A steam heating network trap adjustment method based on an embodied intelligent robot dog

Abstract

The present invention discloses a steam heat network drain valve adjustment method based on an embodied intelligent robot dog, comprising: dividing the steam heat network drain system into multiple inspection sub-areas, and providing a charging station and multiple robot dogs for each inspection sub-area; each robot dog inspects the drain system, obtains multi-source data related to the actual system drain, and transmits it to a remote management background, and simultaneously obtains power information in real time, and cooperates with the charging station to perform autonomous charging; each robot dog obtains real-time drain abnormality information by using a steam heat network drain diagnostic model issued during the inspection, and autonomously controls the robot dog to walk to the drain abnormality position, and after using a target drain valve recognition model to identify a target drain valve object to be adjusted, aligns with and opens the target drain valve by a bionic robot arm to discharge condensate, and simultaneously uses a fluid type monitoring model to monitor the drain port fluid type, so as to control the bionic robot arm to continue to open the target drain valve or close the drain valve.

Description

Technical Field

[0001] The present invention belongs to the technical field of steam heat network drain management, and specifically relates to a steam heat network drain valve adjustment method based on an embodied intelligent robot dog. Background Art

[0002] The heat network drain system is a key link in the heating network. Its function is to remove the condensed water generated in the steam heat network. If the condensed water in the heat network pipes cannot be removed in time, the accumulated condensed water will occupy the pipe space, reducing the amount of steam in the pipes and affecting the heating efficiency. In addition, the condensed water in the heat network pipes will also cause water hammer in the pipes, causing damage to the valves and elbows in the pipes, and affecting the safe and stable operation of the steam heat network.

[0003] At present, the steam heat network trap is mainly regulated by sensing the pressure signal through the sensor, transmitting the pressure signal to the controller, and the controller controls the switch of the trap and the length of operation time. However, this regulation method is more suitable for the steam trap that is an electric regulating valve, not for the manually adjusted steam trap. Moreover, during the entire operation period, it is necessary to continuously inspect the steam heat network and pay attention to the regulation of the steam trap regardless of day or night, which will consume a lot of manpower and material resources.

[0004] With the widespread application of robot dogs in various technical fields, embodied intelligence technology is considered to be the key direction of the new generation of artificial intelligence wave. It can integrate artificial intelligence into the physical entity of robot dogs, giving them the ability to perceive, learn and dynamically interact with the environment like humans, forming embodied intelligent robot dogs. By using multiple types of sensors, efficient motion control systems and robotic arms, and integrating artificial intelligence algorithms for autonomous learning and decision-making, they can replace humans in patrol inspections and monitoring, device grabbing and installation, etc., saving manpower and material resources and improving work efficiency. Therefore, how to use embodied intelligent robot dogs to patrol the steam heat network drain system and adjust the drain valve, saving manpower and material resources and improving the adjustment accuracy and efficiency of the drain valve, and ensuring the safe and stable operation of the steam heat network is an urgent problem to be solved.

[0005] Based on the above technical problems, it is necessary to design a new steam heating network trap adjustment method based on an embodied intelligent robot dog. Summary of the invention

[0006] The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide a steam heating network drain valve adjustment method based on an embodied intelligent robot dog, which can use the embodied intelligent robot dog to inspect the steam heating network drain system and adjust the drain valve, thereby ensuring the complete discharge of condensate and avoiding steam leakage, while improving the accuracy and reliability of the operation, realizing the intelligent management and decision-making of the steam heating network drain system, optimizing the performance of the steam heating network, and ensuring the safe and stable operation of the steam heating network.

[0007] In order to solve the above technical problems, the technical solution of the present invention is:

[0008] The present invention provides a steam heat network trap adjustment method based on an embodied intelligent robot dog, which comprises:

[0009] S1. Divide the steam heating network drainage system into multiple inspection sub-areas, and set up a charging station and multiple robot dogs for each inspection sub-area; each robot dog is equipped with multiple types of sensor equipment, data transmission equipment, bionic mechanical arms, inspection management system and drainage regulation system;

[0010] S2. Each robot dog uses a navigation control algorithm in each inspection sub-area to inspect the steam heat network drainage system according to the inspection path and inspection plan issued by the inspection management system. During the inspection process, multiple types of sensor equipment are used to obtain multi-source data related to the actual system drainage, and the data is transmitted to the remote management background through the data transmission equipment. At the same time, the inspection management system obtains power information in real time and cooperates with the charging station for autonomous charging;

[0011] S3. The remote management background obtains multi-source data related to the actual system draining, and after data processing and model training combined with the draining operation data collected by the system, establishes a steam heat network draining diagnosis model, a target drain valve identification model and a fluid type monitoring model, and sends them to each robot dog draining regulation system;

[0012] S4. During the inspection process, each robot dog uses the steam heat network drain diagnosis model to obtain real-time drain abnormality information, and autonomously controls the robot dog to walk to the drain abnormality position. After using the target drain valve recognition model to identify the target drain valve object to be adjusted, the bionic robot arm is used to align and open the target drain valve to discharge condensate. At the same time, the fluid type monitoring model is used to monitor the drain port fluid type to control the bionic robot arm to continue to open the target drain valve or close the drain valve.

[0013] Furthermore, in S1, the steam heating network drainage system is divided into multiple inspection sub-areas, and a charging station and multiple robot dogs are provided for each inspection sub-area, including:

[0014] The geographic information data of the steam heat network drain system is obtained by using GIS maps, and the physical entities of the system are mapped into virtual space to generate a steam heat network drain system twin. The installed pressure sensors, temperature sensors, and flow sensors are used to collect the drain system operation data in real time, drive the steam heat network drain system twin, and establish a steam heat network drain system digital twin model containing GIS maps.

[0015] The steam heat network drain system digital twin model is used to consider the steam pipeline layout, trap distribution, drain valve location, drain equipment quantity, and geographical characteristics, and the steam heat network drain system is divided into multiple inspection sub-areas.

