Intelligent monitoring and early warning platform for whole process of tunnel lining construction

Abstract

The invention discloses a tunnel lining construction full-process intelligent monitoring and early warning platform, which relates to the technical field of tunnel construction intelligent monitoring and comprises a field sensing unit, an edge computing unit and a central management center. According to the invention, the field sensing unit synchronously acquires multi-source data template stress, template deformation, template displacement, surrounding rock and environment parameters and personnel and equipment positioning, and displacement sensors are alternately arranged to realize temperature drift correction, so that the data precision is improved; through fusion of a dual-spectrum camera and an ultra-wideband label, sub-meter personnel positioning precision and safety protection article identification are realized in a low-illumination environment, a full-dimension sensing network covering physical states, environments and behaviors is formed, and an edge computing unit adopts an embedded chip to provide computing power. And multi-modal feature fusion and physical consistency constraint deep learning reasoning are executed, an online change point detection mechanism is introduced to dynamically adjust an early warning threshold value, and different geological conditions and construction stages are adapted.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent monitoring of tunnel construction, and in particular to an intelligent monitoring and early warning platform for the entire process of tunnel lining construction. Background Art

[0002] Intelligent monitoring technology for tunnel construction refers to a systematic technology that uses modern information technologies such as the Internet of Things, sensor networks, artificial intelligence, and big data analysis to conduct real-time, automated, and remote monitoring and risk warning of the entire tunnel construction process. Its core goal is to improve construction safety, quality control, and efficiency optimization through intelligent means, avoid reliance on human experience and judgment, and reduce construction risks.

[0003] Published patent: A dynamic real-time monitoring system for tunnel construction (publication number: CN114841516A), including at least one camera, a data processing module, a communication module, a storage module and a display module. The data processing module is used to receive image information and analyze changes in image information of the target area using the inter-frame difference method. It also establishes a vertical reference line in the image information. If there is an object moving downward along the reference line in the image information, the outline information of the object is extracted. When a person is about to pass by a drop point, the data processing module sends an illumination signal to the searchlight module. Then, based on the contact points between the person and the outline information in the image information, it determines whether the object is a man-made object. If there is no contact point information, the system determines whether there are sharp points on the edge of the outline information exceeding a threshold number. If the threshold is exceeded, the object is determined to be non-man-made. If the threshold is not exceeded, the drop information is counted. This completes the safety assessment of the construction environment.

[0004] The information source is single and cannot monitor the key physical quantities of tunnel construction. Simple visual analysis cannot predict structural safety risks, resulting in delayed warnings. Fixed thresholds have a high false alarm rate in high-humidity and strong electromagnetic tunnel environments and cannot adapt to different shifts or geological changes. Insufficient CPU computing power leads to high delays in dangerous action recognition. The removal of dangerous areas requires manual on-site inspection and manual input of cancellation commands. The manual links drag down the handling efficiency and are prone to missing risks. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the existing technology and propose an intelligent monitoring and early warning platform for the entire process of tunnel lining construction.

[0006] In order to achieve the above-mentioned objectives, the present invention adopts the following technical solutions: an intelligent monitoring and early warning platform for the entire process of tunnel lining construction, including an on-site perception unit, an edge computing unit and a central management center; the on-site perception unit includes an axial force sensor, a displacement sensor, an environmental monitoring device, a dual-spectrum camera and an ultra-wideband tag, which is used to collect multi-source data during the tunnel lining construction process, including template force, template deformation, template displacement, surrounding rock and environmental parameters, and personnel and equipment positioning information; the edge computing unit receives the data of the on-site perception unit through a dual communication link, performs multimodal feature fusion, physical consistency constraint deep learning reasoning, online change point detection, skeleton posture recognition and dynamic Bayesian reasoning engine, and generates risk warning signals and disposal suggestions; the central management center receives the processing results of the edge computing unit, connects with the BIM / CIM model in real time, constructs a digital twin visualization scene, and stores, analyzes and compares the data horizontally across sections; the on-site perception unit, the edge computing unit and the central management center are connected through a dual communication link, which supports automatic switching to the backup link when the main link fails, thereby achieving high-reliability data transmission.

