A smart city tunnel crack emergency supervision internet of things system, method and medium

Abstract

This invention relates to the field of tunnel crack monitoring, and provides a smart city tunnel crack emergency monitoring IoT system, method, and medium. The system includes an emergency monitoring management platform. The platform is configured to: determine candidate anomaly locations and corresponding first anomaly parameters within the tunnel based on multiple sets of first detection data and train operation data; determine re-inspection parameters based on the first anomaly parameters, and control a robot to detect the candidate anomaly locations according to the re-inspection parameters, obtaining second detection data for the candidate anomaly locations; determine the target anomaly location from the candidate anomaly locations based on the second detection data; and control the traction motor to drive the train through the tunnel and issue an early warning based on the first anomaly parameters corresponding to the target anomaly location. This system can achieve real-time quantitative monitoring, precise location, and trend early warning of tunnel cracks, improving the proactive safety assurance level of tunnel infrastructure.

Description

Technical Field

[0001] This invention relates to the field of tunnel crack monitoring, and in particular to an IoT system, method and medium for emergency monitoring of tunnel cracks in smart cities. Background Technology

[0002] As a critical infrastructure component of urban transportation networks, the structural health of urban tunnels directly impacts urban operational safety and the safety of public life and property. The occurrence and propagation of tunnel cracks can lead to problems such as water seepage, structural deterioration, and reduced load-bearing capacity, and may even result in serious consequences such as localized collapse or overall instability. Traditional tunnel crack monitoring mainly relies on regular manual inspections, using visual checks or simple instruments to measure crack parameters. This method is inefficient, subjective, and exhibits significant time lag, making it difficult to achieve real-time detection and early warning of cracks.

[0003] Therefore, there is a need to provide an IoT system, method, and medium for emergency monitoring of tunnel cracks in smart cities, so as to realize real-time quantitative monitoring, accurate location, and trend early warning of tunnel cracks, provide a scientific basis for tunnel maintenance and emergency response, and improve the proactive safety protection level of tunnel infrastructure. Summary of the Invention

[0004] The invention includes a smart city tunnel crack emergency monitoring IoT system. The IoT system includes an emergency monitoring management platform configured to: acquire train operation data and multiple sets of first detection data corresponding to the tunnel; determine one or more candidate anomaly locations and corresponding one or more first anomaly parameters within the tunnel based on the multiple sets of first detection data and the train operation data; determine re-inspection parameters based on the one or more first anomaly parameters, and control a robot to detect the one or more candidate anomaly locations according to the re-inspection parameters, obtaining second detection data for the one or more candidate anomaly locations, wherein the re-inspection parameters include camera resolution, laser scanner duration, and ultrasonic detector emission frequency; determine one or more target anomaly locations from the one or more candidate anomaly locations based on the second detection data; and generate speed control commands and warning commands based on the one or more first anomaly parameters corresponding to the one or more target anomaly locations, and control a traction motor to drive the train through the tunnel based on the speed control commands and issue a warning based on the warning commands.

[0005] The invention includes a smart city tunnel crack emergency monitoring method, executed by an emergency monitoring and management platform. The method includes: acquiring train operation data and multiple sets of first detection data corresponding to the tunnel; determining one or more candidate anomaly locations and corresponding one or more first anomaly parameters within the tunnel based on the multiple sets of first detection data and the train operation data; determining re-inspection parameters based on the one or more first anomaly parameters, and controlling a robot to detect the one or more candidate anomaly locations according to the re-inspection parameters to obtain second detection data for the one or more candidate anomaly locations, wherein the re-inspection parameters include camera resolution, laser scanner duration, and ultrasonic detector emission frequency; determining one or more target anomaly locations from the one or more candidate anomaly locations based on the second detection data; and generating speed control commands and warning commands based on the one or more first anomaly parameters corresponding to the one or more target anomaly locations, and controlling a traction motor to drive the train through the tunnel based on the speed control commands and issuing a warning based on the warning commands.

