Tunnel health monitoring method and system based on cloud platform

Abstract

The invention relates to the technical field of tunnel engineering monitoring, in particular to a tunnel health monitoring method and system based on a cloud platform, and the method comprises the steps: a data collection module obtains the original state data of a tunnel in real time through a sensor deployment network, and carries out the primary processing of the original state data based on edge calculation; obtaining actual state data of the tunnel; the monitoring cloud platform obtains actual state data of the tunnel in real time, and whether the current tunnel is healthy or not is judged according to a preset first tunnel state threshold value and the actual state data; and when the current tunnel is judged to be healthy, inputting the actual state data into a preset tunnel health prediction model to obtain a tunnel health prediction result used for assisting in judging whether the tunnel is healthy in the future. The method has the advantages that the processing burden of the monitoring cloud platform is effectively reduced, the data transmission pressure is relieved, meanwhile, risk prediction is conducted on the tunnel, potential safety hazards of the tunnel are effectively prevented, and the real-time performance, effectiveness and safety of monitoring and early warning are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel engineering monitoring, and in particular to a tunnel health monitoring method and system based on a cloud platform. Background Art

[0002] Tunnels are critical transportation infrastructure, and their structural safety monitoring is extremely important. At present, the health status of tunnel structures mainly relies on manual inspections and fixed sensor monitoring, which have obvious limitations. Manual inspections are inefficient and cannot capture tiny structural changes in real time. Although fixed sensors can provide real-time data, they are usually limited to local monitoring and have limited data processing capabilities. In addition, the existing system also has shortcomings in remote access, intelligent analysis, data sharing and collaboration, which makes it difficult for tunnel managers to understand the status of the tunnel in real time, affecting the timeliness and effectiveness of decision-making. At the same time, large-scale monitoring systems are expensive and require regular maintenance. For some small and medium-sized tunnel projects, the cost-effectiveness ratio is not high.

[0003] As tunnels expand in size and monitoring requirements increase, existing systems often have difficulty adapting to these changes and lack the necessary flexibility. In addition, existing systems lack early warning of the future health status of tunnels and are unable to meet the growing demand for tunnel health monitoring. Summary of the invention

[0004] 1. Technical issues to be resolved

[0005] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides a tunnel health monitoring method and system based on a cloud platform, which solves the technical problems that the existing system has limitations in data real-time, processing capability, remote monitoring, intelligent analysis, etc., and is difficult to meet the growing demand for tunnel health monitoring, as well as the lack of early warning of the future health status of the tunnel, which in turn increases the safety risks of tunnel facilities and affects the normal operation and safe use of the tunnel.

[0006] (II) Technical solution

[0007] In order to achieve the above object, the main technical solutions adopted by the present invention include:

[0008] In a first aspect, an embodiment of the present invention provides a tunnel health monitoring method based on a cloud platform, the method is implemented based on a pre-deployed sensor deployment network, and the method includes:

[0009] S11, the data acquisition module obtains the original state data of the tunnel in real time through the sensor deployment network, and performs preliminary processing on the original state data based on edge computing to obtain the actual state data of the tunnel;

[0010] The original state data includes data collected by all sensors in the pre-deployed sensor deployment network;

[0011] The preliminary processing includes: data filtering and data anomaly detection;

[0012] S12, the monitoring cloud platform obtains the actual status data of the tunnel in real time, and determines whether the current tunnel is healthy according to the actual status data and a preset first tunnel status threshold;

[0013] When judging the current health of the tunnel, the actual status data is input into a preset tunnel health prediction model to obtain a tunnel health prediction result for assisting in judging whether the tunnel will be healthy in the future.

[0014] Optionally, the S11 includes:

[0015] S11-1, the data acquisition module acquires the original state data of the tunnel in real time through the sensor deployment network, and pre-processes the received original state data; the pre-processing includes data cleaning and data sorting;

[0016] S11-2, the data acquisition module performs data filtering on the pre-processed original state data based on a preset second tunnel state threshold value to filter out valid data in the original state data;

[0017] S11-3. The data acquisition module performs data anomaly detection on the filtered valid data according to a preset anomaly detection algorithm, marks the abnormal data in the valid data, and the valid data that is not marked is the actual status data.

[0018] Optionally, the S11-3 includes:

[0019] The data acquisition module determines whether the filtered valid data is abnormal data based on the filtered valid data, the preset historical status data, and the preset formula 1, and marks the abnormal data;

[0020] The historical status data is the actual status data corresponding to the tunnel within the first preset time in the past;

[0021] The formula 1 is:

[0022]

[0023] Among them, z is the judgment parameter for judging whether the original state data after invalid data is removed is abnormal data, x is the valid data after screening, u is the average value of historical state data, n is the number of historical state data, u is the average value of historical state data, and n is the number of historical state data. i is the historical status data with index i.

[0024] Optionally, the S12 includes:

[0025] S12-1, the monitoring cloud platform obtains the actual status data of the tunnel in real time, and determines whether the current tunnel is healthy according to the actual status data and a preset first tunnel status threshold;

[0026] When it is determined that the current tunnel is healthy, the actual status data is saved as a cache data in a pre-set cache database; the cache database includes all cache data within the past second preset time, and the cache time corresponding to each cache data;

[0027] S12-2, the monitoring cloud platform slices all cache data in the cache database according to a preset time span to obtain at least one time series data and a batch corresponding to each time series data; any of the time series data is a continuous time series;

[0028] S12-3. The monitoring cloud platform inputs all time series data and the batches corresponding to each time series data into a preset tunnel health prediction model to obtain a tunnel health prediction result for assisting in determining whether the tunnel will be healthy in the future.

