A pantograph-catwalk online monitoring system and method based on edge computing

Abstract

The present invention discloses a pantograph-catenary online monitoring system and method based on edge computing, which belongs to the technical field of pantograph-catenary online monitoring. The system includes ground and train; the ground includes cloud server; the train includes end sensor, end controller, edge server and train control and monitoring system; the present invention makes full use of edge computing technology, performs unified data collection on the train, and can process data on-site in time to meet train safety requirements, assist in controlling the train, and avoid the expansion of serious accidents; utilizes end, edge, and cloud collaboration to optimize the train machine learning algorithm model and improve the accuracy of pantograph-catenary fault identification; utilizes edge-edge collaboration, the edge server on the train can connect and communicate with the edge server of the train within a set distance through its own wireless communication, forming an edge server cluster mode, improving data processing efficiency, reducing cloud server resources, and greatly reducing later operation and maintenance costs.

Description

Technical Field

[0001] The present invention belongs to the technical field of bow-catcher online monitoring, and specifically relates to a bow-catcher online monitoring system and method based on edge computing. Background Art

[0002] At present, most rail transit uses electric-driven trains, which usually form the transmission conditions of rated voltage / current through the overhead contact system, and then the train pantograph receives the current to the train's power distribution system for application, thereby providing energy for train operation. Therefore, the interaction between the pantograph and the overhead contact network forms a link between the ground fixed power supply equipment and the mobile train application, and the two mainly constitute the commonly defined "pantograph-network system".

[0003] In rail transit operation, the stable contact state between the pantograph and the contact network is the key to ensuring the excellent quality of train current collection, and reducing the failure of the pantograph-catenary system is one of the key factors to reduce rail transit operation failures. In addition, as the speed of the train increases, the possibility of separation between the pantograph and the contact network increases, the arcing of the pantograph and the contact network is more likely to occur, and the harm caused by the poor operation of the pantograph and the contact network is becoming more and more serious.

[0004] The current hardware device of the pantograph-catenary online monitoring system is installed on the train, and consists of a roof detection module, an in-car data processing module, and a bottom positioning module. The software analysis system is installed on the ground server. The system mainly monitors the dynamic operation status of the pantograph-catenary in real time through sensors such as cameras, acceleration, optical fiber, and current, and can timely and qualitatively detect defect data such as contact network pull-out value, arcing, and conductor height. However, similar methods have the following problems:

[0005] (1) Low safety real-time performance: The ground monitoring system is mainly based on data analysis. The next decision is made after manual data copying and analysis on the train at regular intervals. Once a safety problem occurs in the overhead line or pantograph, the system cannot implement the safety strategy immediately. A large amount of data still needs to be manually copied or uploaded to the ground through wireless networks for decision-making. This will greatly delay the handling of major safety issues, thereby expanding the scope of the accident.

[0006] (2) Low recognition accuracy: The system’s offline image processing recognition algorithm cannot cover all operating conditions under different environments, weather conditions, and lighting conditions. It can only rely on the same, single algorithm to identify a single data source, and cannot be combined with data collected by other trains to optimize the recognition algorithm.

[0007] (3) Performance reduction and cost increase: As the number of trains equipped with the pantograph online monitoring system increases, a large amount of data is transmitted wirelessly to the cloud server on the ground, resulting in the need for continuous expansion of data traffic, servers, and network bandwidth, which will lead to a surge in operating costs.

[0008] In summary, the current online monitoring of the pantograph and the catenary has the above-mentioned shortcomings. Therefore, there is an urgent need for a new system and method for online monitoring of the pantograph and the catenary of rail transit, so that rail transit trains can operate reliably, efficiently and safely, and optimize the benefits as much as possible. Summary of the invention

[0009] The purpose of the present invention is to solve the problems of low security and real-time performance, low recognition accuracy and batch installation of equipment in the existing rail transit bow-net online monitoring system, which leads to reduced performance and increased cost of ground servers. A bow-net online monitoring system and method based on edge computing is proposed.