[0016] The location of the power supply, the time cost of the robot dog returning to charge, and the path length are considered to select the location of the charging station in each inspection sub-area. The number of charging stations required in each inspection sub-area is calculated by considering the task frequency of drainage inspection, the working time of the robot dog, the number of robot dogs, and the battery life. The number of robot dogs is allocated according to the drainage regulation workload, complexity, and emergency response requirements of each inspection sub-area. The model of the robot dog is selected according to the adaptability of the drainage regulation task and geographical characteristics.

[0017] The charging stations are set up according to the location and number of charging stations in each inspection sub-area, and the robot dogs are set up according to the number and model of robot dogs in each inspection sub-area.

[0018] Further, in S2, each robot dog uses a navigation control algorithm in each inspection sub-area to inspect the steam heating network drain system according to the inspection path and inspection plan issued by the inspection management system, including:

[0019] Each inspection sub-area is recorded as , the number of steam traps corresponding to each sub-area is recorded as ;

[0020] According to the entrance and exit positions of each inspection sub-area, a path planning algorithm is used to consider obstacles, path length, and distribution of drainage system equipment, and an inspection path and inspection plan are generated for each robot dog in each inspection sub-area, and sent to the inspection management system of each robot dog;

[0021] Each robot dog uses a navigation control algorithm in the corresponding inspection sub-area to perform inspections according to the inspection path and inspection plan issued by the inspection management system, avoiding obstacles and hydrophobic objects to be inspected.

[0022] Furthermore, when there are multiple robot dogs in each inspection sub-area, one robot dog is set as a leader and the other robot dogs are set as followers to achieve collaborative inspection of the drainage system in each inspection sub-area.

[0023] Furthermore, in S2, multiple types of sensing devices are used to obtain multi-source data related to water drainage in the actual system, including:

[0024] Each robot dog is equipped with a sound sensor, infrared thermal imaging equipment and camera equipment;

[0025] The sound sensor is used to collect the sound signals of condensed water and steam flowing out of the drain port;

[0026] Use infrared thermal imaging equipment to obtain the pipe temperature distribution of the steam heat network drain system, generate thermal imaging images of the drain system, and indirectly calculate the pipe steam pressure using the relationship between temperature and pressure. It is not limited by lighting conditions and can conduct continuous inspections and drain adjustments throughout the day.

[0027] Use video equipment to obtain images or videos related to the operation of the drain system, as well as images or videos of the robot dog adjusting the drain valve.

[0028] Furthermore, in S2, the inspection management system obtains power information in real time and cooperates with the charging station to perform autonomous charging, including:

[0029] The inspection management system obtains the battery information of each robot dog in real time. When the battery is lower than the set threshold, the robot dog automatically returns to the nearest idle charging station for charging.

[0030] Alternatively, each robot dog can reserve a charging period in advance for scheduled charging based on its own remaining power and the expected completion time of the drainage inspection and adjustment task;

[0031] Furthermore, the data transmission device uses a wireless communication module carried by the robot dog to transmit data to a remote management background.

[0032] Further, in S3, a steam heating network drainage diagnosis model is established, including:

[0033] The remote management backend obtains multi-source data related to the actual system drain, and combines the drain temperature, flow and pressure operation data collected by the system to screen the operation data features that are strongly related to the steam heat network drain diagnosis;

[0034] Considering different steam load ranges and different outdoor temperatures, the steam heat network drain system is divided into multiple operating conditions, and the condensate volume of the drain system in each inspection sub-area at each time period under different operating conditions is predicted. When the condensate volume exceeds the set amount, the steam heat network drain abnormality is diagnosed; each operating condition corresponds to a different steam load and outdoor temperature range;

[0035] Taking the operating data characteristics that are strongly correlated with the steam heating network drain diagnosis, steam load, outdoor temperature, and condensate volume as input, and the steam heating network drain diagnosis results as output, a deep learning algorithm is used for model training to establish a steam heating network drain diagnosis model for each inspection sub-area under various operating conditions.

[0036] Further, in S3, a target steam trap identification model is established, including:

[0037] The remote management backend obtains multi-source data related to the actual system drain, and selects multiple drain system images under different lighting conditions in different inspection sub-areas as the original objects for target drain valve identification;

[0038] After grayscale processing, denoising and image enhancement of the steam trap system image, a deep learning algorithm is used to extract shallow features of the steam trap image, fuse deep features, identify steam trap targets and segment instances, and establish a target steam trap recognition model.

[0039] Furthermore, in S3, a fluid type monitoring model is established, including:

[0040] The remote management background monitors the adjustment process of the steam trap and the condensate discharge process of the steam trap, and obtains the fluid image or video data signal and fluid sound signal of the steam trap;

[0041] After preprocessing and feature extraction of the fluid image or video data signal and the fluid sound signal of the drain port, the two types of signals are subjected to feature fusion and dimension reduction processing;

[0042] The fusion data set after dimension reduction is used as data and the fluid type is used as output. The model is trained through deep learning algorithm to establish a fluid type monitoring model.

[0043] Further, the S4 includes:

[0044] During the inspection of the steam heat network drain system in the corresponding inspection sub-area, each robot dog uses the steam heat network drain diagnostic model to identify potential drain abnormality information; when an abnormality is found, the robot dog immediately activates the navigation system and goes to the abnormal drain system location based on the built-in map and path planning algorithm, while using the camera equipment for auxiliary positioning and obstacle avoidance;

[0045] When the robot dog reaches the abnormal drain system location, it uses a camera to obtain the drain system image, and uses the sent target drain valve recognition model to recognize the drain system image to obtain the target drain valve object to be adjusted;

[0046] The robot dog is used as an intelligent agent, the target steam trap object data is used as the input state, and the motion parameters of the end effector of the bionic robot arm are used as the output. The bionic robot arm is trained through a deep reinforcement learning algorithm to continuously interact with the steam heat network steam trap system operating environment. According to the feedback and reward signals in the environment, the bionic robot arm is continuously trained and corrected through trial and error, guiding the bionic robot arm to autonomously align, open and close the target steam trap. At the same time, in the process of opening the target steam trap to discharge condensate, the fluid type monitoring model is used to monitor the fluid type of the steam trap. When the steam trap is detected as condensate, the bionic robot arm is controlled to continue to open the target steam trap to completely discharge the condensate. When the steam trap is detected as steam, the bionic robot arm is controlled to close the steam trap to avoid steam leakage.