[0007] As a further description of the above technical solution: The axial force sensor is a fiber Bragg grating stress gauge, which is arranged in the top wedge or support rod of the key stress-bearing part of the lining formwork to monitor the axial force of the formwork; the displacement sensor is arranged in the middle of the lining formwork and the longitudinal side beam, including a fiber optic displacement meter and a laser displacement meter, which is used to track the deformation of the formwork. The data of the two are corrected in real time to ensure accuracy. The environmental monitoring equipment is used to obtain surrounding rock and environmental information, including a three-dimensional laser scanner, temperature and humidity sensors, air pressure sensors and gas concentration probes, which are used to collect surrounding rock and environmental data. The dual-spectrum camera and ultra-wideband tag are used for personnel and equipment positioning and behavior recognition, and can identify the wearing of safety protection equipment by personnel in low-light environments, and can be integrated with the ultra-wideband tag to achieve sub-meter personnel positioning accuracy.

[0008] As a further description of the above technical solution: The edge computing unit has a built-in embedded chip that supports real-time reasoning of multiple video streams and sensor data fusion algorithms. It is equipped with a deep network model that integrates template stress, displacement and video features, introduces physical consistency constraints, and uses an online change point detection mechanism to dynamically adjust the warning threshold to adapt to different geological conditions and construction stages. Based on the skeletal posture recognition network and the temporal motion analysis framework, it identifies dangerous actions, and the recognition results are input into the dynamic Bayesian inference engine to generate three levels of risk scores from safety, warning to danger, and give corresponding disposal suggestions. When the risk score is higher than the threshold, the emergency broadcast and warning light strip are linked, and the responsible person information is pushed to the mobile terminal.

[0009] As a further description of the above technical solution: The central management center receives processed data from the edge computing unit, maps the laser point cloud, sensor timing curve, risk score, recognition results, warning events and key raw data fragments to the BIM / CIM three-dimensional model, stores data by project dimension and supports horizontal comparative analysis across sections, conducts horizontal comparative analysis of different tunnel sections, automatically generates weekly or monthly reports, and outputs PDF or interactive 3D links for supervisor review; sets a 30-day cycle coverage and automatic desensitization and deletion mechanism for the data collected by the on-site perception unit, and stores the processing results of the edge computing unit, automatically generated weekly and monthly reports, and data in the data lake in the central management center and the cloud in the primary and secondary backup, and preserves them for a long time according to project supervision requirements.

[0010] As a further description of the above technical solution: The dual communication links integrate 5G and Gigabit Ethernet interfaces. When the main link fails, it automatically switches to the 4G backup link, and retransmits the locally cached data after the link is restored.

[0011] As a further description of the above technical solution: The edge computing unit and field perception unit are equipped with dual power supplies and hard disk mirroring. When external power, network or single-point hardware fails, data can be cached locally and retransmitted after the link is restored.

[0012] As a further description of the above technical solution: Dangerous actions include the top support not being locked, the formwork hitting the pole, staying in the danger zone for too long, not wearing safety protection equipment, illegally crossing the warning line, and the identification delay is less than 0.2 seconds.

[0013] As a further description of the above technical solution: The three-level risk score ranges from safe to dangerous. Ten minutes of continuous observation is required for the warning to be lifted, and the data of the three channels of stress, displacement and video must return to normal.

[0014] As a further description of the above technical solution: The data stored in the data lake includes template stress, displacement, surrounding rock and environmental data collected by the on-site sensing unit, fragments of personnel violations, lists of responsible persons and full-process records of warning events, risk scores, disposal recommendations and automatically generated weekly and monthly reports after processing by the edge computing unit.

