An AI-based trackside safety monitoring system

Abstract

The utility model discloses a track transportation technical field's a kind of trackside safety monitoring system based on AI, including front end perception module, front end perception module is set on track monitoring section, and front end perception module is used to perceive the change of track monitoring section and forms monitoring signal.The utility model passes through the variety of monitoring signals that front end perception module to the change of track monitoring section generates, in monitoring signal transmission to front end acquisition module, and front end acquisition module is processed to the uniform processing of multiple different types monitoring signals by signal conversion module, and monitoring signal after uniform processing carries out noise reduction and merging repeated data etc by edge computing module, and further reduce the transmission of monitoring signal, and early warning signal is obtained according to the real-time processing of prediction model after monitoring signal processing, and further through early warning device timely early warning signal is sent, and timely informs maintenance personnel to carry out timely early warning maintenance.

Description

Technical Field

[0001] This utility model relates to the field of rail transit technology, specifically an AI-based trackside safety monitoring system. Background Technology

[0002] With the rapid development and increasing prevalence of urban rail transit systems, ensuring their safe, stable, and efficient operation has become crucial. This is not only related to the travel safety of passengers but also an important guarantee for smooth urban traffic and social stability. Therefore, strengthening the safety management of rail transit, improving its operational support capabilities, and achieving comprehensive and real-time monitoring of the rail transit operating status aim to identify potential safety hazards in advance and provide support for the safe operation of rail transit.

[0003] Existing trackside monitoring is mostly carried out through single monitoring methods, such as video surveillance and sensor monitoring. Although these methods can improve the efficiency of trackside safety monitoring to a certain extent, they still have problems such as limited monitoring range, insufficient data processing capabilities, and low accuracy of early warning. Especially in the complex and ever-changing rail transit environment, single monitoring methods are often unable to fully and accurately reflect the trackside safety status. Utility Model Content

[0004] The purpose of this invention is to provide an AI-based trackside safety monitoring system. By monitoring the vibration and noise data generated by subway trains during operation, it can accumulate on-site data during actual train operation, identify potential safety hazards in advance, provide support for the safe operation of rail transit, and provide data analysis and theoretical development basis for the long-term use and environmental assessment of rails.

[0005] To achieve the above objectives, this utility model provides the following technical solution:

[0006] An AI-based trackside safety monitoring system is used to monitor the stress state of tracks on rail transit lines. The system includes:

[0007] A front-end sensing module is installed on the track monitoring section, and the front-end sensing module is used to sense changes in the track monitoring section and generate monitoring signals.

[0008] A front-end acquisition module is electrically connected to a front-end sensing module. The front-end acquisition module receives monitoring signals and includes a signal conversion module, an edge computing module, and an early warning model. The signal conversion module is electrically connected to the edge computing module and is used to monitor signal conversion. The edge computing module is used to process the converted monitoring signals. The early warning model is used to monitor and generate early warning signals. The early warning model is electrically connected to an early warning device, and the early warning device generates an early warning action based on the early warning signal.

[0009] A communication module, which is electrically connected to the front-end acquisition module, is used for information communication.

[0010] The server is electrically connected to the communication module and is used for intelligent AI data processing.

[0011] As a further embodiment of this utility model: the track monitoring section includes a track bed base, a floating slab, and a rail;

[0012] The front-end sensing module includes a displacement sensor, an acceleration sensor, a pressure sensor, a strain gauge, a temperature sensor, a humidity sensor, a camera, and a microphone. The displacement sensor is installed on one side of the track bed base and is used to monitor the horizontal and vertical displacement of the track bed base. The acceleration sensor is installed on the track bed and the rail and is used to monitor the acceleration of the track bed and the rail. The pressure sensor is installed below the floating slab and is used to monitor the pressure changes borne by the floating slab. The strain gauge is installed on both the inner and outer sides of the rail and is used to monitor the rail strain. The temperature sensor and the humidity sensor are both installed on one side of the rail and are used to monitor humidity and temperature, respectively. The camera is installed on one side of the rail and is used to record video data of the train. The microphone is installed on one side of the rail and is used to record noise signals.

[0013] The displacement sensor, acceleration sensor, pressure sensor, strain gauge, temperature sensor, humidity sensor, camera, and microphone all generate monitoring signals.