[0006] The invention includes providing a computer-readable storage medium that stores computer instructions. When a computer reads the computer instructions from the storage medium, the computer executes the smart city tunnel crack emergency monitoring method described in the above embodiments.

[0007] The beneficial effects of this invention include, but are not limited to: (1) By integrating data collected by distributed fiber optic sensors in the tunnel and train operation data, early and comprehensive monitoring of tunnel structural anomalies is achieved, improving the timeliness and coverage of early warnings and overcoming the shortcomings of low efficiency in traditional manual inspections. (2) By determining the comprehensive difference parameters of the basic location, multiple potential structural anomalies can be quantified into a unified risk indicator, thereby achieving a more scientific and accurate assessment of the overall risk level of the basic location. At the same time, by setting difference conditions to screen the basic location, the efficiency and objectivity of the screening can be significantly improved. (3) By introducing a parameter update model, a secondary refined diagnosis of candidate anomaly locations is performed, significantly improving the accuracy and reliability of anomaly location judgment, greatly reducing the false alarms and missed alarms that may occur if only preliminary screening is relied upon, making the finally determined target anomaly location more credible. Attached Figure Description

[0008] This specification will be further described by way of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. These embodiments are not limiting; in these embodiments, the same reference numerals denote the same structures, wherein:

[0009] Figure 1This is a platform structure diagram of an IoT system for emergency monitoring of tunnel cracks in smart cities, as shown in some embodiments of this specification. Figure 2 This is an exemplary flowchart of an emergency monitoring method for tunnel cracks in smart cities, as shown in some embodiments of this specification. Figure 3 This is an exemplary schematic diagram illustrating the determination of one or more candidate anomaly locations according to some embodiments of this specification; Figure 4 This is an exemplary schematic diagram of a parameter update model based on some embodiments of this specification. Detailed Implementation

[0010] To more clearly illustrate the technical solutions of the embodiments in this specification, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are merely some examples or embodiments of this specification. For those skilled in the art, these drawings can be applied to other similar scenarios without creative effort. Unless obvious from the context or otherwise specified, the same reference numerals in the drawings represent the same structures or operations.

[0011] Flowcharts are used in this specification to illustrate the operations performed by the system according to embodiments of this specification. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, the steps can be processed in reverse order or simultaneously. Furthermore, other operations can be added to these processes, or one or more steps can be removed from them.

[0012] Figure 1 This is a platform structure diagram of an IoT system for emergency monitoring of tunnel cracks in smart cities, as shown in some embodiments of this specification.

[0013] In some embodiments, such as Figure 1 As shown, the smart city tunnel crack emergency monitoring IoT system 100 includes an emergency monitoring user platform 110, an emergency monitoring service platform 120, an emergency monitoring management platform 130, an emergency monitoring sensor network platform 140, and an emergency monitoring perception and control platform 150.

[0014] The Smart City Tunnel Crack Emergency Monitoring IoT System 100 includes an IoT model architecture to enable the efficient operation of large amounts of data within the system. Artificial intelligence models are applied to the IoT model architecture to assist in data perception and processing.

[0015] An emergency monitoring user platform refers to a platform used for interaction with users. In some embodiments, users include tunnel monitoring personnel, etc. The emergency monitoring user platform includes users' terminal devices (such as smartphones and computers).

[0016] An emergency monitoring service platform refers to a platform that provides monitoring information and services. In some embodiments, the emergency monitoring service platform is configured as a server or processor. The emergency monitoring service platform can interact bidirectionally with both the emergency monitoring management platform and the emergency monitoring user platform.

[0017] An emergency monitoring and management platform refers to a platform that comprehensively manages monitoring information. In some embodiments, the emergency monitoring and management platform is configured as a server and / or processor, etc. The emergency monitoring and management platform can communicate with the control system of the train.