[0015] The present invention has the following beneficial effects: In the present invention, first, the on-site perception unit synchronously collects multi-source data template force, template deformation, template displacement, surrounding rock and environmental parameters, and personnel and equipment positioning, and uses alternating displacement sensors to realize temperature drift correction to improve data accuracy. Through the fusion of dual-spectral cameras and ultra-wideband tags, sub-meter personnel positioning accuracy and safety protection equipment identification are achieved in low-light environments, forming a full-dimensional perception network covering physical status, environment and behavior. The edge computing unit uses embedded chips to provide computing power, performs multimodal feature fusion and physical consistency constraint deep learning reasoning, introduces an online change point detection mechanism to dynamically adjust the warning threshold, adapt to different geological conditions and construction stages, and realizes dangerous action recognition with a delay of less than or equal to 0.2 seconds through the skeleton posture recognition network and the time sequence motion analysis framework. The recognition results are input into the dynamic Bayesian reasoning engine to generate three-level risk scores and disposal recommendations for safety, warning, and danger, and based on the scores, emergency broadcasts, warning lights, and pumping suspension and other active intervention measures are linked. It supports dual power supplies and hard disk mirroring mechanisms to ensure that local data cache is continuously flowing in the event of external power or hardware failure.

[0016] In the present invention, the central management center receives edge data, maps laser point clouds, sensor timing curves, risk scores, warning events and key raw data fragments to BIM / CIM models in real time, builds digital twin visualization scenes, dynamically displays template force gradients, displacement arrows and risk heat maps, deposits data in the data lake according to project dimensions, supports cross-section horizontal comparative analysis, and automatically generates PDF or interactive 3D link weekly / monthly reports for supervision review. The on-site perception raw data is covered in a 30-day cycle and automatically desensitized and deleted, while the processing results, weekly and monthly reports and data lake contents are synchronized using the master backup. It is stored in the central management center and the cloud in real time to meet the requirements of long-term engineering supervision. The entire platform ensures high reliability through dual communication links. When the main link fails, it automatically switches to the backup link. After the link is restored, the local cached data is re-transmitted to improve adaptability to extreme environments. Through the unified directory and unified metadata of the data lake, a full-link closed-loop communication, power supply and storage dual redundancy design is realized from original collection to supervision archiving, supplemented by local caching and re-transmission mechanisms to ensure data integrity and risk traceability in extreme environments, thereby improving the real-time perception depth, risk response speed and visual decision-making level of the tunnel lining construction site. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a platform architecture diagram of the present invention. DETAILED DESCRIPTION

[0018] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only 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 making creative efforts are within the scope of protection of the present invention.

[0019] Reference Figure 1 , the present invention provides an embodiment: an intelligent monitoring and early warning platform for the entire process of tunnel lining construction, including an on-site perception unit, an edge computing unit and a central management center; the on-site perception unit includes an axial force sensor, a displacement sensor, an environmental monitoring device, a dual-spectrum camera and an ultra-wideband tag, which is used to collect multi-source data during the tunnel lining construction process, including template force, template deformation, template displacement, surrounding rock and environmental parameters, and personnel and equipment positioning information; the edge computing unit receives data from the on-site perception unit through a dual communication link, performs multimodal feature fusion, physical consistency constraint deep learning reasoning, online change point detection, skeleton posture recognition and a dynamic Bayesian reasoning engine, and generates risk warning signals and disposal suggestions; the central management center receives the processing results of the edge computing unit, connects with the BIM / CIM model in real time, constructs a digital twin visualization scene, and stores, analyzes and compares the data horizontally across sections; the on-site perception unit, the edge computing unit and the central management center are connected through a dual communication link, which supports automatic switching to the backup link when the main link fails, thereby achieving high-reliability data transmission.