[0010] The technical solution of the present invention is: an edge computing-based bow-net online monitoring system, characterized in that it includes a ground cloud server, a train end sensor, a train end controller, a train edge server and a train control monitoring system; each train includes at least one end sensor, at least one end controller and at least one edge server; the end sensors, end controllers and edge servers in the train are communicatively connected in sequence; the edge servers of each train are communicatively connected to each other and the edge servers of each train are communicatively connected to the ground cloud server; the train control monitoring system in the train is communicatively connected to the edge server.

[0011] Furthermore, the end sensor is deployed at the component to be monitored on the train, and is used to collect data of the component to be monitored;

[0012] The end controller is used to obtain the data of the component to be monitored collected by the end sensor, and perform edge computing processing on the data of the component to be monitored through the stored machine learning algorithm model, and store the feature data obtained by the edge computing processing in the edge server;

[0013] The edge server is used to receive work instructions from the ground cloud server, receive and store feature data, and process the feature data according to the stored rule engine configuration, generate corresponding information and control instructions based on the processing results, and upload them to the ground cloud server at the same time.

[0014] Furthermore, the train control monitoring system is used to receive corresponding information and control instructions generated by the edge server, and control the train according to the corresponding information and control instructions.

[0015] Furthermore, the cloud server is used to obtain the information and control instructions uploaded by the edge server, send the corresponding machine learning algorithm model to the end controller through the edge server, and update the rule engine configuration of the edge server.

[0016] Furthermore, the specific method for the side servers of each train to communicate with each other is: when the edge computing power utilization rate of the side server of the train for the feature data exceeds the set threshold, it communicates with the side servers of each train within the set distance to improve the computing power.

[0017] Based on the above system, the present invention also proposes a method for online monitoring of a bow-net based on edge computing, comprising the following steps:

[0018] S1: Use the cloud server to update the rule engine configuration of the edge server, and send the corresponding machine learning algorithm model to the end controller of each train through the edge server;

[0019] S2: When the train is running, the end sensors of the train are used to collect the data of the components to be monitored and send them to the end controller;

[0020] S3: Using the end controller to receive the data of the monitored component from the end sensor, and using the machine learning algorithm model to perform edge computing on the data of the monitored component, and storing the feature data obtained by the edge computing in the edge server;

[0021] S4: Receive feature data using the edge server, process the feature data according to the updated rule engine configuration, and send corresponding information and control instructions to the train control monitoring system based on the processing results, and upload them to the cloud server on the ground at the same time;

[0022] S5: Using the train control monitoring system to obtain information and control instructions generated by the edge server, and controlling the train according to the corresponding information and control instructions;

[0023] S6: Update the bow-net online monitoring system and enter the next bow-net online monitoring.

[0024] Furthermore, in step S6, the specific method for updating is: using the cloud server to obtain the original monitored component data stored from the edge server, and optimizing the corresponding machine learning algorithm model based on the original monitored component data; sending the optimized machine learning algorithm model to the end controller through the edge server to update the machine learning algorithm model in the end controller; using the cloud server to obtain feature data from the edge server, optimizing the corresponding machine learning algorithm model based on the feature data, and sending the optimized machine learning algorithm model to the edge server to update the rule engine configuration of the edge server.

[0025] The beneficial effects of the present invention are:

[0026] (1) Through edge computing technology, the edge server of each train can process the collected data in real time through the rule engine, and generate corresponding train control and monitoring system control instructions and information based on the processing results, so that the train can process the collected data of the end sensor in real time and make timely feedback. The collected data is processed on-site to meet the train safety delay requirements. When major safety problems occur in the contact network and pantograph, the train will carry out corresponding control in time to avoid further expansion of safety accidents;

[0027] (2) Through the collaboration of end, edge, and cloud, the cloud server on the ground can obtain the data of all end sensors and edge servers connected to the train on demand, optimize the machine learning algorithm model, and optimize and customize the machine learning algorithm model for each vehicle and equipment according to the actual operating conditions. The optimized machine learning algorithm model is sent to the corresponding edge server, and the edge server is used to uniformly update its own rule engine configuration and update the machine learning algorithm model of the end controller at the same time, so that the machine learning algorithm model of each vehicle and each device is inconsistent, which has a qualitative improvement in the corresponding vehicle recognition accuracy.