[0014] As a further embodiment of this utility model: the displacement sensor is a non-contact laser displacement sensor, and the laser displacement sensor is installed on one side of the track bed base.

[0015] As a further embodiment of this utility model: multiple sets of springs are provided under the floating plate, and a pressure sensor is installed between one end of the spring and the floating plate.

[0016] As a further embodiment of this utility model: the signal conversion module receives the monitoring signal, the monitoring signal is transmitted to the edge computing module, and the edge computing module is used for intelligent processing of the monitoring signal.

[0017] As a further aspect of this utility model: the monitoring model receives historical data transmitted by the communication module and trains and continuously updates and optimizes it based on AI algorithms. The trained monitoring model is deployed on the front-end acquisition module through the communication module to form an early warning model. The early warning model receives the converted real-time monitoring signal and forms an early warning signal.

[0018] As a further embodiment of this utility model: the warning device includes a warning light, a sound alarm, and a display screen. The warning model generates a warning signal to control the warning light to flash, the sound alarm to emit an alarm sound, and the display screen to display warning information, so that staff can promptly detect and handle safety hazards.

[0019] Compared with the prior art, the beneficial effects of this utility model are:

[0020] 1. In this utility model, the front-end sensing module generates various monitoring signals from changes in the track monitoring section. These signals are then transmitted to the front-end acquisition module, which processes them uniformly using a signal conversion module. The processed signals are then denoised and have duplicate data merged by an edge computing module, thereby reducing the amount of data transmitted. The prediction model generates an early warning signal based on the real-time processed monitoring signal, which is then promptly issued by an early warning device to notify maintenance personnel for timely maintenance.

[0021] 2. In this utility model, a monitoring model is obtained by training a baseline model on a server using historical data. The monitoring model is continuously trained and its parameters are optimized based on the received data. The real-time optimized monitoring model is transmitted to the front-end acquisition module through the communication module. The front-end acquisition module deploys the optimized detection model to form an updated early warning model, thereby improving the accuracy and timeliness of trackside monitoring. Through the continuously optimized monitoring model, the system can more accurately identify abnormal situations, reduce false alarms and missed alarms, and provide more reliable protection for subway operation. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the system structure of this utility model.

[0023] In the diagram: 1. Camera; 2. Displacement sensor; 3. Accelerometer; 4. Pressure sensor; 5. Strain gauge; 6. Temperature sensor; 7. Humidity sensor; 8. Microphone; 9. Front-end acquisition module; 10. Server; 11. Communication module; 12. Warning device. Detailed Implementation

[0024] The technical solutions of the present utility model will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of the present utility model, and not all embodiments. Based on the embodiments of the present utility model, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of the present utility model.

[0025] Example:

[0026] Please see Figure 1 In this embodiment of the invention, an AI-based trackside safety monitoring system is applied to monitor the stress state of rail transit lines. The system includes:

[0027] The front-end sensing module is installed on the track monitoring section. The front-end sensing module is used to sense changes in the track monitoring section and generate monitoring signals.

[0028] The front-end acquisition module 9 is electrically connected to the front-end sensing module. The front-end acquisition module 9 receives monitoring signals and includes a signal conversion module, an edge computing module, and an early warning model. The signal conversion module is electrically connected to the edge computing module and is used to convert the monitoring signals. The edge computing module is used to process the converted monitoring signals. The early warning model is used to monitor and issue early warnings and generate early warning signals. The early warning model is electrically connected to the early warning device 12, and the early warning device 12 generates early warning actions based on the early warning signals.

[0029] Communication module 11 is electrically connected to front-end acquisition module 9 and is used for information communication.

[0030] Server 10 is electrically connected to communication module 11 and is used for intelligent AI data processing.

[0031] Specifically, this system uses a front-end sensing module to detect various monitoring signals generated by changes in the track monitoring section. These signals are then transmitted to the front-end acquisition module 9, which uses a signal conversion module to process the different types of monitoring signals in a unified manner. The processed monitoring signals are then processed by an edge computing module to reduce noise and merge duplicate data, thereby reducing the amount of monitoring signal transmission. The prediction model generates an early warning signal based on the real-time processed monitoring signal, which is then promptly issued by the early warning device 12 to notify maintenance personnel for timely early warning and maintenance.