[0020] The axial force sensor is a fiber Bragg grating stress gauge, which is arranged in the top wedge or support rod of the key stress-bearing part of the lining formwork to monitor the axial force of the formwork; the displacement sensor is arranged in the middle of the lining formwork and the longitudinal side beam, including a fiber optic displacement meter and a laser displacement meter, which is used to track the deformation of the formwork. The data of the two are corrected in real time to ensure accuracy. The environmental monitoring equipment is used to obtain surrounding rock and environmental information, including a three-dimensional laser scanner, temperature and humidity sensors, air pressure sensors and gas concentration probes, which are used to collect surrounding rock and environmental data. The dual-spectrum camera and ultra-wideband tag are used for personnel and equipment positioning and behavior recognition, and can identify the wearing of safety protection equipment by personnel in low-light environments, and can be integrated with the ultra-wideband tag to achieve sub-meter personnel positioning accuracy.

[0021] The edge computing unit has a built-in embedded chip that supports real-time reasoning of multiple video streams and sensor data fusion algorithms. It is equipped with a deep network model that integrates template stress, displacement and video features, introduces physical consistency constraints, and uses an online change point detection mechanism to dynamically adjust the warning threshold to adapt to different geological conditions and construction stages. Based on the skeletal posture recognition network and the temporal motion analysis framework, it identifies dangerous actions, and the recognition results are input into the dynamic Bayesian inference engine to generate three levels of risk scores from safety, warning to danger, and give corresponding disposal suggestions. When the risk score is higher than the threshold, the emergency broadcast and warning light strip are linked, and the responsible person information is pushed to the mobile terminal.

[0022] The central management center receives processed data from the edge computing unit, maps the laser point cloud, sensor timing curve, risk score, recognition results, warning events and key raw data fragments to the BIM / CIM three-dimensional model, stores data by project dimension and supports horizontal comparative analysis across sections, conducts horizontal comparative analysis of different tunnel sections, automatically generates weekly or monthly reports, and outputs PDF or interactive 3D links for supervisor review; sets a 30-day cycle coverage and automatic desensitization and deletion mechanism for the data collected by the on-site perception unit, and stores the processing results of the edge computing unit, automatically generated weekly and monthly reports, and data in the data lake in the central management center and the cloud in the primary and secondary backup, and preserves them for a long time according to project supervision requirements.

[0023] The dual communication links integrate 5G and Gigabit Ethernet interfaces. When the main link fails, it automatically switches to the 4G backup link, and retransmits the locally cached data after the link is restored.

[0024] The edge computing unit and field perception unit are equipped with dual power supplies and hard disk mirroring. When external power, network or single-point hardware fails, data can be cached locally and retransmitted after the link is restored.

[0025] Dangerous actions include the top support not being locked, the formwork hitting the pole, staying in the danger zone for too long, not wearing safety protection equipment, illegally crossing the warning line, and the identification delay is less than 0.2 seconds.

[0026] The three-level risk score ranges from safe to dangerous. Ten minutes of continuous observation is required for the warning to be lifted, and the data of the three channels of stress, displacement and video must return to normal.

[0027] The data stored in the data lake includes template stress, displacement, surrounding rock and environmental data collected by the on-site sensing unit, fragments of personnel violations, lists of responsible persons and full-process records of warning events, risk scores, disposal recommendations and automatically generated weekly and monthly reports after processing by the edge computing unit.

[0028] Example 1: A fiber Bragg grating stress meter is embedded in the middle of each support rod of the lining formwork to continuously monitor the axial force of the formwork; fiber optic displacement meters and laser displacement meters are installed alternately on the longitudinal side beams and the middle of the formwork to track the deformation of the formwork with millimeter-level accuracy, and the temperature drift error is corrected in real time using dual-sensor data; surrounding rock and environmental information is obtained by three-dimensional laser scanners, temperature and humidity sensors, air pressure sensors, and gas concentration probes; workers and equipment are positioned using a fusion of dual-spectral cameras and ultra-wideband tags to achieve sub-meter positioning and identify the wearing of safety protective equipment. All sensors sample at a frequency of 10Hz, and the integrated timestamp is written into the time series database.