[0028] (3) When the computing power of the edge server on the train is insufficient, through edge-to-edge collaboration, it can connect and communicate with the edge server of the train within a set distance through its own wireless communication, forming a cluster mode, enhancing computing power and speeding up data processing. The processing results are transmitted to the cloud server on the ground, and the cloud server obtains the corresponding data on demand, which greatly reduces the communication data traffic, the number of cloud servers and the network bandwidth, and greatly reduces the subsequent operating costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 This is the structural diagram of the pantograph-catenary online monitoring system;

[0030] Figure 2 This is a schematic diagram of the deployment method of the bow-net online monitoring system;

[0031] Figure 3 This is a flow chart of the on-line monitoring method for the bow-catcher. DETAILED DESCRIPTION

[0032] The embodiments of the present invention will be further described below in conjunction with the accompanying drawings.

[0033] Before describing the specific embodiments of the present invention, in order to make the scheme of the present invention clearer and more complete, the abbreviations and key terms used in the present invention are first defined as follows:

[0034] Train Control and Monitoring System: (TCMS) is an integrated control system that integrates train control, monitoring and diagnosis. It provides real-time control signals for various train systems to complete train control.

[0035] Edge computing: refers to the use of an open platform that integrates network, computing, storage and application core capabilities to provide the nearest end service near the source of objects or data. Data is processed in real time near the data source (on the train). Through edge computing, data processing decisions can be made directly without being transmitted to the cloud server on the ground.

[0036] Rule engine: refers to a component embedded in the application, which separates business decisions from the application code and uses predefined semantic modules to write business decisions. The edge server receives the characteristic data of the end controller, interprets the business rule configuration sent by the ground cloud server through the rule engine, processes and identifies it, and makes certain decisions (notification, alarm, control instructions, etc.) for data that meets certain specific conditions.

[0037] Machine learning: An intelligent recognition method that is currently widely studied and applied. As a core component in the field of artificial intelligence, it is a non-explicit computer program learning data experience to optimize its own algorithm to learn to handle tasks. It has a high recognition accuracy through training with a large amount of data. The machine learning algorithm model of this application refers to commonly used algorithms, such as: decision tree, random forest, K-means, support vector machine, neural network (deep learning), linear regression, kernel density estimation and extreme learning machine algorithms.

[0038] like Figure 1 As shown, the present invention provides a pantograph-catwalk online monitoring system based on edge computing, including two parts: ground and train;

[0039] The ground mainly includes cloud servers, and the train mainly includes end sensors, end controllers, edge servers and train control monitoring systems; each train includes at least one end sensor, at least one end controller and at least one edge server; the end sensors, end controllers and edge servers in the train are communicatively connected in sequence; the edge servers of each train are communicatively connected to each other and the edge servers of each train are communicatively connected to the ground cloud server; the train control monitoring system in the train is communicatively connected to the edge servers.

[0040] The connection method between the end sensor and the end controller includes wired connection or wireless connection, the connection method between the end controller and the edge server includes wired connection or wireless connection, the connection method between the edge server and the ground cloud server and the edge servers of other trains within a set distance includes wireless connection, and the connection method between the train control monitoring system and the edge server includes wired connection or wireless connection.

[0041] The deployment locations of ground cloud servers include monitoring centers, owner group companies and OEM operation and maintenance centers.

[0042] The end sensor types include acceleration sensors, industrial cameras, thermal imaging cameras, fiber optic sensors, ultraviolet sensors, position sensors and dynamic measurement compensation sensors. The monitoring component on the train is the pantograph system, and the end sensor collects the contact network pull-out value, conductor height, contact line slope, wear, arcing photos, video images, pressure, hard points, temperature, current and positioning information data.