[0032] The front-end sensing module is an embedded system based on the ARM architecture, running a real-time Linux operating system. It features a built-in CPU+FPGA dual-processing capability, making it a robust, wide-temperature-range, high-processing-capability, and customizable front-end data acquisition unit. The FPGA (Field-Programmable Gate Array) is responsible for data preprocessing and trigger judgment. It rapidly processes monitored signals to extract key feature information, such as vibration frequency, peak pressure, and strain amplitude. Simultaneously, the early warning model determines whether to trigger an abnormal alarm based on preset threshold conditions. For example, when the vibration amplitude exceeds the trigger threshold, it indicates an approaching train and activates the data acquisition function; or when the real-time values ​​of certain sensors reach a dangerous threshold, it promptly sends an early warning signal to the early warning device 12 and issues a timely warning.

[0033] RTOS and Embedded CPU: As the core unit of the front-end acquisition device, it is responsible for controlling the FPGA, acquiring real-time acquisition data, processing, storing and transmitting it, including sending the processed data to the server 10 through the network interface and storing the data in the local file system in the form of files.

[0034] The front-end acquisition module 9 adopts an embedded CPU + FPGA + acquisition module architecture. Leveraging the powerful parallel processing capabilities of the FPGA for data preprocessing and real-time trigger judgment, it significantly shortens data processing time and achieves rapid response. Combined with a lightweight AI algorithm deployed on RT (Real-Time Operating System), it can perform preliminary intelligent analysis of the track status at the front end, promptly detect anomalies and issue warnings. Compared to traditional systems that transmit large amounts of data to the back end for processing, this greatly improves the timeliness of warnings and the real-time performance of the system.

[0035] Furthermore, the monitoring signals are transmitted via wired or wireless data through the communication module 11. The transmission of monitoring signals to the server 10 includes two methods: first, utilizing the public network of the operator and employing secure and reliable communication protocols, such as 4G / 5G networks, to achieve remote and rapid data transmission, suitable for areas where access to the subway system's internal network is not possible; second, using the subway system's internal network, such as dedicated wired or wireless communication networks (e.g., Wi-Fi, LTE-M), to ensure the stability and security of data transmission, meeting the requirements for data transmission reliability and confidentiality in the subway operating environment.

[0036] Furthermore, the server 10 obtains a monitoring model based on historical data and a baseline model. The monitoring model is continuously trained and optimized based on the received data. The real-time optimized monitoring model is then transmitted to the front-end acquisition module 9 via the communication module 11. The front-end acquisition module 9 deploys the optimized detection model to form an updated early warning model, thereby improving the accuracy and timeliness of trackside monitoring. Through the continuously optimized monitoring model, the system can more accurately identify abnormal situations, reduce false alarms and missed alarms, and provide more reliable protection for subway operation. The baseline model refers to a reference model built based on the normal operating status data of the track, which is used as a benchmark for judging whether the track is abnormal. Its core is to determine the normal fluctuation range and characteristic laws of various track parameters (such as vibration, pressure, strain, temperature, etc.) by analyzing historical or real-time normal data, thereby providing a comparison standard for real-time monitoring data.

[0037] An AI-powered monitoring model runs on server 10, performing in-depth data mining and trend analysis based on long-term accumulated monitoring data. Through the continuous learning process of the AI ​​algorithm, the track normal state model is updated and optimized, enabling not only accurate judgment of current anomalies but also prediction of future track operation trends, early detection of potential safety risks, and strong support for preventative maintenance. This trend analysis function based on big data and AI technology is significantly innovative and leading among similar systems, contributing to improved safety and reliability of rail transit operations and reduced maintenance costs.

[0038] Employing an autoencoder algorithm, it can concatenate multi-sensor features (such as vibration frequency, average pressure, and peak strain) into an input vector. Through the hidden layers of a neural network, it automatically learns the inherent relationships between these features (without requiring manually defined rules). This approach better reflects the overall state of the monitored object than monitoring with a single sensor independently.