[0029] Optionally, the S12-3 includes:

[0030] The monitoring cloud platform extracts features from all time series data and obtains the feature parameters corresponding to each time series data;

[0031] The monitoring cloud platform obtains the attention matrix corresponding to each time series data according to the characteristic parameters corresponding to each time series data and the pre-set attention mechanism;

[0032] The monitoring cloud platform inputs the attention weight matrix corresponding to each time series data into a preset first fully connected layer to obtain a feature transformation result corresponding to each attention weight matrix; the first fully connected layer is used to perform feature transformation on the attention weight matrix according to the preset feature transformation weight matrix;

[0033] The monitoring cloud platform performs residual connection on the attention matrix corresponding to each time series data and the corresponding feature transformation result to obtain the residual connection result corresponding to each time series data;

[0034] The monitoring cloud platform obtains tunnel health prediction results to assist in judging whether the tunnel will be healthy in the future based on the residual connection results corresponding to all time series data, the batch weights corresponding to each time series data, and the pre-set weighted attention mechanism;

[0035] The batch weight is a corresponding weight assigned to each time series data based on the batch corresponding to each time series data and a pre-set weight allocation strategy table; the weight allocation strategy table is a correspondence table between batch quantity and batch, and batch weight.

[0036] Optionally, in S12-3, the monitoring cloud platform obtains a tunnel health prediction result for assisting in determining whether the tunnel will be healthy in the future according to the residual connection results corresponding to all the time series data and the batch weight corresponding to each time series data, and a preset weighted attention mechanism, including:

[0037] The monitoring cloud platform obtains the global context vector corresponding to the actual state data according to the residual connection results corresponding to all time series data and the batch weight corresponding to each time series data, as well as the preset formula 2; the formula 2 is:

[0038]

[0039] Among them, W score is the pre-set attention score weight matrix, R k is the residual connection result corresponding to the time series data of batch k, W k is the batch weight corresponding to the time series data with batch k, B is the number of batches, and c is the global context vector corresponding to the actual state data;

[0040] The monitoring cloud platform obtains the connection vector corresponding to the actual state data according to the global context vector corresponding to the actual state data and the actual state data, as well as the preset formula three; the formula three is:

[0041] L = concat(c,D);

[0042] Among them, L is the connection vector corresponding to the actual state data, c is the global context vector corresponding to the actual state data, and D is the actual state data;

[0043] The monitoring cloud platform inputs the connection vector corresponding to the actual status data into a preset second fully connected layer to obtain a tunnel health prediction result for assisting in determining whether the tunnel will be healthy in the future;

[0044] The second fully connected layer is used to output a regularized output result corresponding to the connection vector, that is, a tunnel health prediction result.

[0045] Optionally, the S12-1 further includes:

[0046] The monitoring cloud platform determines in real time whether the cache time corresponding to each cache data in the cache database exceeds the preset time threshold;

[0047] If it exceeds, the cached data and the cached time corresponding to the cached data will be deleted.

[0048] Optionally, the S12 further includes:

[0049] The monitoring cloud platform aggregates, stores and analyzes the actual status data received;

[0050] and, establishing a data visualization interface based on the actual status data after aggregation, data storage and data analysis processing to enable management personnel to detect tunnel status in real time;

[0051] and, based on the received user input information, generating corresponding real-time reports and data statistics results;

[0052] The visualization interface includes real-time maps, charts, and reports.

[0053] Optionally, the S12 further includes:

[0054] When the monitoring cloud platform obtains the actual status data of the tunnel, the signal-to-noise ratio of the actual status data is obtained;

[0055] Furthermore, the monitoring cloud platform selects a corresponding denoising process based on the signal-to-noise ratio of the actual state data and a preset signal-to-noise ratio gradient threshold, and performs denoising on the actual state data; the denoising process includes: one or more of moving average filtering, weighted average filtering, median filtering, Wiener filtering and wavelet threshold denoising;

[0056] The monitoring cloud platform eliminates dispersion, fills in missing values ​​and normalizes the actual status data after denoising.

[0057] In a second aspect, an embodiment of the present invention provides a tunnel health monitoring system based on a cloud platform, comprising:

[0058] Sensor deployment network, data acquisition modules and monitoring cloud platform;

[0059] The sensor deployment network is set at a designated location in the tunnel, and includes a tilt sensor, a strain sensor, a temperature sensor, a vibration sensor, and a displacement sensor;

[0060] The sensor deployment network is used to detect various data in the tunnel, that is, the original state data of the tunnel, and send the original state data of the tunnel to the data acquisition module;

[0061] The data acquisition module is used to perform preliminary processing on the original status data based on edge computing to obtain actual status data of the tunnel;

[0062] The monitoring cloud platform is used to obtain the actual status data of the tunnel in real time, and judge whether the current tunnel is healthy according to the preset first tunnel status threshold and the actual status data;

[0063] When judging the current health of the tunnel, the actual status data is input into a preset tunnel health prediction model to obtain a tunnel health prediction result for assisting in judging whether the tunnel will be healthy in the future.