[0043] like Figure 2 As shown, the end sensor is connected to the end controller through a communication cable, and the end controller is connected to the edge server through a communication cable. Wireless connection can also be used. The end sensor is located at the position of the component to be monitored on the train, and is used to collect data of the component to be monitored. The end controller is located near the component to be monitored. For example, to monitor the arcing of the bow and net, the end sensor is installed at the pantograph position on the top of the train, and the end controller is installed near the roof or on the top of the carriage.

[0044] Each edge server is connected to a corresponding end controller on the train, and in order to facilitate viewing and maintenance of the edge server, each edge server is installed in a carriage of the corresponding train.

[0045] In an embodiment of the present invention, the end controller is used to obtain the data of the component to be monitored collected by the end sensor, and perform edge computing processing on the data of the component to be monitored through the stored machine learning algorithm model, and store the feature data obtained by the edge computing processing in the edge server;

[0046] The edge server is used to receive work instructions from the ground cloud server, receive and store feature data, and process the feature data according to the stored rule engine configuration, generate corresponding information and control instructions based on the processing results, and upload them to the ground cloud server at the same time.

[0047] The specific method for the side servers of each train to communicate with each other is: when the edge computing power utilization rate of the train's side server for feature data exceeds a set threshold, it communicates with the side servers of each train within a set distance to increase computing power.

[0048] When the edge server is unable to process data in a timely manner due to insufficient computing power due to an increase in the amount of data to be processed, it can connect with the edge servers of other trains within a set distance to form an edge-to-edge collaborative cluster mode to speed up data processing.

[0049] In an embodiment of the present invention, a train control monitoring system is connected to an edge server, and the train control monitoring system is installed in a cab, so that the train driver can view data and make decisions. The train control monitoring system is used to receive corresponding information and control instructions generated by the edge server, and control the train according to the corresponding information and control instructions.

[0050] In the embodiment of the present invention, the cloud server is used to obtain the information and control instructions uploaded by the edge server, send the corresponding machine learning algorithm model to the end controller through the edge server, and update the rule engine configuration of the edge server. The cloud server can also obtain the end sensor data in the edge server on demand to allow the machine learning algorithm model to continue learning and optimizing, thereby improving recognition accuracy.

[0051] Based on the above system, the present invention also proposes a method for online monitoring of the bow-net based on edge computing, such as Figure 3 As shown, the following steps are included:

[0052] S1: Use the cloud server to update the rule engine configuration of the edge server, and send the corresponding machine learning algorithm model to the end controller of each train through the edge server;

[0053] S2: When the train is running, the end sensors of the train are used to collect the data of the components to be monitored and send them to the end controller;

[0054] S3: Using the end controller to receive the data of the monitored component from the end sensor, and using the machine learning algorithm model to perform edge computing on the data of the monitored component, and storing the feature data obtained by the edge computing in the edge server;

[0055] S4: Receive feature data using the edge server, process the feature data according to the updated rule engine configuration, and send corresponding information and control instructions to the train control monitoring system based on the processing results, and upload them to the cloud server on the ground at the same time;

[0056] S5: using the train control monitoring system to obtain information and control instructions generated by the edge server, and controlling the train according to the corresponding information and control instructions;

[0057] S6: Update the bow-net online monitoring system and enter the next bow-net online monitoring.

[0058] In an embodiment of the present invention, in step S6, the specific method for updating is: using the cloud server to obtain the original monitored component data stored from the edge server, and optimizing the corresponding machine learning algorithm model based on the original monitored component data; sending the optimized machine learning algorithm model to the end controller through the edge server to update the machine learning algorithm model in the end controller; using the cloud server to obtain feature data from the edge server, optimizing the corresponding machine learning algorithm model based on the feature data, and sending the optimized machine learning algorithm model to the edge server to update the rule engine configuration of the edge server.

[0059] Therefore, through the above method, each train can give feedback on the pantograph failure immediately and quickly to avoid further expansion of safety accidents. The ground cloud server can also combine the specific operation data of each train to optimize the machine learning algorithm model in a targeted manner, and update the end controller model in time to make fault prediction more accurate. Making full use of edge computing can greatly save the system's later operating costs and make train operation more intelligent and efficient.