[0029] Each edge computing unit is equipped with a Jetson Orin NX embedded chip, which can provide computing power of hundreds of TOPS and support real-time reasoning of multiple video streams and sensor data fusion algorithms; the whole machine integrates 5G, Gigabit Ethernet and Time-sensitive Ethernet interfaces, and reserves a 4G backup link. The deep network simultaneously integrates template stress, displacement and video features, and introduces physical consistency constraints in the loss function; the online change point detection mechanism adjusts the warning threshold in real time according to geological conditions and construction stages; the skeletal posture recognition network and the time-series motion analysis framework can detect dangerous actions such as unlocked top supports, template collisions, excessive stays in dangerous areas, failure to wear safety protection equipment, and illegal crossing of warning lines within a delay of less than or equal to 0.2 seconds. The recognition results are input into the dynamic Bayesian inference engine to generate a three-level risk score of safety-warning-danger and give disposal recommendations.

[0030] When the edge computing unit determines that the risk score exceeds the threshold and the formwork is in the pouring stage, the emergency broadcast and warning light strips are immediately linked and the pumping is suspended. At the same time, the list of responsible persons is pushed to the safety officer's mobile terminal. After the construction personnel complete the reinforcement, the system continues to observe for ten minutes. As long as the stress, displacement and video channels return to normal, the warning is automatically lifted and the entire process of the incident is fully recorded. The central management center receives the processing results of the edge computing unit through dual communication links, and maps the laser point cloud, sensor timing curve, risk score, warning events and key raw data fragments to the BIM / CIM model to form a digital twin visualization scene. Supervisors can view the template force gradient, displacement arrows and risk heat map in real time on the browser or mobile terminal; the system deposits the original data, processing results and analysis reports in the data lake according to the project dimension, and can make horizontal comparisons across sections and automatically generate weekly or monthly reports, and output PDF or interactive 3D links for main review.

[0031] The field perception unit, edge computing unit and central management center are interconnected through dual communication links. When the main link fails, it automatically switches to the 4G backup link, and retransmits the local cache after the link is restored. The edge computing unit and field perception unit are equipped with dual power supplies and hard disk mirroring, which can cache data locally and maintain continuous flow in the event of external power, network or single-point hardware failure.

[0032] The platform performs a 30-day cycle coverage and automatic desensitization deletion on the raw data collected by the on-site sensing unit; it adopts a master-backup simultaneous storage strategy for the edge computing unit processing results, weekly reports, monthly reports and data lake content, and preserves them for a long time in the central management center and the cloud to meet engineering supervision requirements.

[0033] Example 2: In hard rock tunnel construction sites at high altitudes and high humidity, the platform first embedded fiber Bragg grating strain gauges in the top wedges and struts of each lining formwork. Fiber optic displacement gauges and laser displacement meters were alternately placed at the ends of the longitudinal beams. A 3D laser scanner, temperature and humidity sensors, air pressure sensors, and gas concentration probes were installed around the vault to comprehensively monitor formwork stress, formwork deformation, and the surrounding rock environment. Personnel and large-scale shotcrete machinery were equipped with ultra-wideband tags, which, combined with dual-spectral cameras, enabled sub-meter positioning and PPE identification.

[0034] The edge computing unit uses Jetson Orin NX, which relies on a computing power of hundreds of TOPS to fuse template stress, displacement and video features within a frequency of ten Hz. It also combines the physical consistency constraint deep network, online change point detection and skeletal posture recognition network to perform zero-point-two-second recognition of dangerous actions such as "top support not locked" and "template hitting the rod", and inputs the results into the dynamic Bayesian inference engine to generate a three-level risk score; when the score is high-risk and the tunnel is in the pouring stage, the system links the emergency broadcast and warning light strips and suspends pumping, and then continues to observe for ten minutes. After the data of the three channels return to normal, the warning is automatically lifted.

