ≥ 0.9 is qualified), and the physical constraint loss is used to ensure that the repair result conforms to physical rules (such as the welding temperature not exceeding the melting point of the material). For example, if SSIM is 0.85 and the physical constraint loss is 0.05, then L me = 1 - 0.85 + 0.05 = 0.2.

[0068] S135: Based on the repair accuracy loss function and the operation accuracy loss function, determine the total repair loss function.

[0069] Specifically, the total repair loss function refers to the weighted loss that comprehensively considers repair accuracy and operation accuracy, and is used to jointly optimize the model parameters; the calculation formula of the total repair loss function is: L me = μ1 × L im + μ2 × L me , where the weights are set according to the task requirements. For example, μ1 = 0.7, μ2 = 0.3, and both μ1 and μ2 conform to the dynamic adjustment mechanism.

[0070] S136: Optimize the total repair loss function to train the repair robot operation control model until the training of the repair robot operation control model is completed.

[0071] S14: Input the multi-segment second-low-definition damaged pipeline images to be repaired into the trained repair robot operation control model to obtain the repaired high-definition pipeline images.

[0072] In this embodiment, the second-low-definition damaged pipeline image refers to the low-quality image of the actual pipeline to be repaired; the high-definition pipeline image refers to the repaired image generated by the model, which is used to guide the robot to perform repair operations.

[0073] Specifically, the image recognition network analyzes the low-definition image to generate an ideal pipeline model (such as the three-dimensional structure after crack closure), and the mechanical operation model generates the repaired high-definition image (such as the pipeline surface without cracks) according to the ideal model. The repaired high-definition image is transmitted to the robot control system to guide the tool to perform repair operations (such as locating the crack position and controlling the spraying trajectory of the sealant). The mechanical operation model of this application simulates the welding repair process based on Finite Element Analysis (FEA) to generate the repaired surface image.

[0074] It should be understood that the sequence numbers of the steps in the above embodiments do not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of this application.

[0075] Those skilled in the art can clearly understand that, for the convenience and conciseness of description, only the above-mentioned division of each functional unit and module is used as an example. In actual applications, the above functions can be allocated to different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above.

[0076] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A robot for repairing pipeline drip leakage, characterized in that, It includes a maintenance robot and a robot control mechanism. The robot control mechanism obtains pipeline detection information and maintenance task requirements, and plans the traveling path and operation mode of the maintenance robot according to the pipeline detection information and the maintenance task requirements. According to the traveling path and operation mode, predict the first predicted moving speed and the second predicted operation speed of the maintenance robot in the operation and maintenance area. According to the maintenance task requirements, the first predicted moving speed and the second predicted operation speed, generate a control parameter table for adjusting the operation efficiency of the maintenance robot. Obtain the pipeline environment reference range representing the pipeline environment change threshold, and optimize the preset robot operation control model according to the pipeline environment reference range and the control parameter table. Based on the robot operation control model, control the maintenance robot to perform pipeline maintenance operations; obtain real-time pipeline environment data, and input the real-time pipeline environment data into the maintenance robot operation control model to adjust the operation state of the maintenance robot.

2. The robot for pipeline drip repair according to claim 1, characterized in that, The obtaining of the pipeline detection information and the maintenance task requirements, and the planning of the traveling path and operation mode of the maintenance robot according to the pipeline detection information and the maintenance task requirements specifically include: Obtain pipeline detection information, where the pipeline detection information includes pipeline material, pipeline diameter, pipeline curvature, and the position of obstacles in the pipeline, and the maintenance task requirements include maintenance type, maintenance position, and required tools. According to the pipeline diameter, the pipeline curvature, and the position of obstacles in the pipeline, plan the traveling path of the maintenance robot; according to the maintenance type, the maintenance position, and the required tools, determine the operation mode of the maintenance robot.

3. The robot for pipeline drip repair according to claim 2, wherein, The predicting of the first predicted moving speed and the second predicted operation speed of the maintenance robot in the pipeline according to the traveling path and operation mode specifically includes: According to the pipeline diameter, the pipeline curvature, and the traveling path, predict the first predicted moving speed of the maintenance robot in the pipeline. According to the maintenance type, the required tools, and the operation mode, predict the second predicted operation speed of the maintenance robot at the operation point.

4. A robot for pipeline drip repair according to claim 1, characterized in that, The generating of a control parameter table for adjusting the operation efficiency of the maintenance robot according to the maintenance task requirements, the first predicted moving speed, and the second predicted operation speed specifically includes: According to the maintenance position, the first predicted moving speed, and the second predicted operation speed, calculate the total time required for the maintenance task. According to the total time required and the maintenance task requirements, obtain the target operation efficiency of the maintenance robot. According to the target operation efficiency, the first predicted moving speed, and the second predicted operation speed, adjust the moving speed and operation speed parameters of the maintenance robot, and generate a control parameter table for adjusting the operation efficiency of the maintenance robot.