[0060] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific variations and combinations that do not deviate from the essence of the present invention based on the technical revelations disclosed by the present invention, and these variations and combinations are still within the protection scope of the present invention.

Claims

1. A pantograph-catwalk online monitoring system based on edge computing, characterized in that: It includes a cloud server on the ground, a terminal sensor of the train, a terminal controller of the train, a side server of the train and a train control monitoring system; each train includes at least one terminal sensor, at least one terminal controller and at least one side server; the terminal sensor, the terminal controller and the side server in the train are communicatively connected in sequence; the side servers of each train are communicatively connected to each other and the side servers of each train are communicatively connected to the cloud server on the ground; the train control monitoring system in the train is communicatively connected to the side server; The deployment locations of the ground cloud server include the monitoring center, the owner group company and the OEM operation and maintenance center; the terminal sensor is deployed at the components to be monitored on the train to collect data of the components to be monitored; The end sensors include acceleration sensors, industrial cameras, thermal imaging cameras, optical fiber sensors, ultraviolet sensors, position sensors and dynamic measurement compensation sensors, which are used to collect contact network pull-out values, conductor heights, contact line slopes, wear amounts, arcing photos, video images, pressures, hard points, temperatures, currents and positioning information data; The end controller is used to obtain the data of the component to be monitored collected by the end sensor, and perform edge computing processing on the data of the component to be monitored through the stored machine learning algorithm model, and store the feature data obtained by the edge computing processing in the edge server; The edge server is used to receive the work instructions of the cloud server on the ground, receive and store the feature data, and process the feature data according to the stored rule engine configuration, generate corresponding information and control instructions according to the processing results, and upload them to the cloud server on the ground at the same time; The specific method for the edge servers of each train to communicate with each other is: when the edge computing power utilization rate of the edge server of the train for the feature data exceeds a set threshold, it communicates with the edge servers of each train within a set distance; The train control monitoring system is connected to the edge server, and the train control monitoring system is installed in the cab. The train control monitoring system is used to receive corresponding information and control instructions generated by the edge server, and control the train according to the corresponding information and control instructions; The cloud server is used to obtain information and control instructions uploaded by the edge server, send the corresponding machine learning algorithm model to the end controller through the edge server, and update the rule engine configuration of the edge server; The cloud server is also used to obtain end sensor data in the edge server to allow the machine learning algorithm model to continue learning and optimizing.

2. A method for online monitoring of a bow and catenary based on the edge computing-based online monitoring system of bow and catenary according to claim 1, characterized in that: The following steps are involved: S1: Use the cloud server to update the rule engine configuration of the edge server, and send the corresponding machine learning algorithm model to the end controller of each train through the edge server; S2: When the train is running, the end sensors of the train are used to collect the data of the components to be monitored and send them to the end controller; S3: Using the end controller to receive the data of the monitored component from the end sensor, and using the machine learning algorithm model to perform edge computing on the data of the monitored component, and storing the feature data obtained by the edge computing in the edge server; S4: Receive feature data using the edge server, process the feature data according to the updated rule engine configuration, and send corresponding information and control instructions to the train control monitoring system based on the processing results, and upload them to the cloud server on the ground at the same time; S5: using the train control monitoring system to obtain information and control instructions generated by the edge server, and controlling the train according to the corresponding information and control instructions; S6: Update the bow-net online monitoring system and enter the next bow-net online monitoring; In step S6, the specific method for updating is: using the cloud server to obtain the original monitored component data stored from the edge server, and optimizing the corresponding machine learning algorithm model based on the original monitored component data; sending the optimized machine learning algorithm model to the end controller through the edge server to update the machine learning algorithm model in the end controller; using the cloud server to obtain feature data from the edge server, optimizing the corresponding machine learning algorithm model based on the feature data, and sending the optimized machine learning algorithm model to the edge server to update the rule engine configuration of the edge server.

Images