[0047] Among them, the motion parameters of the end effector of the bionic robotic arm include the action behaviors of each joint during the training process of aligning, opening and closing the steam trap: in each frame of the bionic robotic arm, it can rotate clockwise or counterclockwise to open and close the steam trap, and follow the corresponding rotation arc limit.

[0048] The beneficial effects of the present invention are:

[0049] (1) The present invention divides the steam heat network drain system into multiple inspection sub-areas, and sets a charging station and multiple robot dogs for each inspection sub-area. The robot dogs can flexibly shuttle through every corner of the steam heat network drain system to ensure the comprehensiveness of the steam trap adjustment, greatly reduce the frequency and intensity of manual inspection and adjustment, reduce operation and maintenance costs, and avoid safety risks during manual inspection and adjustment. The robot dogs in each inspection sub-area can simultaneously and efficiently inspect and adjust their respective areas, thereby improving work efficiency.

[0050] (2) Each robot dog of the present invention uses a navigation control algorithm in each inspection sub-area to inspect the steam heat network drain system according to the inspection path and inspection plan issued by the inspection management system. During the inspection process, multiple types of sensor equipment are used to obtain multi-source data related to the actual system drain, and the data is transmitted to the remote management background through the data transmission equipment. The multi-source data of the on-site drain system can be obtained in real time through the robot dog, providing a data basis for the subsequent adjustment of the drain valve; the robot dog uses advanced navigation control algorithms to autonomously navigate and avoid obstacles in the complex steam heat network drain system to ensure the smooth progress of the inspection; the inspection management system obtains power information in real time, cooperates with the charging station to perform autonomous charging, and ensures that sufficient power is always maintained during the inspection and adjustment process to avoid inspection and adjustment interruptions due to power exhaustion;

[0051] (3) The remote management background of the present invention obtains multi-source data related to the actual system drain, and after combining the drain operation data collected by the system for data processing and model training, a steam heat network drain diagnosis model, a target drain valve identification model and a fluid type monitoring model are established respectively, and sent to each robot dog drain adjustment system. During the inspection process, each robot dog uses the sent steam heat network drain diagnosis model to obtain real-time drain abnormality information, and autonomously controls the robot dog to walk to the drain abnormality position. After using the target drain valve identification model to identify the target drain valve object to be adjusted, the bionic robot arm is used to align and open the target drain valve to discharge condensate. At the same time, the fluid type monitoring model is used to monitor the drain port fluid type to control the bionic robot arm to continue to open the target drain valve or Close the steam trap; it can use deep learning algorithms, image recognition, sound recognition and other algorithms to first diagnose the steam heat network drain. The robot dog can obtain the abnormal drain information in real time, without the need for manual investigation one by one, thus improving the inspection efficiency. When diagnosing the abnormality, the robot dog can quickly reach the abnormal drain position, and then perform image recognition on the target steam trap to obtain the target steam trap object to be adjusted, while monitoring the type of fluid at the drain port. Finally, the bionic robotic arm is controlled to open and close the steam trap to ensure the complete discharge of condensate and avoid steam leakage, thus avoiding the safety risks brought by manual operation and improving the accuracy and reliability of the operation, thus realizing the intelligent management and decision-making of the steam heat network drain system, optimizing the performance of the steam heat network and ensuring the safe and stable operation of the steam heat network.

[0052] Other features and advantages will be described in the following description, and partly become apparent from the description, or understood by practicing the invention. The purpose and other advantages of the invention are realized and obtained by the structures particularly pointed out in the description and the drawings.

[0053] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0055] Figure 1 This is a flow chart of a steam heating network trap adjustment method based on an embodied intelligent robot dog according to the present invention;

[0056] Figure 2It is a schematic diagram of the regulating structure of the steam heat network drain valve based on the embodied intelligent robot dog of the present invention;

[0057] Figure 3 It is a schematic diagram of the structure of the steam heating network drainage system of the present invention;

[0058] Figure 4 This is a flow chart of the method for adjusting a steam trap by the robot dog of the present invention. DETAILED DESCRIPTION

[0059] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0060] Example 1

[0061] like Figure 1 As shown, this embodiment 1 provides a steam heat network trap adjustment method based on an embodied intelligent robot dog, which includes:

[0062] S1. Divide the steam heating network drainage system into multiple inspection sub-areas, and set up a charging station and multiple robot dogs for each inspection sub-area; each robot dog is equipped with multiple types of sensor equipment, data transmission equipment, bionic mechanical arms, inspection management system and drainage regulation system;

[0063] S2. Each robot dog uses a navigation control algorithm in each inspection sub-area to inspect the steam heat network drainage system according to the inspection path and inspection plan issued by the inspection management system. During the inspection process, multiple types of sensor equipment are used to obtain multi-source data related to the actual system drainage, and the data is transmitted to the remote management background through the data transmission equipment. At the same time, the inspection management system obtains power information in real time and cooperates with the charging station for autonomous charging;

[0064] S3. The remote management background obtains multi-source data related to the actual system draining, and after data processing and model training combined with the draining operation data collected by the system, establishes a steam heat network draining diagnosis model, a target drain valve identification model and a fluid type monitoring model, and sends them to each robot dog draining regulation system;

[0065] S4. During the inspection process, each robot dog uses the steam heat network drain diagnosis model to obtain real-time drain abnormality information, and autonomously controls the robot dog to walk to the drain abnormality position. After using the target drain valve recognition model to identify the target drain valve object to be adjusted, the bionic robot arm is used to align and open the target drain valve to discharge condensate. At the same time, the fluid type monitoring model is used to monitor the drain port fluid type to control the bionic robot arm to continue to open the target drain valve or close the drain valve.