[0064] (III) Beneficial effects

[0065] The beneficial effects of the present invention are as follows: a cloud platform-based tunnel health monitoring method of the present invention collects various data in the tunnel in real time through a sensor deployment network, and while delegating part of the computing tasks to the data acquisition module, performs future predictions through the monitoring cloud platform. Compared with the existing technology, it can effectively reduce the processing burden of the monitoring cloud platform, alleviate the pressure of data transmission, and predict risks for the tunnel, effectively prevent safety hazards in the tunnel, and improve the real-time, effectiveness and safety of monitoring and early warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 A flow chart of a tunnel health monitoring method based on a cloud platform provided in an embodiment of the present invention;

[0067] Figure 2 A schematic diagram of the structure of a tunnel health monitoring system based on a cloud platform provided in an embodiment of the present invention;

[0068] Figure 3 A schematic diagram of a sensor deployment network provided in an embodiment of the present invention;

[0069] Figure 4 This is a schematic diagram of the monitoring cloud platform structure. DETAILED DESCRIPTION

[0070] In order to better explain the present invention and facilitate understanding, the present invention is described in detail below through specific implementation modes in conjunction with the accompanying drawings.

[0071] A cloud platform-based tunnel health monitoring method proposed in an embodiment of the present invention collects various data in the tunnel in real time through a sensor deployment network, and while delegating some computing tasks to the data acquisition module, performs future predictions through a monitoring cloud platform. Compared with the existing technology, it can effectively reduce the processing burden of the monitoring cloud platform, alleviate the pressure of data transmission, and predict risks for the tunnel, effectively prevent safety hazards in the tunnel, and improve the real-time, effectiveness and safety of monitoring and early warning.

[0072] In order to better understand the above technical solution, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to enable a clearer and more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0073] Example 1

[0074] An embodiment of the present invention provides a tunnel health monitoring method based on edge computing. The method is implemented based on a pre-deployed sensor deployment network. The sensor deployment network includes: tilt sensors (including fixed inclinometers, MEMS inclinometers), strain sensors, temperature sensors (including thermometers and fiber grating thermometers), vibration sensors and displacement sensors, etc.; each sensor is set at a specified position in the tunnel.

[0075] The method flow is as follows Figure 1 As shown, including:

[0076] S11, the data acquisition module obtains the original state data of the tunnel in real time through the sensor deployment network, and performs preliminary processing on the original state data based on edge computing to obtain the actual state data of the tunnel;

[0077] The original state data includes data collected by all sensors in the pre-deployed sensor deployment network;

[0078] The preliminary processing includes: data filtering and data anomaly detection;

[0079] S12, the monitoring cloud platform obtains the actual status data of the tunnel in real time, and determines whether the current tunnel is healthy according to the actual status data and a preset first tunnel status threshold;

[0080] When judging the current health of the tunnel, the actual status data is input into a preset tunnel health prediction model to obtain a tunnel health prediction result for assisting in judging whether the tunnel will be healthy in the future.

[0081] This embodiment provides a tunnel health monitoring method based on edge computing, which collects various data in the tunnel in real time through a sensor deployment network, and while delegating some computing tasks to the data acquisition module, performs future predictions through a monitoring cloud platform. It can effectively reduce the processing burden of the monitoring cloud platform, alleviate the pressure of data transmission, and predict risks for the tunnel, effectively prevent safety hazards in the tunnel, and improve the real-time, effectiveness and security of monitoring and early warning.

[0082] Furthermore, the S11 includes:

[0083] S11-1, the data acquisition module obtains the original state data of the tunnel in real time through the sensor deployment network, and pre-processes the received original state data; the original state data includes the data corresponding to all sensors in the pre-deployed sensor deployment network;

[0084] The preprocessing includes data cleaning and data arrangement;

[0085] By cleaning and arranging the raw state data, the raw state data can be screened, converted and corrected, and errors, missing values, anomalies or duplications in the data can be processed. Among them, common data cleaning methods include: missing value processing, outlier processing, noise data processing, data deduplication, data normalization (normalization and standardization, etc.); common data arranging methods include: data format conversion, data integration, data aggregation, data encoding and feature extraction. In general, the raw state data is a data set, including the corresponding data collected by all sensors in the sensor deployment network.

[0086] S11-2, the data acquisition module performs data filtering on the pre-processed original state data based on a preset second tunnel state threshold value to filter out valid data in the original state data;

[0087] S11-3. The data acquisition module performs data anomaly detection on the filtered valid data according to a preset anomaly detection algorithm, marks the abnormal data in the valid data, and the valid data that is not marked is the actual status data.

[0088] Furthermore, the S11-1 further includes:

[0089] When the data acquisition module obtains the original state data of the tunnel through the sensor deployment network, the obtained original state data is stored in a pre-set local database.

[0090] Furthermore, the S11-2 includes:

[0091] The data acquisition module compares the pre-processed raw state data with the preset second tunnel state threshold to determine whether it is valid data; wherein the second tunnel state threshold is a state threshold set in one-to-one correspondence with all sensors in the sensor deployment network, and each second tunnel state threshold includes an upper limit and a lower limit, which are used to preliminarily determine whether the data is valid;

[0092] If it is invalid data, the data will be marked as invalid data and removed. The invalid data after removal will be saved to the preset data cache area accordingly, and the invalid data in the data cache area will be cleared periodically to facilitate the detection of possible problems.