[0035] The central management center receives processing results and key raw data fragments via a dual 5G-Gigabit Ethernet communication link. In the event of a failure, it automatically switches to the 4G backup link, and retransmits the cached data upon link recovery. Laser point clouds, sensor timing curves, and risk scores are mapped to the BIM / CIM 3D digital twin model in real time, allowing supervisors to view force gradients and displacement arrows in a browser. All data is written to the data lake by project dimension, with raw sensor data overwritten every 30 days. Weekly and monthly reports, as well as cross-section comparisons, are stored in encrypted, long-term storage.

[0036] Example 3: In densely populated urban areas with good communications, the platform's field sensing unit features sensors including axial force sensors, displacement sensors, environmental monitoring equipment, dual-spectrum cameras, and ultra-wideband tags. Fiber optic displacement meters are densely deployed for shallow soft soil layers to enhance sensitivity to instantaneous formwork deformation. The dual-spectrum cameras utilize optical cables installed in ventilation shafts within station open-cut shafts, ensuring detection of PPE wear and excessive time spent in hazardous areas even in low-light conditions.

[0037] An edge computing unit is located in the station's construction power room, connected to the metropolitan area network via Gigabit Ethernet and backed up by 5G. During overnight "fast-paced" shifts, online change-point detection automatically adjusts the template force threshold based on soil moisture fluctuations to avoid false alarms. A skeletal posture recognition network and a temporal motion analysis framework perform real-time behavioral reviews on track-track transport vehicle drivers operating in cross-operation situations. Any violations crossing the warning line are immediately notified to the responsible person via a mobile app.

[0038] A single board can provide computing power of hundreds of TOPS, and can still run multiple video streams and sensor data fusion algorithms smoothly in a narrow tunnel guide. The whole machine integrates 5G and Gigabit Ethernet interfaces, and reserves time-sensitive Ethernet ports to meet real-time requirements. When the main link is disconnected, it can automatically switch to the 4G backup link to ensure continuous data flow. The edge computing unit uses a deep network to fuse template stress, displacement and video features, and adds physical consistency constraints to the loss function, so that the model not only follows the laws of mechanics but also retains the advantages of data-driven. When the sensor or visual data shows an abnormal trend, the algorithm will introduce an online change point detection mechanism to adjust the threshold in real time to adapt to different geological conditions and construction stages. For the compliance of personnel operations, based on the skeleton posture recognition network and the time-series motion analysis framework, it can be used Dangerous actions such as unlocked top supports, template collisions, excessive stays in dangerous areas, failure to wear safety protective equipment, and illegal crossing of warning lines are identified within a delay of less than 0.2 seconds. All identification results are input into the dynamic Bayesian inference engine to generate three-level risk scores from safety, warning to danger, and corresponding handling suggestions. When the risk score is higher than the preset threshold, the platform immediately links the emergency broadcast and warning light strips in the tunnel, and pushes the work number of the person located to the mobile phone application. After the construction personnel complete the reinforcement or adjustment, the system continues to observe for ten minutes. As long as the three channels of stress, displacement and video return to normal, the warning will be automatically lifted, and the entire process from the personnel's illegal operation to the completion of the disposal will be recorded. At the same time, the abnormal sensor or visual data will be recorded and uploaded to the central management center for preservation according to the requirements of engineering supervision.

[0039] The central management center stores risk scores from multiple tunnel sections in a single data lake. This data lake is a centralized, project-based data repository in the cloud. It integrates raw data, processing results, and analysis reports from multiple tunnel sections. The data lake also uses color gradients to contrast peak stresses across template sections within the digital twin. The system automatically summarizes key indicators daily and generates a PDF weekly report with an interactive 3D link for supervisor review. All field sensor data is desensitized and deleted every 30 days. Weekly and monthly reports are stored both centrally and in the cloud for long-term backup.