[0066] like Figure 2 As shown in the figure, the steam heat network has multiple drain pipes and corresponding drain valves, through which condensed water is discharged, and each robot dog is charged through the deployed charging stations; the remote management background can be used to receive steam heat network related drain operation data, monitor the robot dog's path, inspection adjustment parameters and power information, and after multi-source data fusion and deep learning algorithm analysis and processing, establish the steam heat network drain diagnosis model, target drain valve identification model and fluid type monitoring model, and send them to each robot dog. In actual applications, when the drain valve is closed, the robot dog should continue to stay for a period of time, and use the steam heat network drain diagnosis model again to determine whether the abnormal information has really disappeared.

[0067] like Figure 3 As shown in the figure, the main functions of the stop valve, check valve, steam trap and exhaust valve in the steam hot network drain system are as follows:

[0068] 1) Stop valve: The stop valve relies on the valve stem pressure to make the valve disc sealing surface fit tightly with the valve seat sealing surface to prevent the flow of the medium. It is mainly used to cut off the fluid passage and regulate the fluid flow;

[0069] 2) Check valve: The function of the check valve is to prevent the backflow of the medium. In the steam heating system, the check valve can prevent the backflow of steam or other media to ensure the normal operation and safety of the system;

[0070] 3) Steam trap: The main function of the steam trap is to automatically remove condensed water from the steam pipe and prevent steam leakage;

[0071] 4) Exhaust valve: The function of the exhaust valve is to prevent the formation of air blockage due to the accumulation of air during the operation of the system, which will hinder the circulation of hot water and generate noise. In addition, the exhaust valve is used to remove air from the pipeline before the pipeline water pressure test.

[0072] In this embodiment, in S1, the steam heating network drainage system is divided into multiple inspection sub-areas, and each inspection sub-area is provided with a charging station and multiple robot dogs, including:

[0073] The geographic information data of the steam heat network drain system is obtained by using GIS maps, and the physical entities of the system are mapped into virtual space to generate a steam heat network drain system twin. The installed pressure sensors, temperature sensors, and flow sensors are used to collect the drain system operation data in real time, drive the steam heat network drain system twin, and establish a steam heat network drain system digital twin model containing GIS maps.

[0074] The steam heat network drain system digital twin model is used to consider the steam pipeline layout, trap distribution, drain valve location, drain equipment quantity, and geographical characteristics, and the steam heat network drain system is divided into multiple inspection sub-areas.

[0075] The location of the power supply, the time cost of the robot dog returning to charge, and the path length are considered to select the location of the charging station in each inspection sub-area. The number of charging stations required in each inspection sub-area is calculated by considering the task frequency of drainage inspection, the working time of the robot dog, the number of robot dogs, and the battery life. The number of robot dogs is allocated according to the drainage regulation workload, complexity, and emergency response requirements of each inspection sub-area. The model of the robot dog is selected according to the adaptability of the drainage regulation task and geographical characteristics.

[0076] The charging stations are set up according to the location and number of charging stations in each inspection sub-area, and the robot dogs are set up according to the number and model of robot dogs in each inspection sub-area.

[0077] It should be noted that the deployment of steam traps includes:

[0078] 1) Low point: The steam trap should be installed at the lowest point of the steam heating network, because condensate will flow to these locations under the action of gravity;

[0079] 2) Bottom of the riser: For pipelines with a rising section, install a steam trap at the bottom of the riser to prevent condensate accumulation and cause water hammer;

[0080] 3) Horizontal pipes at certain intervals: On long horizontal pipes, a steam trap is usually installed at certain intervals. The specific interval depends on the pipe diameter and steam flow rate.

[0081] Therefore, the inspection route of the robot dog should be based on the deployment of the steam trap to ensure that each steam trap can be inspected. The robot dogs can coordinate inspections and adjustments, and collect and upload relevant operating information of the steam trap system in real time during the inspection process.

[0082] In this embodiment, in S2, each robot dog uses a navigation control algorithm in each inspection sub-area to inspect the steam heating network drain system according to the inspection path and inspection plan issued by the inspection management system, including:

[0083] Each inspection sub-area is recorded as , the number of steam traps corresponding to each sub-area is recorded as ;

[0084] According to the entrance and exit positions of each inspection sub-area, a path planning algorithm is used to consider obstacles, path length, and distribution of drainage system equipment, and an inspection path and inspection plan are generated for each robot dog in each inspection sub-area, and sent to the inspection management system of each robot dog;

[0085] Each robot dog uses a navigation control algorithm in the corresponding inspection sub-area to perform inspections according to the inspection path and inspection plan issued by the inspection management system, avoiding obstacles and hydrophobic objects to be inspected.

[0086] In this embodiment, when there are multiple robot dogs in each inspection sub-area, one robot dog is set as a leader and the other robot dogs are set as followers to achieve collaborative inspection of the drainage system in each inspection sub-area.

[0087] In this embodiment, in S2, multiple types of sensing devices are used to obtain multiple source data related to water drainage in the actual system, including:

[0088] Each robot dog is equipped with a sound sensor, infrared thermal imaging equipment and camera equipment;

[0089] The sound sensor is used to collect the sound signals of condensed water and steam flowing out of the drain port;

[0090] Use infrared thermal imaging equipment to obtain the pipe temperature distribution of the steam heat network drain system, generate thermal imaging images of the drain system, and indirectly calculate the pipe steam pressure using the relationship between temperature and pressure. It is not limited by lighting conditions and can conduct continuous inspections and drain adjustments throughout the day.

[0091] Use video equipment to obtain images or videos related to the operation of the drain system, as well as images or videos of the robot dog adjusting the drain valve.