[0093] If it is valid data, subsequent abnormal data detection is performed.

[0094] Furthermore, when the original state data corresponding to any sensor contains continuous invalid data, or the original state data corresponding to the sensor contains invalid data for multiple times (preset number of times) within a period of time, it is determined that the sensor may be abnormal; all data corresponding to the sensor that are screened out within this period of time are recovered, and an abnormal file is generated; the abnormal file includes the collection time corresponding to each data, the data value, and the judgment result of whether it is invalid data, and is saved in the local database to facilitate subsequent fault type troubleshooting.

[0095] Furthermore, the S11-3 includes:

[0096] Common anomaly detection algorithms include: mean-variance method and box plot method based on statistics, support vector machine (SVM) anomaly detection model based on machine learning, clustering algorithm (such as K-means and DBSCAN, etc.) and isolation forest, simple exponential smoothing and ARIMA model based on time series.

[0097] In this embodiment, the data acquisition module determines whether the filtered valid data is abnormal data based on the filtered valid data and the preset historical status data, as well as the preset formula 1, and marks the abnormal data;

[0098] The historical status data is the actual status data corresponding to the tunnel within the first preset time in the past;

[0099] The formula 1 is:

[0100]

[0101] Among them, z is the judgment parameter for judging whether the original state data after invalid data is removed is abnormal data, x is the valid data after screening, u is the average value of historical state data, n is the number of historical state data, u is the average value of historical state data, and n is the number of historical state data. i is the historical status data with index i.

[0102] When z < 1, the data acquisition module marks the original state data as normal data;

[0103] When 1≤z<2, the data acquisition module marks the original state data as slightly abnormal data;

[0104] When 2≤z<3, the data acquisition module marks the original state data as moderately abnormal data;

[0105] When z≥3, the data acquisition module marks the original state data as severely abnormal data.

[0106] Furthermore, the S11 further includes:

[0107] S11-4. The data acquisition module determines the working status of the instrument corresponding to the abnormal data according to the abnormal degree marked by the original status data and the pre-set abnormal diagnosis strategy.

[0108] S11-5. The data acquisition module stores the actual state data after preliminary processing in a local database, and establishes a corresponding relationship between the actual state data and the original state data.

[0109] The local database includes: a MySQL database, a MongoDB database, or a Solr database.

[0110] Further, the S12 includes:

[0111] S12-1, the monitoring cloud platform obtains the actual status data of the tunnel in real time, and determines whether the current tunnel is healthy according to the actual status data and a preset first tunnel status threshold;

[0112] When it is determined that the current tunnel is healthy, the actual status data is saved as a cache data in a pre-set cache database; the cache database includes all cache data within the past second preset time, and the cache time corresponding to each cache data;

[0113] Among them, the data acquisition module is connected to the monitoring cloud platform through a 4G / 5G smart gateway or wired network.

[0114] S12-2, the monitoring cloud platform slices all cache data in the cache database according to a preset time span to obtain at least one time series data and a batch corresponding to each time series data; any of the time series data is a continuous time series;

[0115] S12-3. The monitoring cloud platform inputs all time series data and the batches corresponding to each time series data into a preset tunnel health prediction model to obtain a tunnel health prediction result for assisting in determining whether the tunnel will be healthy in the future.

[0116] Furthermore, before the actual status data is saved as a cache data in a pre-set cache database, it is necessary to perform a secondary evaluation on the actual status data; that is, the monitoring cloud platform obtains the data change values ​​of the detection objects corresponding to various sensors in the sensor deployment network based on the actual status data and all the cache data in the cache database; and the monitoring cloud platform evaluates the health and risk status of the tunnel based on the data change values ​​of the detection objects corresponding to each sensor in the sensor deployment network and the pre-set change warning values ​​of each detection object.

[0117] Further, the S12-3 includes:

[0118] The monitoring cloud platform extracts features from all time series data and obtains the feature parameters corresponding to each time series data;

[0119] The monitoring cloud platform obtains the attention matrix corresponding to each time series data according to the characteristic parameters corresponding to each time series data and the pre-set attention mechanism;

[0120] The monitoring cloud platform inputs the attention weight matrix corresponding to each time series data into a preset first fully connected layer to obtain a feature transformation result corresponding to each attention weight matrix; the first fully connected layer is used to perform feature transformation on the attention weight matrix according to the preset feature transformation weight matrix;

[0121] The monitoring cloud platform performs residual connection on the attention matrix corresponding to each time series data and the corresponding feature transformation result to obtain the residual connection result corresponding to each time series data;

[0122] The monitoring cloud platform obtains the global context vector corresponding to the actual state data according to the residual connection results corresponding to all time series data and the batch weight corresponding to each time series data, as well as the preset formula 2; the formula 2 is:

[0123]

[0124] Among them, W score is the pre-set attention score weight matrix, R k is the residual connection result corresponding to the time series data of batch k, W k is the batch weight corresponding to the time series data with batch k, B is the number of batches, and c is the global context vector corresponding to the actual state data;

[0125] The monitoring cloud platform obtains the connection vector corresponding to the actual state data according to the global context vector corresponding to the actual state data and the actual state data, as well as the preset formula three; the formula three is:

[0126] L = concat(c,D);

[0127] Among them, L is the connection vector corresponding to the actual state data, c is the global context vector corresponding to the actual state data, and D is the actual state data;

[0128] The monitoring cloud platform inputs the connection vector corresponding to the actual status data into a preset second fully connected layer to obtain a tunnel health prediction result for assisting in determining whether the tunnel will be healthy in the future;

[0129] The second fully connected layer is used to output a regularized output result corresponding to the connection vector, that is, a tunnel health prediction result;

[0130] The batch weight is a corresponding weight assigned to each time series data based on the batch corresponding to each time series data and a pre-set weight allocation strategy table; the weight allocation strategy table is a correspondence table between batch quantity and batch, and batch weight.