[0018] An emergency monitoring sensor network platform refers to a platform that comprehensively manages monitoring sensor information. In some embodiments, the emergency monitoring sensor network platform is configured as a communication network or gateway, etc. The emergency monitoring sensor network platform can interact bidirectionally with the emergency monitoring management platform and the emergency monitoring perception and control platform.

[0019] An emergency monitoring and control platform refers to a platform that generates monitoring information and executes control information. In some embodiments, the emergency monitoring and control platform includes robots, detection devices carried by the train itself, and sensing devices installed in tunnels.

[0020] The robot is configured to detect candidate anomaly locations according to re-inspection parameters. In some embodiments, the robot detects candidate anomaly locations using structures such as cameras, laser scanners, and ultrasonic detectors mounted on its body.

[0021] In some embodiments, the sensing devices carried by the train itself include force-measuring wheels, vibration sensors, odometers, and force sensors built into the pantograph. The sensing devices installed in the tunnel include image recognition devices (such as high-speed cameras) and distributed optical fiber acoustic sensors (DAS).

[0022] In some embodiments, multiple distributed fiber optic acoustic sensors can be deployed along the longitudinal and circumferential directions inside the tunnel lining. The data collected by each distributed fiber optic acoustic sensor are the data from all locations along its deployment path.

[0023] In some embodiments, the smart city tunnel crack emergency monitoring IoT system 100 further includes a processor and a memory. The processor is configured to process information and / or data related to the smart city tunnel crack emergency monitoring IoT system 100. By way of example only, the processor includes a central processing unit (CPU), an application-specific integrated circuit (ASIC), an application-specific instruction processor (ASIP), a graphics processing unit (GPU), a physical processor (PPU), a digital signal processor (DSP), or any combination thereof.

[0024] The memory is configured to store information and / or data related to the smart city tunnel crack emergency monitoring IoT system 100. The memory includes mass storage, removable memory, and any combination thereof. The memory may be integrated into the processor.

[0025] For a detailed explanation of the foregoing, please refer to Figures 2 to 4 Related descriptions.

[0026] Some embodiments in this specification demonstrate that the smart city tunnel crack emergency monitoring IoT system 100 can form an information operation closed loop between various functional platforms and operate in a coordinated and regular manner under the unified management of the emergency monitoring and management platform, thereby realizing the informatization and intelligentization of urban tunnel crack monitoring.

[0027] Figure 2 This is an exemplary flowchart of a smart city tunnel crack emergency monitoring method according to some embodiments of this specification. In some embodiments, process 200 of the smart city tunnel crack emergency monitoring method can be executed by the emergency monitoring management platform 130 (hereinafter referred to as the management platform). For details on the smart city tunnel crack emergency monitoring IoT system and its various platforms, please refer to... Figure 1 And related explanations. For example... Figure 2 As shown, the process 200 of the emergency monitoring method for tunnel cracks in smart cities includes the following steps.

[0028] Step 210: Obtain the train operation data of the train in the tunnel and multiple sets of first detection data corresponding to the tunnel.

[0029] Tunnels can be underground passages in urban infrastructure, such as subway tunnels or railway tunnels. Trains are vehicles that travel inside tunnels; for example, trains can be electric multiple units (EMUs) or subway trains.

[0030] Train operation data refers to data reflecting the train's operating status when passing through a tunnel. In some embodiments, train operation data may include wheel-rail force, car body vibration data, pantograph-catenary relationship data, train positioning data, and train speed sequence when passing through the tunnel.

[0031] Wheel-rail forces can include vertical forces, lateral forces, or derailment coefficients generated when the vehicle wheels contact the rails. Wheel-rail forces can reflect track irregularities and changes in tunnel structure.

[0032] Vibration data can include the vibration characteristics (such as vibration acceleration, frequency, and amplitude) of parts like carriages and bogies during train operation. Changes in tunnel structure can alter the vibration characteristics of the train.