[0092] In this embodiment, in S2, the inspection management system obtains power information in real time and cooperates with the charging station to perform autonomous charging, including:

[0093] The inspection management system obtains the battery information of each robot dog in real time. When the battery is lower than the set threshold, the robot dog automatically returns to the nearest idle charging station for charging.

[0094] Alternatively, each robot dog can reserve a charging period in advance for scheduled charging based on its own remaining power and the expected completion time of the drainage inspection and adjustment task;

[0095] Furthermore, the data transmission device uses a wireless communication module carried by the robot dog to transmit data to a remote management background.

[0096] It should be noted that automatic return to charge: when the battery level of the robot dog is lower than the set threshold, it will automatically return to the nearest charging station for charging. The location of the charging station should be planned in advance and can be accurately located by GPS or other positioning means.

[0097] Appointment-based charging: In order to avoid congestion caused by multiple robot dogs arriving at the same charging station at the same time, appointment-based charging management can be implemented, that is, each robot dog can reserve a charging period in advance based on its own remaining power and the expected completion time of the drainage task.

[0098] In this embodiment, in S3, a steam heating network drainage diagnosis model is established, including:

[0099] The remote management backend obtains multi-source data related to the actual system drain, and combines the drain temperature, flow and pressure operation data collected by the system to screen the operation data features that are strongly related to the steam heat network drain diagnosis;

[0100] Considering different steam load ranges and different outdoor temperatures, the steam heat network drain system is divided into multiple operating conditions, and the condensate volume of the drain system in each inspection sub-area at each time period under different operating conditions is predicted. When the condensate volume exceeds the set amount, the steam heat network drain abnormality is diagnosed; each operating condition corresponds to a different steam load and outdoor temperature range;

[0101] Taking the operating data characteristics that are strongly correlated with the steam heating network drain diagnosis, steam load, outdoor temperature, and condensate volume as input, and the steam heating network drain diagnosis results as output, a deep learning algorithm is used for model training to establish a steam heating network drain diagnosis model for each inspection sub-area under various operating conditions.

[0102] It should be noted that the amount of condensate in the steam heat network drain system is correlated with the steam load and meteorological parameters, as shown in the following:

[0103] 1) Steam heating network load:

[0104] Direct correlation: The load of the steam heating network, i.e. the amount of steam consumed in the system, directly affects the amount of condensate produced;

[0105] Efficiency factor: The overall efficiency of the system will also affect the amount of condensate. If the steam utilization efficiency is high, then the amount of condensate produced may be relatively small under the same load; vice versa.

[0106] 2) Meteorological parameters:

[0107] Outdoor temperature: Temperature changes in weather conditions have a great impact on the steam heating network. For example, in cold seasons or regions, in order to maintain the indoor temperature, the heating network needs to provide more heat, which will lead to an increase in steam consumption and thus increase the production of condensed water. On the contrary, in warm climate conditions, the load of the heating network is usually reduced, and the corresponding condensed water will also decrease;

[0108] Humidity: Air humidity can also affect the amount of condensate. Higher humidity levels may cause the air to contain more moisture, which is more likely to condense when it encounters a cooler surface. Although this is not a major factor that directly affects the amount of condensate within the steam heating network, it still plays a role in certain specific application scenarios (such as open cooling towers).

[0109] In actual applications, deep learning algorithms are used for model training to establish steam heat network drainage diagnosis models for each inspection sub-area under various operating conditions, including:

[0110] 1) Algorithm selection: Use deep learning algorithms such as long short-term memory networks (LSTM), convolutional neural networks (CNN), or hybrid models, which are good at processing time series data and complex nonlinear relationships;

[0111] 2) Input and output definition:

[0112] Input: Obtain operating data features that are strongly related to steam heat network drainage diagnosis, as well as information such as steam load, outdoor temperature, condensate volume, current time period, etc.;

[0113] Output: steam trap diagnosis result (normal / abnormal). When the result is abnormal, steam trap adjustment is required.

[0114] Model training: Use the collected historical data as the training set, divide the data into training set, validation set and test set. Optimize hyperparameters through cross-validation to ensure good generalization ability of the model;

[0115] Performance evaluation: Use indicators such as mean square error (MSE) and mean absolute error (MAE) to evaluate the accuracy of the model in predicting the amount of condensed water;

[0116] 3) Real-time diagnosis

[0117] Set thresholds: Based on historical data and business needs, set reasonable upper thresholds for the amount of condensed water under each operating condition;

[0118] Online diagnosis: Deploy the trained model to the actual operating environment, receive data from sensors in real time, and perform hydrophobic diagnosis;

[0119] Abnormal alarm: When diagnosed as abnormal, the alarm mechanism is triggered to indicate possible hydrophobic abnormality.

[0120] In this embodiment, in S3, a target steam trap identification model is established, including:

[0121] The remote management backend obtains multi-source data related to the actual system drain, and selects multiple drain system images under different lighting conditions in different inspection sub-areas as the original objects for target drain valve identification;

[0122] After grayscale processing, denoising and image enhancement of the steam trap system image, a deep learning algorithm is used to extract shallow features of the steam trap image, fuse deep features, identify steam trap targets and segment instances, and establish a target steam trap recognition model.

[0123] In practical applications, a dual-channel pixel attention mechanism network is used to denoise the hydrophobic system image. The network includes a feature extraction unit and a pixel attention feature enhancement unit. The feature extraction unit includes two subnetworks for extracting local features and global features. The first subnetwork consists of Conv+Relu and Conv, and the second subnetwork consists of Conv+BN+ Relu, Dilated Conv+BN+ Relu, and Conv. The pixel attention feature enhancement unit is used to splice the extracted different features, and at the same time introduces the pixel attention mechanism to pay more attention to the pixel information in the image. Finally, the denoised hydrophobic system image is obtained by subtracting the output noise feature map from the input noise image, which is expressed as:

[0124] ;

[0125] This is the denoised image of the hydrophobic system; This is the original hydrophobic system image; is the convolution operation; , They are the first feature splicing and the second feature splicing operation respectively; The features extracted by the feature extraction unit; , They are the two sub-networks of the feature extraction unit.