[0131] Furthermore, the S12-1 further includes:

[0132] The monitoring cloud platform determines in real time whether the cache time corresponding to each cache data in the cache database exceeds the preset time threshold;

[0133] If it exceeds, the cached data and the cached time corresponding to the cached data will be deleted.

[0134] By regularly deleting cached data, the data storage pressure of the monitoring cloud platform can be reduced.

[0135] Furthermore, S12 also includes:

[0136] The monitoring cloud platform aggregates, stores and analyzes the actual status data received;

[0137] and, establishing a data visualization interface based on the actual status data after aggregation, data storage and data analysis processing to enable management personnel to detect tunnel status in real time;

[0138] and, based on the received user input information, generating corresponding real-time reports and data statistics results;

[0139] The visualization interface includes real-time maps, charts, and reports.

[0140] The monitoring cloud platform aggregates, stores and analyzes data, making it easier for users to view the data through mobile terminals or PCs, thereby enhancing the convenience and practicality of actual use.

[0141] Furthermore, S12 also includes:

[0142] When the monitoring cloud platform obtains the actual status data of the tunnel, the signal-to-noise ratio of the actual status data is obtained;

[0143] Furthermore, the monitoring cloud platform selects a corresponding denoising process based on the signal-to-noise ratio of the actual state data and a preset signal-to-noise ratio gradient threshold, and performs denoising on the actual state data; the denoising process includes: one or more of moving average filtering, weighted average filtering, median filtering, Wiener filtering and wavelet threshold denoising;

[0144] The monitoring cloud platform eliminates dispersion, fills in missing values ​​and normalizes the actual status data after denoising. Based on the signal-to-noise ratio of the received data, the appropriate denoising process is selected, which greatly enhances the data processing speed, improves the implementation efficiency, and greatly enhances the real-time performance of tunnel monitoring.

[0145] Furthermore, before the data acquisition module sends the actual status data stored in the local database to the monitoring cloud platform, it is necessary to compress the data to reduce possible IO congestion and improve data transmission efficiency.

[0146] Therefore, the compressed sensing algorithm is used to compress the actual state data. Common compressed sensing algorithms include coefficient representation algorithm, OMP algorithm, L1 norm minimization algorithm and wavelet transform algorithm.

[0147] The present embodiment provides a tunnel health monitoring method based on edge computing, which receives real-time data (i.e., original status data) collected by all sensors in the sensor deployment network through a data acquisition module and performs preliminary processing. Through the above-mentioned edge computing method, the pressure of data transmission is greatly reduced, the processing burden of the monitoring cloud platform is reduced, and the data processing efficiency is improved; and the data acquisition module establishes a corresponding relationship between the original status data and the actual status data by setting up a database, which is convenient for on-site personnel to view and maintain and to find out the cause of the problem; and, through the tunnel health prediction model, the prediction of the future risk status of the tunnel is realized, thereby enhancing the safety of the tunnel and reducing the risk of accidents.

[0148] Example 2

[0149] This embodiment provides a tunnel health monitoring system based on edge computing, whose structure is as follows: Figure 2As shown, it includes: a monitoring cloud platform and at least one group of sensor deployment networks, and a data acquisition module corresponding to each sensor deployment network, and a user terminal that can access the monitoring cloud platform for interaction. Each sensor deployment network is connected to the corresponding data acquisition module in communication, and all data acquisition modules are connected to the monitoring cloud platform in communication. All sensor deployment networks are distributed in different locations of the tunnel, and each sensor deployment network includes: a tilt sensor, a strain sensor, a temperature sensor, a vibration sensor and a displacement sensor; wherein the tilt sensor includes a fixed inclinometer and a MEMS inclinometer, and the tilt sensor is used to measure the tilt angle of the tunnel. The change in the tilt angle can reflect the deformation of the tunnel, thereby judging the structural health of the tunnel. The strain sensor includes a vibrating string strain gauge, a vibrating string surface strain gauge, a fiber Bragg grating surface strain gauge, a vibrating string embedded strain gauge, and a fiber Bragg grating embedded strain gauge. The strain sensor is used to measure the strain of the tunnel structure. By monitoring the strain change of the tunnel structure, the structural health of the tunnel can be judged, and maintenance measures can be taken in time. The temperature sensor includes a thermometer and a fiber Bragg grating thermometer, and the temperature sensor is used to measure the temperature change of the tunnel. Temperature changes in tunnels can affect the stability of tunnel structures. Therefore, monitoring of temperature sensors is very important for judging the health of tunnels. Vibration sensors are used to monitor the vibration of tunnel structures. Vibration changes in tunnel structures can reflect the structural health of tunnels. Timely detection of abnormal vibrations allows for appropriate maintenance measures. Displacement sensors include vibrating-string displacement meters, vibrating-string embedded joint meters, vibrating-string surface joint meters, and potentiometer-type large-range displacement meters. All sensors in any sensor deployment network are connected to the data acquisition module corresponding to the sensor deployment network. The deployment status of the sensor deployment network is as follows: Figure 3 shown.