[0033] Pantograph-catenary interaction data can be data related to the interaction between the train's pantograph and the overhead contact system in the tunnel, such as the pantograph's vertical displacement and contact pressure, and the contact status between the pantograph and the contact system (e.g., offline status). Changes in the tunnel structure may cause changes in the contact system, thus affecting the pantograph-catenary interaction data.

[0034] Train positioning data can be the geographical environment of the train when it is traveling in a tunnel.

[0035] In some embodiments, the management platform can obtain the speed of the train at a preset frequency by using odometers or similar devices deployed on the train, thereby obtaining a speed sequence.

[0036] In some embodiments, the management platform can acquire train operation data through sensors mounted on the train itself. For example, the management platform acquires train body vibration data through vibration sensors installed on the car body or bogies, and acquires train positioning data through a global positioning system (GPS system, etc.). Another example is that the management platform monitors the contact status between the pantograph and the overhead contact line using a high-speed camera, and monitors the pantograph's vertical displacement and contact pressure using force sensors built into the pantograph. Yet another example is that the management platform acquires wheel-rail forces through piezoelectric sensors deployed inside the train's force-measuring wheels.

[0037] The first detection data refers to the data collected by the distributed fiber optic acoustic sensor. The distributed fiber optic acoustic sensor can characterize the vibration signal at the sensor's location by detecting changes in the collected optical signal.

[0038] In some embodiments, the management platform can acquire multiple sets of first detection data within a preset time period and at a preset acquisition frequency. The preset acquisition frequency and preset time period can be preset, such as a preset time period of 1 hour and a preset acquisition frequency of acquiring a set of first detection data every 10 seconds. The management platform can process the first detection data through digital filtering or other methods to remove environmental noise from the first detection data.

[0039] Step 220: Based on multiple sets of first detection data and train operation data, determine one or more candidate anomaly locations and one or more corresponding first anomaly parameters within the tunnel.

[0040] Candidate anomaly locations refer to tunnel areas where structural anomalies may occur. Structural anomalies can include water seepage, cracks, cavities, or deformation.

[0041] In some embodiments, the management platform can determine one or more candidate anomaly locations from multiple basic locations within the tunnel based on multiple sets of first detection data and train operation data.

[0042] A basic location refers to a unit area for monitoring a tunnel. In some embodiments, the management platform divides the tunnel area deployed by DAS into multiple basic locations according to a preset area size (such as a cube with a preset side length).

[0043] The first anomaly parameter refers to the multiple probabilities corresponding to various structural anomalies occurring at a candidate anomaly location. The first anomaly parameter can be represented in vector form, where each element in the vector corresponds to the probability of one type of structural anomaly occurring.

[0044] In some embodiments, for each basic location, the management platform selects multiple sets of first detection data corresponding to the basic location and train operation data when the train passes through the basic location from the train operation data and multiple sets of first detection data. The management platform extracts the mean and peak values ​​of the vibration signal from the multiple sets of first detection data corresponding to the basic location, and extracts the vibration amplitude and train speed change rate (e.g., the ratio of the train speed when passing through the basic location to the speed in the previous period in the speed sequence) from the train operation data corresponding to the basic location. The management platform performs normalization processing on the extracted data, and performs weighted summation on the normalized feature data (denoted as target feature data) according to preset weight coefficients to calculate the anomaly index corresponding to the basic location.

[0045] It is understandable that when a train passes through a basic location in a tunnel where structural anomalies may occur, it may experience abnormal vibrations or speed changes. Therefore, the probability of structural anomalies at the basic location can be characterized by the mean and peak values ​​of the vibration signal when the train passes, the vibration amplitude of the train, and the rate of change of the train's speed.

[0046] In some embodiments, the management platform can compare the anomaly indices of multiple basic locations with a first anomaly threshold, and identify one or more basic locations whose anomaly indices are higher than the first anomaly threshold as one or more candidate anomaly locations. The first anomaly threshold is preset.