[0040] The central management center receives data processed by the edge computing unit, including laser point cloud data of surrounding rocks obtained through 3D laser scanning, sensor timing curve composed of time series data of physical quantity template stress, displacement, temperature and humidity, air pressure and gas concentration, three-level risk scores from safe to dangerous generated by the edge computing unit, as well as identification results and early warning events, including identification results of personnel violations, fragments of dangerous actions and lists of responsible persons, status of early warning events, key original data fragments retained by the edge computing unit when uploading to the cloud, and real-time docking with the pre-established BIM / CIM model to form a 3D model. The data stored in the central management center include template stress, displacement data, surrounding rock and environmental data, personnel and equipment positioning information, and the entire process of key events occurring during construction collected by the on-site perception unit. They are encrypted and stored, and a 30-day cycle coverage strategy is set. They are automatically desensitized and deleted upon expiration. The edge computing unit processes the results, including risk scores and disposal recommendations and warning event records, including warning time, disposal process, and release time. Automatically generated weekly and monthly reports and data in the data lake are used for horizontal comparative analysis of different tunnel sections. The main backup is stored in the central management center and the cloud at the same time, and is preserved for a long time according to project supervision requirements.

[0041] Example 4: To withstand high pressure, high salt fog, and long underwater distances, the platform's fiber Bragg grating strain gauges and displacement sensors feature corrosion-resistant sheaths, while the 3D laser scanner and dual-spectrum camera are equipped with anti-fog film. Air pressure sensors and gas concentration probes simultaneously monitor the pressure differential inside and outside the pipe joints and the presence of combustible gases. Ultra-wideband tags, combined with a metal reflection compensation algorithm, maintain sub-meter personnel positioning accuracy.

[0042] The edge computing unit utilizes the AIEdge platform, with an embedded chip housed in a high-pressure reinforced chamber. It connects to a immersed fiber optic cable via Gigabit Ethernet, and transmits data externally via dual communication links: 5G primary and 4G backup. Online change-point detection dynamically updates the danger threshold based on pressure differential changes. If an "overstay in the danger zone" is detected and gas concentrations suddenly increase, the three-level risk score jumps to the highest level, and the platform immediately activates emergency broadcasts, warning lights, and ventilation systems.

[0043] The central management center superimposes real-time pressure difference cloud maps and template displacement fields on the basis of BIM / CIM to generate digital twin visualization, which can simultaneously compare the risk levels of multiple pipe sections; all key raw data fragments and the entire disposal process are uploaded to the data lake to meet the requirements of cross-section horizontal comparison and engineering supervision, and data integrity is ensured through hard disk mirroring and dual power supply mechanisms.

[0044] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. Intelligent monitoring and early warning platform for the entire tunnel lining construction process, featuring: It includes a field perception unit, an edge computing unit and a central management center; the field perception unit includes an axial force sensor, a displacement sensor, an environmental monitoring device, a dual-spectrum camera and an ultra-wideband tag, which is used to collect multi-source data during the tunnel lining construction process, including template force, template deformation, template displacement, surrounding rock and environmental parameters, and personnel and equipment positioning information; the edge computing unit receives data from the field perception unit through a dual communication link, performs multimodal feature fusion, physical consistency constraint deep learning reasoning, online change point detection, skeleton posture recognition and dynamic Bayesian reasoning engine, and generates risk warning signals and disposal suggestions; the central management center receives the processing results of the edge computing unit, connects with the BIM / CIM model in real time, builds a digital twin visualization scene, and stores, analyzes and compares the data across sections; the field perception unit, edge computing unit and central management center are connected through a dual communication link, which supports automatic switching to the backup link when the main link fails, and realizes high-reliability data transmission.