[0039] Preferred, such as Figure 1 As shown, the track monitoring section includes the track bed base, floating slab, and rails;

[0040] The front-end sensing module includes a displacement sensor 2, an acceleration sensor 3, a pressure sensor 4, a strain gauge 5, a temperature sensor 6, a humidity sensor 7, a camera 1, and a microphone 8. The displacement sensor 2 is installed on one side of the track bed base and is used to monitor the horizontal and vertical displacement of the track bed base. The acceleration sensor 3 is installed on the track bed and the rail and is used to monitor the acceleration of the track bed and the rail. The pressure sensor 4 is installed below the floating slab and is used to monitor the pressure changes borne by the floating slab. The strain gauge 5 is installed on both the inner and outer sides of the rail and is used to monitor the rail strain. The temperature sensor 6 and the humidity sensor 7 are both installed on one side of the rail and are used for humidity and temperature monitoring, respectively. The camera 1 is installed on one side of the rail and is used to record video data of the train. The microphone 8 is installed on one side of the rail and is used to record noise signals.

[0041] Displacement sensor 2, acceleration sensor 3, pressure sensor 4, strain gauge 5, temperature sensor 6, humidity sensor 7, camera 1, and microphone 8 all generate monitoring signals.

[0042] Specifically, the monitoring signals undergo preliminary processing through the signal conversion module within the front-end sensing module, including amplification, filtering, and analog-to-digital conversion, to ensure data accuracy and stability. An array of various types of sensors, such as vibration sensors, floating slab pressure sensors, rail strain sensors, and temperature sensors, is used to achieve comprehensive, multi-parameter monitoring of the track. Compared to traditional monitoring systems using single or a few types of sensors, this system can more comprehensively and accurately reflect the actual operating status of the track, improving monitoring reliability and accuracy, and effectively reducing missed detections of safety hazards. The processed monitoring signals are transmitted to the edge computing module, which can receive and store this multi-dimensional monitoring data in real time. Advanced algorithms are used for in-depth analysis and processing of the data. Combining video data recorded by camera 1 and noise signals recorded by microphone 8, the system can further analyze train operating status and changes in the surrounding environment, providing comprehensive and reliable protection for the safe operation of rail transit.

[0043] Preferably (not shown in the figure), the displacement sensor 2 is a non-contact laser displacement sensor 2, which is installed on one side of the track bed base.

[0044] Specifically, the laser displacement sensor 2 has the advantages of high precision, high speed, and non-contact operation, which can effectively avoid the wear and errors that may be caused by traditional contact sensors. The laser displacement sensor 2 emits a laser beam and receives the reflected light signal. It calculates the displacement change of the target object based on the time difference or phase difference of the light signal. In the rail transit monitoring system, the laser displacement sensor 2 can monitor the minute displacement of the track bed base in real time, providing important data support for assessing the stability and safety of the track structure. In addition, the non-contact design also enables it to operate stably under harsh environmental conditions, such as high temperature, humidity, or dusty environments, thereby ensuring the continuity and reliability of monitoring data.

[0045] Preferably (not shown in the figure), multiple sets of springs are installed under the floating plate, and a pressure sensor 4 is installed between one end of the spring and the floating plate.

[0046] Specifically, pressure sensor 4 is used to monitor the pressure changes on the spring in real time, reflecting the dynamic response of the floating plate during train operation. By accurately measuring the pressure changes of the spring, the system can indirectly assess the track vibration and the smoothness of train operation. The collection of multiple sets of data from multiple springs and pressure sensors 4 can further improve the accuracy and reliability of track monitoring. The data from each set of springs and pressure sensors 4 can be cross-checked, reducing inaccurate data caused by single sensor failure or error. The collection of multiple sets of data can also provide more comprehensive track status information, helping the system to more accurately analyze the track vibration characteristics, assess the smoothness of train operation, and predict potential track defects, thus providing strong support for the safe operation of rail transit.

[0047] Preferred, such as Figure 1 As shown, the signal conversion module receives the monitoring signal, and the monitoring signal is transmitted to the edge computing module, which is used to intelligently process the monitoring signal.

[0048] Preferred, such as Figure 1 As shown, the monitoring model receives historical data transmitted by the communication module 11 and trains and continuously updates and optimizes it based on AI algorithms. The trained monitoring model is deployed on the front-end acquisition module 9 through the communication module 11 and forms an early warning model. The early warning model receives the converted real-time monitoring signal and forms an early warning signal.