[0047] In some embodiments, for each candidate anomaly location, the management platform can construct a matching vector based on target feature data, match a reference feature vector in a vector library that satisfies the matching conditions, and determine the first anomaly parameter of the candidate anomaly location based on the historical anomaly cases corresponding to the reference feature vectors. Matching conditions include a vector similarity greater than a similarity threshold. Vector similarity is negatively correlated with vector distance. Vector distance includes Euclidean distance, etc.

[0048] In some embodiments, the vector library is pre-configured based on historical data, including multiple reference feature vectors corresponding to each basic position and historical anomalies (the actual situation of various structural anomalies occurring at historical time, such as whether they occurred or not). The reference feature vectors can be feature vectors constructed based on reference feature data in the historical data.

[0049] In some embodiments, for each structural anomaly, the management platform counts the total number of matched reference feature vectors (denoted as the first number) and determines the number of reference feature vectors among the matched reference feature vectors that exhibit the structural anomaly based on historical anomaly data (denoted as the second number). The ratio of the second number to the first number is taken as the probability of the structural anomaly occurring. The management platform determines the probability of each structural anomaly occurring in the first anomaly parameter using the above method.

[0050] In some embodiments, the management platform can determine the second abnormal parameters corresponding to multiple basic locations in the tunnel based on multiple sets of first detection data and train operation data through an abnormal parameter model, and determine one or more candidate abnormal locations and one or more first abnormal parameters based on the second abnormal parameters.

[0051] In some embodiments, the anomaly parameter model can be a machine learning model, including a feature extraction layer and an anomaly detection layer. For example, the feature extraction layer is a Convolutional Neural Network (CNN) model, and the anomaly detection layer is a Support Vector Machine (SVM) model. The input to the feature extraction layer includes multiple sets of first detection data and train operation data, and the output includes feature vectors representing the multiple sets of first detection data and train operation data. The input to the anomaly detection layer includes the feature vectors output by the feature extraction layer, and the output includes second anomaly parameters corresponding to multiple basic positions.

[0052] The second anomaly parameter refers to the multiple probabilities corresponding to various structural anomalies occurring at a basic location. The form of the second anomaly parameter is similar to that of the first anomaly parameter. For example, the second anomaly parameter corresponding to a single basic location can be ((seepage, 10%), (crack, 10%), (void, 70%)), indicating that the probability of seepage and cracks occurring at the basic location is 10%, and the probability of voids occurring is 70%.

[0053] In some embodiments, the management platform can obtain an anomaly parameter model through joint training using methods such as gradient descent, based on multiple first training samples with first labels. For example, the management platform can input multiple first training samples into an initial feature extraction layer, use the feature vector output by the initial feature extraction layer as the input to an initial anomaly detection layer, construct a loss function based on the output of the initial anomaly detection layer and the first labels, iteratively update the parameters of the initial anomaly detection layer and the initial feature extraction layer based on the loss function, and terminate the iteration when the loss function meets the iteration completion condition, thus obtaining the trained anomaly parameter model. The iteration completion condition includes the convergence of the loss function or the number of iterations reaching a threshold.

[0054] The first training sample includes the first detection data and the train operation data of the sample period. The first label includes the actual occurrence of structural anomalies at multiple basic locations within the sample period.

[0055] In some embodiments, the first training sample and the first label are determined based on historical data. For example, the processor uses a historical time period as a sample time period and multiple sets of first detection data and train operation data within the sample time period as the first training samples. As another example, for multiple structural anomalies corresponding to a basic location, the processor can set the label of structural anomalies that actually occurred at that basic location within the sample time period to 1, and set the label of structural anomalies that did not occur to 0.

[0056] For example, the first label corresponding to a basic location can be ((seepage, 0), (crack, 1), (void, 0)), indicating that the basic location had a crack during the sample period, but no seepage or void occurred.