[0150] The data acquisition module integrates 4G / 5G smart gateway, CAN bus interface and other data transmission interfaces. The data acquisition module can realize functions such as data acquisition, edge computing, local storage and data transmission, that is, data collection from various sensors, and data processing and transmission to the monitoring cloud platform through wireless or wired networks.

[0151] The monitoring cloud platform uses cloud computing and IoT technologies to process and store the received data in real time, provide services such as data collection, early warning, viewing, report printing, information aggregation, etc., and supports multiple browser versions. The monitoring cloud platform has built-in data analysis functions for model training, evaluation, predictive analysis, etc.

[0152] The user end is usually a mobile device or PC device, including control panel, data center, alarm center, measurement point management, enterprise engineering management, report management, and develops a friendly interface for managers to view.

[0153] Furthermore, data sharing is achieved among the data acquisition module, the monitoring cloud platform and the user end, and data management permissions are set among the three. Remote operation can be achieved through permission authentication, that is, functions such as uploading and downloading, collaborative communication, annotation, comparative analysis, case sharing, and knowledge base sharing are realized.

[0154] Further, the data acquisition module includes a data acquisition unit, a data preprocessing unit and a data storage unit;

[0155] The data collection unit is used to receive the original state data of the tunnel in real time and send it to the edge computing unit and the data storage unit;

[0156] The edge computing unit is used to filter the original state data based on a preset state data threshold, remove invalid data in the original state data, and mark abnormal data in the valid data according to a preset abnormality detection algorithm. The remaining unmarked original state data is the actual state data of the tunnel, and the data processing unit sends the actual state data to the data storage unit;

[0157] The data storage unit is used to store the acquired original state data in a preset local database when original state data is received; and to store the actual state data in the local database when actual state data is received, and to establish a corresponding relationship between the actual state data and the original state data.

[0158] Normally, the data acquisition module also includes a data acquisition workstation located at the tunnel site. The data acquisition workstation uses different conditioning methods to condition the signals of various types of sensors, and processes and converts the conditioned sensor signals to finally form digital signals that can be transmitted remotely. On this basis, according to the different signal input methods and monitored physical quantities, the collected signals are divided into digital signals, analog signals and optical signals, and processed by pre-set digital signal data acquisition software, analog signal data acquisition software and optical signal data acquisition software. The three software have similar business functions, but different algorithms for processing data and different hardware devices they rely on, so they will be developed separately.

[0159] Furthermore, the monitoring cloud platform uses the powerful computing and storage capabilities of the cloud computing platform to process and store the collected data in real time. The monitoring cloud platform is based on cloud computing and Internet of Things technology to provide effective, reliable and convenient data monitoring services to obtain relevant monitoring data and alarm information in the first time, thereby realizing the safety monitoring of various engineering projects. The main functions include data collection, data warning, data viewing, report printing, and information aggregation. The monitoring cloud platform supports browser versions: IE9 and above, Firefox11 and above, Chrome10 and above.

[0160] Monitoring cloud platform, its structure is as follows Figure 4 As shown, it is generally divided into: data persistence layer, basic support layer, business layer and presentation layer.

[0161] The data persistence layer is the bottom layer of the entire platform. It obtains data through scheduled task applications, determines the data preprocessing steps based on the actual data quality, and stores the collected data persistently. That is, it determines how to proceed with the noise reduction process based on the signal-to-noise ratio of the actual state data received.

[0162] The basic support layer is the general basic support function of the software system, which is an abstract function independent of the business field. Establishing a perfect basic support layer lays a solid foundation for system development, debugging and maintenance, making time controllable and quality reliable.

[0163] In the business layer, various types of sensor equipment are managed to implement specific business operations, and business modules will be customized and developed in combination with the actual management needs of tunnel health monitoring. In the business layer, prediction models can be set up, including time series analysis models, regression analysis models, neural network prediction models, decision trees, etc., to analyze actual status data to predict future tunnel conditions. The tunnel health prediction model mentioned in Example 1 is used to obtain tunnel health prediction results for assisting in determining whether the tunnel will be healthy in the future, which is implemented in the business layer.

[0164] In the presentation layer, in order to meet the needs of business management, comprehensive management of various businesses is realized, as well as data interaction, business linkage and comprehensive analysis and decision-making between multiple business units. There are mainly Web applications and mobile applications.

[0165] Furthermore, the user end includes equipment control panel, data center, alarm center, measurement point management, enterprise engineering management, and report management. The purpose is to develop a user-friendly interface so that tunnel managers can access the cloud platform through the Internet and view the tunnel health status and analysis reports in real time.

[0166] Equipment control panel: Displays the working status of each sensor in an intuitive chart, including whether it is operating normally, signal strength, etc. Provides a switch control button for the device, so that managers can remotely start and stop specific sensors when necessary. Displays basic information about the device, such as device model, installation location, installation time, etc.

[0167] Data Center: Presents real-time data collected by sensors in the form of dynamic charts and data tables, such as displacement, stress, vibration, temperature and humidity, water level, etc. Supports time filtering of data, and users can choose to view data change trends within a specific time period. Provides data download function for users to conduct further offline analysis.