[0126] When performing image enhancement processing on the hydrophobic system image, a Retinex image enhancement algorithm is used. This algorithm is based on the HSI color space and adaptively enhances the separated brightness component and saturation component. The Canny operator is then used to extract the edge details of the entire image and perform weighted fusion with the enhanced components to form a low-illumination image enhancement algorithm. This achieves the purpose of enhancing the detail information of the low-illumination image and adjusting the light and dark contrast information, so that the image can well display the detail information of the steam trap and provide the bionic robotic arm with information such as the alignment direction.

[0127] A deep learning algorithm is used to extract shallow features of steam trap images, fuse deep features, identify steam trap targets and segment instances. When establishing the target steam trap recognition model, an improved YOLOv5 network algorithm is used. The YOLOv5 network structure is mainly composed of three parts: Backbone, Neck and Head. These three parts work together to achieve target recognition of the input image. By adding a mixed superposition attention mechanism to the shallow backbone network and the deep neck network, the feature extraction capability of the steam trap is enhanced, the perception of the deep network on global semantic information is improved, and a new feature extraction network and feature fusion network are formed; the instance segmentation head network is added, and the receptive field is increased by using the dilated convolutional pooling pyramid to ensure that the improved network can complete target recognition and instance segmentation, and obtain the coordinates of the center point of the steam trap in the image and the segmentation results of the edge details.

[0128] The hybrid superposition attention mechanism is an attention module for feedforward convolutional neural networks. It uses the progression of the convolutional network structure to adaptively match the learning target features, performs weighted fusion of channel attention weights and spatial attention weights, concatenates and activates the feature maps, and obtains a hybrid attention map with two-dimensional outputs of channel and space. The final result is multiplied with the input map, which can improve the accuracy of target steam trap identification in different scenarios.

[0129] In the instance segmentation feature fusion stage, a path aggregation network is used to add a bottom-up path to the original FPN feature pyramid, and a horizontal link is added to each layer of the feature extraction network to strengthen the contextual semantic information and spatial information. The maximum pooling layer is used to fuse the feature map to ensure that the contextual information can be fully utilized. The path aggregation network is used to accurately retain the spatial information and semantic information, and the global target information is used to locate the pixel points and the target steam trap and form a mask.

[0130] In this embodiment, in S3, a fluid type monitoring model is established, including:

[0131] The remote management background monitors the adjustment process of the steam trap and the condensate discharge process of the steam trap, and obtains the fluid image or video data signal and fluid sound signal of the steam trap;

[0132] After preprocessing and feature extraction of the fluid image or video data signal and the fluid sound signal of the drain port, the two types of signals are subjected to feature fusion and dimension reduction processing;

[0133] The fusion data set after dimension reduction is used as data and the fluid type is used as output. The model is trained through deep learning algorithm to establish a fluid type monitoring model.

[0134] In practical applications, in order to further analyze whether the water flowing out of the drain outlet is condensed water or steam, a robot dog is used to obtain the image signal, video signal and sound signal of the drain outlet. First, the denoised sound signal is framed, and each sound sample contains a preset number of sampling points. Then, the sound signal characteristics of the drain outlet are extracted based on each sample, including root amplitude, standard deviation, absolute mean, variance, root mean square, etc., as the time domain characteristics of the sound signal; secondly, the image segmentation algorithm is used to obtain the area of ​​interest for the image signal or video signal, and then the image signal feature extraction algorithm is used to extract features; then, the feature-level fusion method is used to fuse the image signal, video signal and sound signal of the drain outlet, and after obtaining the visual and sound high-dimensional feature data sets, the PCA algorithm is used to perform feature dimensionality reduction processing; finally, the XGBoost algorithm is used to train the fluid type monitoring model to obtain the monitoring result of whether the drain outlet fluid is condensed water or steam.

[0135] In this embodiment, if Figure 4 As shown, the specific process of the S4, the method for adjusting the steam trap by the robot dog includes:

[0136] During the inspection of the steam heat network drain system in the corresponding inspection sub-area, each robot dog uses the steam heat network drain diagnostic model to identify potential drain abnormality information; when an abnormality is found, the robot dog immediately activates the navigation system and goes to the abnormal drain system location based on the built-in map and path planning algorithm, while using the camera equipment for auxiliary positioning and obstacle avoidance;

[0137] When the robot dog reaches the abnormal drain system location, it uses a camera to obtain the drain system image, and uses the sent target drain valve recognition model to recognize the drain system image to obtain the target drain valve object to be adjusted;

[0138] The robot dog is used as an intelligent agent, the target steam trap object data is used as the input state, and the motion parameters of the end effector of the bionic robot arm are used as the output. The bionic robot arm is trained through a deep reinforcement learning algorithm to continuously interact with the steam heat network steam trap system operating environment. According to the feedback and reward signals in the environment, the bionic robot arm is continuously trained and corrected through trial and error, guiding the bionic robot arm to autonomously align, open and close the target steam trap. At the same time, in the process of opening the target steam trap to discharge condensate, the fluid type monitoring model is used to monitor the fluid type of the steam trap. When the steam trap is detected as condensate, the bionic robot arm is controlled to continue to open the target steam trap to completely discharge the condensate. When the steam trap is detected as steam, the bionic robot arm is controlled to close the steam trap to avoid steam leakage.

[0139] Among them, the motion parameters of the end effector of the bionic robotic arm include the action behaviors of each joint during the training process of aligning, opening and closing the steam trap: in each frame of the bionic robotic arm, it can rotate clockwise or counterclockwise to open and close the steam trap, and follow the corresponding rotation arc limit.

[0140] In actual applications, the process of opening and closing the steam trap by the bionic robotic arm includes: moving the bionic robotic arm to the top of the steam trap and opening the end effector; moving the end effector downward to clamp the steam trap; and rotating the end effector to the left or right to the set threshold.