[0049] Specifically, the edge computing capabilities of the edge computing module in front-end acquisition module 9 are used to embed lightweight AI algorithms, enabling real-time identification and prediction of early track faults. The early warning algorithm can be implemented through a combination of real-time data recording and front-end unsupervised learning algorithms. The technical approach is as follows: using normal data from the initial system operation phase as a baseline, a "normal state model" is dynamically constructed through unsupervised learning, and anomalies are identified based on the deviation between real-time data and the model. The specific implementation process is as follows:

[0050] 1. Historical data collection and accumulation stage

[0051] The front-end acquisition module 9 records various sensor data (including vibration, pressure, strain, temperature, etc.) in real time as the train passes by. During the acquisition process, it uses FPGA and CPU to calculate feature values ​​of different sensor data, such as peak value, spectrum, octave band, etc. The raw data and feature value data are labeled (e.g., peak time, winter / summer, vehicle type, etc.) as training samples.

[0052] Displacement signal Extreme values, effective values Spectrum Vibration signal Extreme values, effective values Octave strain signal Extreme values, effective values none noise signal Extreme values, effective values Octave pressure signal Extreme values, effective values Spectrum Temperature and humidity signal mean none

[0053] 2. Selection of baseline model

[0054] A baseline model is a reference model built based on data from the normal operation of the track, used as a benchmark to determine whether the track is abnormal. Its core is to analyze historical or real-time collected normal data to determine the normal fluctuation range and characteristic patterns of various track parameters (such as vibration, pressure, strain, temperature, etc.), thereby providing a comparison standard for real-time monitoring data.

[0055] This scheme employs an autoencoder algorithm, which can concatenate multi-sensor features (such as vibration frequency, average pressure, and peak strain) into an input vector. Through the hidden layer of a neural network, it automatically learns the inherent correlation between these features (without requiring manual rule definition), thus reflecting the overall state of the monitored object better than independent monitoring by a single sensor.

[0056] 3. Model training and deployment

[0057] The training process of autoencoders (especially the iterative optimization of model parameters) requires significant computing power (such as matrix operations and gradient descent calculations). A monitoring model was trained on server 10 using effective data accumulated over a period of time.

[0058] The trained model is deployed on the front-end data acquisition device. Using the data obtained from the FPGA and CPU, the reconstruction error is calculated. If the error exceeds the threshold, an early warning is issued.

[0059] The newly accumulated data is used regularly to retrain the monitoring model on server 10, and the optimized model parameters are updated to the early warning model of the front-end acquisition module 9.

[0060] This achieves a division of labor where "the front end is responsible for real-time inference, and the server 10 end is responsible for model training and updating," which is more in line with the architecture design of "front-end intelligent processing + back-end deep optimization," satisfying both the need for lightweight design and ensuring the accuracy of early warnings.

[0061] Server 10 can build a deep analysis model based on the feature data and historical full data (including historical normal data and subsequent abnormal data) uploaded by front-end acquisition module 9, achieving more accurate fault location, trend prediction, and maintenance decision support. This system uses the Transformer deep learning algorithm, which has the ability to fuse multi-sensor data and predict time-series trends, enabling extended fault location and decision support. It also boasts high economic efficiency and feasibility in engineering implementation.

[0062] Fault marking

[0063] Collect data reported from the front end, combine it with manual inspection and maintenance records (such as "loose rail joints" corresponding to "increased high-frequency vibration components + abrupt strain change"), label the fault type and associated features, realize supervised learning, and improve the performance of AI algorithms.

[0064] 2. Full life cycle trend prediction

[0065] Remaining life assessment:

[0066] Using the Transformer model to perform time-series modeling on historical data, inputting current track state characteristics (such as cumulative strain value, temperature cycle number) to predict the remaining safe operating mileage of key components (such as rails, floating slabs) (e.g., "the rails may develop fatigue cracks after 1500 train cycles").

[0067] Maintenance priority sorting:

[0068] Combining the scope of the fault impact (e.g., main line track vs. branch line track) and the predicted fault time, a reinforcement learning algorithm is used to generate maintenance work order priorities (e.g., "handle the section with uneven pressure on the floating slab within 3 days, and check the section with abnormal vibration within 7 days").

[0069] Whether to introduce popular large models, given that early warning and maintenance functions have already been implemented using the Transformer deep learning algorithm, requires comprehensive consideration from multiple perspectives. Large models do not simply replace Transformers, but rather play an expanding and supplementary role in the system, enhancing the breadth and depth of its functionality.