[0168] Alarm center: Displays the current alarm information in real time, including alarm type (such as displacement exceeding the limit, stress abnormality, etc.), alarm level (level 1 to 5), alarm time and location. Records and classifies historical alarm information to facilitate users to query and analyze alarm patterns and trends. Provides alarm confirmation and processing buttons, and managers confirm receipt of alarms and record processing measures.

[0169] Measuring point management: Displays the distribution of each measuring point in the tunnel in the form of a map and a list. Supports adding, deleting, and modifying measuring points, making it easy to adjust the measuring point layout according to actual monitoring needs. Displays detailed information for each measuring point, including sensor type, installation angle, monitoring parameters, etc.

[0170] Enterprise project management: Displays a list of tunnel projects being monitored, including basic information such as project name, location, tunnel type, etc. Provides the functions of adding, editing, and deleting projects to facilitate the management of multiple tunnel projects. For each project, displays key information such as project progress and key areas of monitoring.

[0171] Report management: Automatically generate various monitoring reports, such as daily reports, weekly reports, monthly reports, etc. The report content includes monitoring data summary, data analysis results, alarm conditions, etc. Support users to customize the format and content of reports to meet different needs. Provide report printing and export functions to facilitate users to archive and share.

[0172] The present embodiment provides a tunnel health monitoring system based on edge computing, which utilizes a monitoring cloud platform to realize real-time processing and storage of data, greatly improving the monitoring efficiency. It is no longer limited by the limitations of traditional storage and processing methods, and can quickly respond to large amounts of data influx to ensure the timeliness and effectiveness of the data. The tunnel health prediction model is applied to significantly improve the accuracy of the tunnel health status prediction and accurately predict the changing trend of the tunnel structure. Through edge computing, part of the computing tasks are delegated to the data acquisition module, effectively reducing the processing burden of the monitoring cloud platform, alleviating the pressure of the data transmission process, reducing the processing and transmission delays, and improving the real-time and effectiveness of monitoring warnings.

[0173] In the description of the present invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0174] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0175] In the present invention, unless otherwise clearly specified and limited, when a first feature is “on” or “below” a second feature, it may be that the first and second features are in direct contact, or that the first and second features are in indirect contact through an intermediate medium. Moreover, when a first feature is “above”, “above” or “above” a second feature, it may be that the first feature is directly above or obliquely above the second feature, or it may simply mean that the first feature is higher in level than the second feature. When a first feature is “below”, “below” or “below” a second feature, it may be that the first feature is directly below or obliquely below the second feature, or it may simply mean that the first feature is lower in level than the second feature.

[0176] In the description of this specification, the description of the terms "one embodiment", "some embodiments", "embodiment", "example", "specific example" or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, unless they are contradictory.

[0177] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may alter, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A tunnel health monitoring method based on a cloud platform, the method is implemented based on a pre-deployed sensor deployment network, characterized in that: The method comprises: S11, the data acquisition module obtains the original state data of the tunnel in real time through the sensor deployment network, and performs preliminary processing on the original state data based on edge computing to obtain the actual state data of the tunnel; The original state data includes data collected by all sensors in the pre-deployed sensor deployment network; The preliminary processing includes: data filtering and data anomaly detection; S12, the monitoring cloud platform obtains the actual status data of the tunnel in real time, and determines whether the current tunnel is healthy according to the actual status data and a preset first tunnel status threshold; When judging the current health of the tunnel, the actual status data is input into a preset tunnel health prediction model to obtain a tunnel health prediction result for assisting in judging whether the tunnel will be healthy in the future.

2. The cloud platform-based tunnel health monitoring method according to claim 1, characterized in that: The S11 includes: S11-1, the data acquisition module acquires the original state data of the tunnel in real time through the sensor deployment network, and pre-processes the received original state data; the pre-processing includes data cleaning and data sorting; S11-2, the data acquisition module performs data filtering on the pre-processed original state data based on a preset second tunnel state threshold value to filter out valid data in the original state data; S11-3. The data acquisition module performs data anomaly detection on the filtered valid data according to a preset anomaly detection algorithm, marks the abnormal data in the valid data, and the valid data that is not marked is the actual status data.

3. The cloud platform-based tunnel health monitoring method according to claim 2, characterized in that: The S11-3 includes: The data acquisition module determines whether the filtered valid data is abnormal data based on the filtered valid data, the preset historical status data, and the preset formula 1, and marks the abnormal data; The historical status data is the actual status data corresponding to the tunnel within the first preset time in the past; The formula 1 is: Among them, z is the judgment parameter for judging whether the original state data after invalid data is removed is abnormal data, x is the valid data after screening, u is the average value of historical state data, n is the number of historical state data, u is the average value of historical state data, and n is the number of historical state data. i is the historical status data with index i.

4. The cloud platform-based tunnel health monitoring method according to claim 1, characterized in that: The S12 includes: S12-1, the monitoring cloud platform obtains the actual status data of the tunnel in real time, and determines whether the current tunnel is healthy according to the actual status data and a preset first tunnel status threshold; When it is determined that the current tunnel is healthy, the actual status data is saved as a cache data in a pre-set cache database; the cache database includes all cache data within the past second preset time, and the cache time corresponding to each cache data; S12-2, the monitoring cloud platform slices all cache data in the cache database according to a preset time span to obtain at least one time series data and a batch corresponding to each time series data; any of the time series data is a continuous time series; S12-3. The monitoring cloud platform inputs all time series data and the batches corresponding to each time series data into a preset tunnel health prediction model to obtain a tunnel health prediction result for assisting in determining whether the tunnel will be healthy in the future.