5. A control method for a robot applied to pipeline drip repair, characterized in that, Applied to a robot for pipeline drip repair as described in any one of claims 1-4, the control method includes: Obtain pipeline detection information and maintenance task requirements. According to the pipeline detection information and the maintenance task requirements, plan the traveling path and operation mode of the maintenance robot. Predict the first predicted moving speed and the second predicted operation speed of the maintenance robot in the operation and maintenance area according to the planned travel path and operation mode; Generate a control parameter table for adjusting the operation efficiency of the maintenance robot according to the maintenance task requirements, the first predicted moving speed, and the second predicted operation speed; Obtain a pipeline environment reference range representing the pipeline environment change threshold, and optimize the preset robot operation control model according to the pipeline environment reference range and the control parameter table; Control the maintenance robot to perform pipeline maintenance operations based on the optimized robot operation control model; During the execution of the maintenance operation, obtain real-time pipeline environment data; input the real-time pipeline environment data into the robot operation control model to dynamically adjust the operation state of the maintenance robot.

6. The control method of a robot applied to pipeline drip repair according to claim 5, characterized in that The pipeline detection information includes low-clarity damaged pipeline images. After obtaining the pipeline detection information, the method further includes: determining the forward repair relationship between multiple segments of the first low-clarity damaged pipeline images and the first intact state pipeline model, and determining the reverse verification relationship between the first intact state pipeline model and multiple segments of the first repaired pipeline images; Construct a maintenance robot operation control model based on the image recognition network corresponding to the forward repair relationship and the mechanical operation model corresponding to the reverse verification relationship, where the image recognition network includes a repair accuracy loss function, and the machine operation model includes an operation accuracy loss function; Obtain a training data set for which the forward repair relationship and the reverse verification relationship have been determined; train the maintenance robot operation control model based on the training data set, where the repair accuracy loss function and the operation accuracy loss function are jointly optimized during training; Input multiple segments of the second low-clarity damaged pipeline images to be repaired into the trained maintenance robot operation control model to obtain repaired high-clarity pipeline images.

7. The control method of a robot applied to pipeline drip repair according to claim 6, characterized in that, The training the maintenance robot operation control model based on the training data set includes: Based on the training data set, determine the training high-clarity repaired pipeline images, the training low-clarity damaged pipeline images, and the training intact state pipeline model; where the training low-clarity damaged pipeline images and the training intact state pipeline model satisfy the forward repair relationship, and the training intact state pipeline model and the training high-clarity repaired pipeline images satisfy the reverse verification relationship; Input multiple segments of the training low-clarity damaged pipeline images into the maintenance robot operation control model, the image recognition network outputs a test recovery model based on multiple segments of the training low-clarity damaged pipeline images, and the mechanical operation model outputs multiple segments of the test repaired pipeline images based on the test repair model; Determine the repair accuracy loss function based on the training intact state pipeline model and the test repair model; Determine the operation accuracy loss function based on the training high-clarity repaired pipeline images and the test repaired pipeline images; Determine the total maintenance loss function based on the repair accuracy loss function and the operation accuracy loss function; Optimize the total maintenance loss function to train the maintenance robot operation control model until the training of the maintenance robot operation control model is completed.

8. The control method of a robot applied to pipeline drip repair according to claim 7, characterized in that, Based on the test repair model, the mechanical operation model outputs multiple segments of the pipeline images after test repair, including: Input the test repair model into the mechanical operation model to obtain multiple segments of the pipeline images after test repair, where the mechanical operation model is a mechanical model simulating the repair process constructed based on the robot operation path and repair tool application parameters.

9. The control method of a robot applied to pipeline drip repair according to claim 8, characterized in that, The step of inputting the test repair model into the mechanical operation model to obtain multiple segments of the pipeline images after test repair includes: Input the damage type and location information corresponding to the training low-definition damaged pipeline images input into the image recognition network, and the test repair model output by the image recognition network into the mechanical operation model to obtain multiple segments of the pipeline images after test repair, where the mechanical operation model adjusts the robot operation path according to the damage type, location information, and application parameters of the repair tool.

10. The control method of a robot applied to pipeline drip repair according to claim 6, characterized in that, The first predicted moving speed v m and the second predicted operating speed ω j are calculated by the following formula: v m = 200 / D eq + 50×(1 - α / 1 + α) (1) Among them, D eq is the equivalent diameter of the pipeline; α is the obstacle density index, α = N / S r where N is the number of obstacles, and S r is the specified unit area based on the travel path; T j is the tool torque requirement of the maintenance tool (unit: N·m); D j is the pipeline diameter at the operation point; S j is the contact area.

Images