[0141] State space design: In the simulation experiment environment of opening and closing the steam trap by the bionic robot arm, it is necessary to obtain the environmental information observed by the robot dog agent and the state information of the agent itself, including the coordinate position of the end effector, each value of the end effector action, the distance and coordinate position information between the target steam trap and the end effector, and the rotation angle and coordinate position of each joint of the bionic robot arm, etc.

[0142] Action space design: The action is set as the end effector of the bionic robot arm moves to the coordinate position of the target steam trap, and adjusts the position and rotation angle of the clamped steam trap at the same time;

[0143] Reward function design: The reward function is designed so that the agent can learn the optimal strategy in a small number of time steps to align the gripping of the steam trap and the opening and closing of the steam trap to maximize the cumulative reward value and reach convergence. The design of the reward function will affect the convergence time and effect of the deep reinforcement learning algorithm. When the action of the end effector clamping the steam trap occurs in the target area, a reward of 0.5 is given; when the end effector successfully clamps the steam trap and rotates the set angle, a reward value of 1 is given, and the reward value is 0 in other cases.

[0144] The deep reinforcement learning algorithm used is the improved TD3 algorithm, which introduces a prediction module to predict the state and reward value at the next moment. The bionic robot arm obtains the current state by interacting with the steam heat network drainage system environment. , and used as the input of the deep reinforcement learning actor-critic network to obtain the action of the bionic robotic arm , perform the action Get reward value . The state at time t ,action , Reward Value and the state at the next moment t+1 As experience value Stored in the experience pool, when the amount of data in the experience pool reaches the set threshold, the algorithm parameters are updated: the state prediction value at the next moment Input the policy network to get the action , the state-action Input into the prediction module to obtain the state prediction value at time t+2 ; The predicted state value at time t+1 , the state prediction value at time t+2 and the predicted reward value at time t+1 Input into the policy network, update the parameters of the actor network, and update the parameters of the prediction module at the same time.

[0145] The prediction module is expressed as:

[0146] ;

[0147] , They are the state and reward value prediction functions respectively.

[0148] In several embodiments provided in the present application, it should be understood that the disclosed system and method can also be implemented in other ways. The system embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the systems, methods and computer program products according to multiple embodiments of the present invention. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of the code, and a part of the module, program segment or code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0149] In addition, the functional modules in each embodiment of the present invention can be integrated together to form an independent part, or each module can exist separately, or two or more modules can be integrated to form an independent part. If the function is implemented in the form of a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, RandomAccess Memory), disk or optical disk and other media that can store program codes.

[0150] Based on the above ideal embodiments of the present invention, the relevant staff can make various changes and modifications without departing from the technical concept of the present invention through the above description. The technical scope of the present invention is not limited to the contents of the specification, and its technical scope must be determined according to the scope of the claims.

Claims

1. A steam heat network trap adjustment method based on an embodied intelligent robot dog, characterized in that: It includes: S1. Divide the steam heating network drainage system into multiple inspection sub-areas, and set up a charging station and multiple robot dogs for each inspection sub-area; each robot dog is equipped with multiple types of sensor equipment, data transmission equipment, bionic mechanical arms, inspection management system and drainage regulation system; S2. Each robot dog uses a navigation control algorithm in each inspection sub-area to inspect the steam heat network drainage system according to the inspection path and inspection plan issued by the inspection management system. During the inspection process, multiple types of sensor equipment are used to obtain multi-source data related to the actual system drainage, and the data is transmitted to the remote management background through the data transmission equipment. At the same time, the inspection management system obtains power information in real time and cooperates with the charging station for autonomous charging; S3. The remote management background obtains multi-source data related to the actual system draining, and after data processing and model training combined with the draining operation data collected by the system, establishes a steam heat network draining diagnosis model, a target drain valve identification model and a fluid type monitoring model, and sends them to each robot dog draining regulation system; S4. During the inspection process, each robot dog uses the steam heat network drain diagnosis model to obtain real-time drain abnormality information, and autonomously controls the robot dog to walk to the drain abnormality position. After using the target drain valve recognition model to identify the target drain valve object to be adjusted, the bionic robot arm is used to align and open the target drain valve to discharge condensate. At the same time, the fluid type monitoring model is used to monitor the drain port fluid type to control the bionic robot arm to continue to open the target drain valve or close the drain valve.

2. The steam heating network trap adjustment method according to claim 1, characterized in that: In S1, the steam heating network drainage system is divided into multiple inspection sub-areas, and each inspection sub-area is provided with a charging station and multiple robot dogs, including: The geographic information data of the steam heat network drain system is obtained by using GIS maps, and the physical entities of the system are mapped into virtual space to generate a steam heat network drain system twin. The installed pressure sensors, temperature sensors, and flow sensors are used to collect the drain system operation data in real time, drive the steam heat network drain system twin, and establish a steam heat network drain system digital twin model containing GIS maps. The steam heat network drain system digital twin model is used to consider the steam pipeline layout, trap distribution, drain valve location, drain equipment quantity, and geographical characteristics, and the steam heat network drain system is divided into multiple inspection sub-areas. The location of the power supply, the time cost of the robot dog returning to charge, and the path length are considered to select the location of the charging station in each inspection sub-area. The number of charging stations required in each inspection sub-area is calculated by considering the task frequency of drainage inspection, the working time of the robot dog, the number of robot dogs, and the battery life. The number of robot dogs is allocated according to the drainage regulation workload, complexity, and emergency response requirements of each inspection sub-area. The model of the robot dog is selected according to the adaptability of the drainage regulation task and geographical characteristics. The charging stations are set up according to the location and number of charging stations in each inspection sub-area, and the robot dogs are set up according to the number and model of robot dogs in each inspection sub-area.