5. The cloud platform-based tunnel health monitoring method according to claim 4, characterized in that: The S12-3 includes: The monitoring cloud platform extracts features from all time series data and obtains the feature parameters corresponding to each time series data; The monitoring cloud platform obtains the attention matrix corresponding to each time series data according to the characteristic parameters corresponding to each time series data and the pre-set attention mechanism; The monitoring cloud platform inputs the attention weight matrix corresponding to each time series data into a preset first fully connected layer to obtain a feature transformation result corresponding to each attention weight matrix; the first fully connected layer is used to perform feature transformation on the attention weight matrix according to the preset feature transformation weight matrix; The monitoring cloud platform performs residual connection on the attention matrix corresponding to each time series data and the corresponding feature transformation result to obtain the residual connection result corresponding to each time series data; The monitoring cloud platform obtains tunnel health prediction results to assist in judging whether the tunnel will be healthy in the future based on the residual connection results corresponding to all time series data, the batch weights corresponding to each time series data, and the pre-set weighted attention mechanism; The batch weight is a corresponding weight assigned to each time series data based on the batch corresponding to each time series data and a pre-set weight allocation strategy table; the weight allocation strategy table is a correspondence table between batch quantity and batch, and batch weight.

6. The cloud platform-based tunnel health monitoring method according to claim 5, characterized in that: In S12-3, the monitoring cloud platform obtains a tunnel health prediction result for assisting in judging whether the tunnel will be healthy in the future according to the residual connection results corresponding to all time series data and the batch weight corresponding to each time series data, as well as a pre-set weighted attention mechanism, including: The monitoring cloud platform obtains the global context vector corresponding to the actual state data according to the residual connection results corresponding to all time series data and the batch weight corresponding to each time series data, as well as the preset formula 2; the formula 2 is: Among them, W score is the pre-set attention score weight matrix, R k is the residual connection result corresponding to the time series data of batch k, W k is the batch weight corresponding to the time series data with batch k, B is the number of batches, and c is the global context vector corresponding to the actual state data; The monitoring cloud platform obtains the connection vector corresponding to the actual state data according to the global context vector corresponding to the actual state data and the actual state data, as well as the preset formula three; the formula three is: L = concat(c,D); Among them, L is the connection vector corresponding to the actual state data, c is the global context vector corresponding to the actual state data, and D is the actual state data; The monitoring cloud platform inputs the connection vector corresponding to the actual status data into a preset second fully connected layer to obtain a tunnel health prediction result for assisting in determining whether the tunnel will be healthy in the future; The second fully connected layer is used to output a regularized output result corresponding to the connection vector, that is, a tunnel health prediction result.

7. The cloud platform-based tunnel health monitoring method according to claim 4, characterized in that: The S12-1 also includes: The monitoring cloud platform determines in real time whether the cache time corresponding to each cache data in the cache database exceeds the preset time threshold; If it exceeds, the cached data and the cached time corresponding to the cached data will be deleted.

8. The cloud platform-based tunnel health monitoring method according to claim 1, characterized in that: The S12 further includes: The monitoring cloud platform aggregates, stores and analyzes the actual status data received; and, establishing a data visualization interface based on the actual status data after aggregation, data storage and data analysis processing to enable management personnel to detect tunnel status in real time; and, based on the received user input information, generating corresponding real-time reports and data statistics results; The visualization interface includes real-time maps, charts, and reports.

9. The cloud platform-based tunnel health monitoring method according to claim 1, characterized in that: The S12 further includes: When the monitoring cloud platform obtains the actual status data of the tunnel, the signal-to-noise ratio of the actual status data is obtained; Furthermore, the monitoring cloud platform selects a corresponding denoising process based on the signal-to-noise ratio of the actual state data and a preset signal-to-noise ratio gradient threshold, and performs denoising on the actual state data; the denoising process includes: one or more of moving average filtering, weighted average filtering, median filtering, Wiener filtering and wavelet threshold denoising; The monitoring cloud platform eliminates dispersion, fills in missing values ​​and normalizes the actual status data after denoising.

10. A tunnel health monitoring system based on a cloud platform, characterized in that: include: Sensor deployment network, data collection modules and monitoring cloud platform; The sensor deployment network is set at a designated location in the tunnel, and includes a tilt sensor, a strain sensor, a temperature sensor, a vibration sensor, and a displacement sensor; The sensor deployment network is used to detect various data in the tunnel, that is, the original state data of the tunnel, and send the original state data of the tunnel to the data acquisition module; The data acquisition module is used to perform preliminary processing on the original status data based on edge computing to obtain the actual status data of the tunnel; The monitoring cloud platform is used to obtain the actual status data of the tunnel in real time, and judge whether the current tunnel is healthy according to the preset first tunnel status threshold and the actual status data; When judging the current health of the tunnel, the actual status data is input into a preset tunnel health prediction model to obtain a tunnel health prediction result for assisting in judging whether the tunnel will be healthy in the future.

Images