2. The intelligent monitoring and early warning platform for the entire tunnel lining construction process according to claim 1 is characterized by: The axial force sensor is a fiber Bragg grating stress gauge, which is arranged in the top wedge or support rod of the key stress-bearing part of the lining formwork to monitor the axial force of the formwork; the displacement sensor is arranged in the middle of the lining formwork and the longitudinal side beam, including a fiber optic displacement meter and a laser displacement meter, which is used to track the deformation of the formwork. The data of the two are corrected in real time to ensure accuracy. The environmental monitoring equipment is used to obtain surrounding rock and environmental information, including a three-dimensional laser scanner, temperature and humidity sensors, air pressure sensors and gas concentration probes, which are used to collect surrounding rock and environmental data. The dual-spectrum camera and ultra-wideband tag are used for personnel and equipment positioning and behavior recognition, and can identify the wearing of safety protection equipment by personnel in low-light environments, and can be integrated with the ultra-wideband tag to achieve sub-meter personnel positioning accuracy.

3. The intelligent monitoring and early warning platform for the entire tunnel lining construction process according to claim 1 is characterized by: The edge computing unit has a built-in embedded chip that supports real-time reasoning of multiple video streams and sensor data fusion algorithms. It is equipped with a deep network model that integrates template stress, displacement and video features, introduces physical consistency constraints, and uses an online change point detection mechanism to dynamically adjust the warning threshold to adapt to different geological conditions and construction stages. Based on the skeletal posture recognition network and the temporal motion analysis framework, it identifies dangerous actions, and the recognition results are input into the dynamic Bayesian inference engine to generate three levels of risk scores from safety, warning to danger, and give corresponding disposal suggestions. When the risk score is higher than the threshold, the emergency broadcast and warning light strip are linked, and the responsible person information is pushed to the mobile terminal.

4. The intelligent monitoring and early warning platform for the entire tunnel lining construction process according to claim 1 is characterized by: The central management center receives processed data from the edge computing unit, maps the laser point cloud, sensor timing curve, risk score, recognition results, warning events and key raw data fragments to the BIM / CIM three-dimensional model, stores data by project dimension and supports horizontal comparative analysis across sections, conducts horizontal comparative analysis of different tunnel sections, automatically generates weekly or monthly reports, and outputs PDF or interactive 3D links for supervisor review; sets a 30-day cycle coverage and automatic desensitization and deletion mechanism for the data collected by the on-site perception unit, and stores the processing results of the edge computing unit, automatically generated weekly and monthly reports, and data in the data lake in the central management center and the cloud in the primary and secondary backup, and preserves them for a long time according to project supervision requirements.

5. The intelligent monitoring and early warning platform for the entire tunnel lining construction process according to claim 1 is characterized by: The dual communication links integrate 5G and Gigabit Ethernet interfaces. When the main link fails, it automatically switches to the 4G backup link, and retransmits the locally cached data after the link is restored.

6. The intelligent monitoring and early warning platform for the entire tunnel lining construction process according to claim 1 is characterized by: The edge computing unit and field perception unit are equipped with dual power supplies and hard disk mirroring. When external power, network or single-point hardware fails, data can be cached locally and retransmitted after the link is restored.

7. The intelligent monitoring and early warning platform for the entire tunnel lining construction process according to claim 3 is characterized by: Dangerous actions include the top support not being locked, the formwork hitting the pole, staying in the danger zone for too long, not wearing safety protection equipment, illegally crossing the warning line, and the identification delay is less than 0.2 seconds.

8. The intelligent monitoring and early warning platform for the entire tunnel lining construction process according to claim 3 is characterized by: The three-level risk score ranges from safe to dangerous. Ten minutes of continuous observation is required for the warning to be lifted, and the data of the three channels of stress, displacement and video must return to normal.

9. The intelligent monitoring and early warning platform for the entire tunnel lining construction process according to claim 4 is characterized by: The data stored in the data lake includes template stress, displacement, surrounding rock and environmental data collected by the on-site sensing unit, fragments of personnel violations, lists of responsible persons and full-process records of warning events, risk scores, disposal recommendations and automatically generated weekly and monthly reports after processing by the edge computing unit.

Images