[0070] Preferably (not shown in the figure), the warning device 12 includes a warning light, a sound alarm, and a display screen. The warning model generates a warning signal to control the warning light to flash, the sound alarm to sound an alarm, and the display screen to display warning information so that staff can promptly detect and deal with safety hazards.

[0071] Specifically, the design of the warning device 12 emphasizes human-computer interaction. The warning lights use striking colors and a high-frequency flashing pattern, the audible alarm provides a clear and easily identifiable sound, and the display screen uses a high-resolution and intuitive information display method to ensure effective transmission of warning information in various lighting and noise environments. In addition, the warning system also has a historical warning record query function, facilitating staff to trace and analyze warning events and further optimize maintenance strategies.

[0072] The above description is only a preferred embodiment of the present utility model, but the protection scope of the present utility model is not limited thereto. Any equivalent substitutions or changes made by those skilled in the art within the technical scope disclosed in the present utility model, based on the technical solution and the inventive concept of the present utility model, should be included within the protection scope of the present utility model.

Claims

1. An AI-based trackside safety monitoring system, characterized by, The system, used for monitoring the stress state of rail transit lines, includes: A front-end sensing module is installed on the track monitoring section, and the front-end sensing module is used to sense changes in the track monitoring section and generate monitoring signals. A front-end acquisition module is electrically connected to a front-end sensing module. The front-end acquisition module receives monitoring signals and includes a signal conversion module, an edge computing module, and an early warning model. The signal conversion module is electrically connected to the edge computing module and is used to monitor signal conversion. The edge computing module is used to process the converted monitoring signals. The early warning model is used to monitor and generate early warning signals. The early warning model is electrically connected to an early warning device, and the early warning device generates an early warning action based on the early warning signal. A communication module, which is electrically connected to the front-end acquisition module, is used for information communication. The server is electrically connected to the communication module and is used for intelligent AI data processing. 2.The AI-based trackside safety monitoring system of claim 1, wherein: The track monitoring section includes the track bed base, floating slab, and rails; The front-end sensing module includes a displacement sensor, an acceleration sensor, a pressure sensor, a strain gauge, a temperature sensor, a humidity sensor, a camera, and a microphone. The displacement sensor is installed on one side of the track bed base and is used to monitor the horizontal and vertical displacement of the track bed base. The acceleration sensor is installed on the track bed and the rail and is used to monitor the acceleration of the track bed and the rail. The pressure sensor is installed below the floating slab and is used to monitor the pressure changes borne by the floating slab. The strain gauge is installed on both the inner and outer sides of the rail and is used to monitor the rail strain. The temperature sensor and the humidity sensor are both installed on one side of the rail and are used to monitor humidity and temperature, respectively. The camera is installed on one side of the rail and is used to record video data of the train. The microphone is installed on one side of the rail and is used to record noise signals. The displacement sensor, acceleration sensor, pressure sensor, strain gauge, temperature sensor, humidity sensor, camera, and microphone all generate monitoring signals. 3.The AI-based trackside safety monitoring system of claim 2, wherein: The displacement sensor is a non-contact laser displacement sensor, which is installed on one side of the track bed base.

4. The AI-based trackside safety monitoring system of claim 3, wherein: Multiple sets of springs are installed under the floating plate, and a pressure sensor is installed between one end of the spring and the floating plate.

5. The AI-based trackside safety monitoring system of claim 4, wherein: The signal conversion module receives the monitoring signal, and the monitoring signal is transmitted to the edge computing module, which is used to intelligently process the monitoring signal. 6.The AI-based trackside safety monitoring system of claim 5, wherein: The monitoring model receives historical data transmitted by the communication module, trains and continuously updates and optimizes it based on AI algorithms, and the trained monitoring model is deployed on the front-end acquisition module through the communication module to form an early warning model. The early warning model receives the converted real-time monitoring signal and forms an early warning signal.

7. The AI-based trackside safety monitoring system of claim 6, wherein: The early warning device comprises an early warning light, a sound alarm and a display screen, and the early warning model controls the early warning light to flicker, the sound alarm to emit an alarm sound and the display screen to display early warning information, so that the staff can discover and handle the safety hidden danger in time.

Images