3. The steam heating network trap adjustment method according to claim 1, characterized in that: In S2, each robot dog uses a navigation control algorithm in each inspection sub-area to inspect the steam heating network drain system according to the inspection path and inspection plan issued by the inspection management system, including: Each inspection sub-area is recorded as , the number of steam traps corresponding to each sub-area is recorded as ; According to the entrance and exit positions of each inspection sub-area, a path planning algorithm is used to consider obstacles, path length, and distribution of drainage system equipment, and an inspection path and inspection plan are generated for each robot dog in each inspection sub-area, and sent to the inspection management system of each robot dog; Each robot dog uses a navigation control algorithm in the corresponding inspection sub-area to perform inspections according to the inspection path and inspection plan issued by the inspection management system, avoiding obstacles and hydrophobic objects to be inspected.

4. The steam heating network trap adjustment method according to claim 3, characterized in that: When there are multiple robot dogs in each inspection sub-area, one robot dog is set as a leader and the other robot dogs are set as followers to achieve collaborative inspection of the drainage system in each inspection sub-area.

5. The steam heating network trap adjustment method according to claim 1, characterized in that: In S2, multiple types of sensing devices are used to obtain multi-source data related to water drainage in the actual system, including: Each robot dog is equipped with a sound sensor, infrared thermal imaging equipment and camera equipment; The sound sensor is used to collect the sound signals of condensed water and steam flowing out of the drain port; Use infrared thermal imaging equipment to obtain the pipe temperature distribution of the steam heat network drain system, generate thermal imaging images of the drain system, and indirectly calculate the pipe steam pressure using the relationship between temperature and pressure. It is not limited by lighting conditions and can conduct continuous inspections and drain adjustments throughout the day. Use video equipment to obtain images or videos related to the operation of the drain system, as well as images or videos of the robot dog adjusting the drain valve.

6. The steam heating network trap adjustment method according to claim 1, characterized in that: In S2, the inspection management system obtains power information in real time and cooperates with the charging station to perform autonomous charging, including: The inspection management system obtains the battery information of each robot dog in real time. When the battery is lower than the set threshold, the robot dog automatically returns to the nearest idle charging station for charging. Alternatively, each robot dog can reserve a charging period in advance for scheduled charging based on its own remaining power and the expected completion time of the drainage inspection and adjustment task; Furthermore, the data transmission device uses a wireless communication module carried by the robot dog to transmit data to a remote management background.

7. The steam heating network trap adjustment method according to claim 1, characterized in that: In S3, a steam heating network drainage diagnosis model is established, including: The remote management backend obtains multi-source data related to the actual system drain, and combines the drain temperature, flow and pressure operation data collected by the system to screen the operation data features that are strongly related to the steam heat network drain diagnosis; Considering different steam load ranges and different outdoor temperatures, the steam heat network drain system is divided into multiple operating conditions, and the condensate volume of the drain system in each inspection sub-area at each time period under different operating conditions is predicted. When the condensate volume exceeds the set amount, the steam heat network drain abnormality is diagnosed; each operating condition corresponds to a different steam load and outdoor temperature range; Taking the operating data characteristics that are strongly correlated with the steam heating network drain diagnosis, steam load, outdoor temperature, and condensate volume as input, and the steam heating network drain diagnosis results as output, a deep learning algorithm is used for model training to establish a steam heating network drain diagnosis model for each inspection sub-area under various operating conditions.

8. The steam heating network trap adjustment method according to claim 1, characterized in that: In S3, a target steam trap identification model is established, including: The remote management backend obtains multi-source data related to the actual system drain, and selects multiple drain system images under different lighting conditions in different inspection sub-areas as the original objects for target drain valve identification; After grayscale processing, denoising and image enhancement of the steam trap system image, a deep learning algorithm is used to extract shallow features of the steam trap image, fuse deep features, identify steam trap targets and segment instances, and establish a target steam trap recognition model.

9. The steam heating network trap adjustment method according to claim 1, characterized in that: In S3, a fluid type monitoring model is established, including: The remote management background monitors the adjustment process of the steam trap and the condensate discharge process of the steam trap, and obtains the fluid image or video data signal and fluid sound signal of the steam trap; After preprocessing and feature extraction of the fluid image or video data signal and the fluid sound signal of the drain port, the two types of signals are subjected to feature fusion and dimension reduction processing; The fusion data set after dimension reduction is used as data and the fluid type is used as output. The model is trained through deep learning algorithm to establish a fluid type monitoring model.

10. The steam heating network trap adjustment method according to claim 1, characterized in that: The S4 includes: During the inspection of the steam heat network drain system in the corresponding inspection sub-area, each robot dog uses the steam heat network drain diagnostic model to identify potential drain abnormality information; when an abnormality is found, the robot dog immediately activates the navigation system and goes to the abnormal drain system location based on the built-in map and path planning algorithm, while using the camera equipment for auxiliary positioning and obstacle avoidance; When the robot dog reaches the abnormal drain system location, it uses a camera to obtain the drain system image, and uses the sent target drain valve recognition model to recognize the drain system image to obtain the target drain valve object to be adjusted; The robot dog is used as an intelligent agent, the target steam trap object data is used as the input state, and the motion parameters of the end effector of the bionic robot arm are used as the output. The bionic robot arm is trained through a deep reinforcement learning algorithm to continuously interact with the steam heat network steam trap system operating environment. According to the feedback and reward signals in the environment, the bionic robot arm is continuously trained and corrected through trial and error, guiding the bionic robot arm to autonomously align, open and close the target steam trap. At the same time, in the process of opening the target steam trap to discharge condensate, the fluid type monitoring model is used to monitor the fluid type of the steam trap. When the steam trap is detected as condensate, the bionic robot arm is controlled to continue to open the target steam trap to completely discharge the condensate. When the steam trap is detected as steam, the bionic robot arm is controlled to close the steam trap to avoid steam leakage. Among them, the motion parameters of the end effector of the bionic robotic arm include the action behaviors of each joint during the training process of aligning, opening and closing the steam trap: in each frame of the bionic robotic arm, it can rotate clockwise or counterclockwise to open and close the steam trap, and follow the corresponding rotation arc limit.

Images