[0057] In some embodiments, the input to the anomaly parameter model (i.e., the input to the feature extraction layer) also includes deployment information of multiple sensing devices within the tunnel, train traction data, and the geographical environment of the tunnel. The sensing devices may be DAS (Digital Animation System).

[0058] The deployment information for multiple sensors includes data such as the physical layout and service life of the sensors. The physical layout refers to the specific installation location of the DAS within the tunnel (e.g., longitudinal mileage and circumferential angle). Deployment information can be pre-stored on the management platform.

[0059] Train traction data refers to data related to train traction, such as the traction force, current, voltage, and temperature of the traction motor during train operation. In some embodiments, the management platform can acquire train traction data through strain gauges, current sensors, voltage sensors, and temperature sensors deployed on the traction motor.

[0060] The geographical environment of a tunnel refers to information related to the environment in which the tunnel is located, such as geological structure, climate type, and surface traffic load. The geographical environment of a tunnel can be pre-stored in a management platform.

[0061] In some embodiments, the first training sample may also include the geographical environment of the sample tunnel, sample deployment information for the sample time period, and sample train traction data.

[0062] In some embodiments of this specification, by integrating deployment information of sensing devices, train traction data, and tunnel geographical environment, collaborative analysis of multi-dimensional data is achieved, optimizing the robustness of the anomaly parameter model and improving its accuracy.

[0063] In some embodiments, the loss function during the training process of the anomaly parameter model includes a weighted sum of multiple sub-loss terms, where one of the multiple sub-loss terms corresponds to one of the multiple basic positions.

[0064] In some embodiments, the loss function of the anomaly parameter model is the sum of the products of each sub-loss term and its corresponding weight. The weights of the sub-loss terms are positively correlated with the criticality of the base position corresponding to the sub-loss term.

[0065] The criticality of basic locations is determined manually. For example, basic locations located in the main load-bearing structure area of ​​a tunnel have higher criticality. Similarly, basic locations where structural anomalies have occurred have higher criticality. And basic locations traversing water-rich strata and with large structures above them also have higher criticality.

[0066] In some embodiments, the sub-loss term is represented by the following formula (1): (1) in, Represents a sub-loss term. This represents the first element in the output of the initial anomaly detection layer. This indicates the first element within the first tag. This represents the first element in the output of the initial anomaly detection layer. One element, This represents the nth element in the first tag. The number of various structural anomalies.

[0067] For example, if the multiple first labels corresponding to multiple basic positions are (A1, A2, ..., Am), then the multiple sub-loss terms in the loss function are (S1, S2, ..., Sm), where A1, A2, ..., Am represent the first labels corresponding to the first to the m-th basic positions, and S1, S2, ..., Sm represent the sub-loss terms corresponding to the first to the m-th basic positions, where m is the total number of basic positions.

[0068] Some embodiments in this specification achieve refined guidance of the model training process by constructing the loss function as a weighted sum of sub-loss terms corresponding to multiple basic positions.

[0069] In some embodiments, the management platform determines the basic locations containing one or more outlier values ​​among the second anomaly parameters corresponding to multiple basic locations as candidate anomaly locations, and uses the second anomaly parameters corresponding to the candidate anomaly locations as first anomaly parameters. If the probability of a structural anomaly occurring among the second anomaly parameters is greater than the corresponding probability threshold, then that structural anomaly is an anomaly value. The probability threshold is preset.

[0070] In some embodiments, the management platform can also determine one or more candidate anomaly locations from multiple basic locations based on comprehensive difference parameters and difference conditions. See [link to relevant documentation] for further details on this section. Figure 3 And its related descriptions.

[0071] Some embodiments in this specification determine candidate anomaly locations and their first anomaly parameters through anomaly parameter models, thereby achieving automated and intelligent preliminary identification of potential anomaly locations within tunnels. This provides clear and reliable targets for subsequent robot re-inspection, thus improving the accuracy and efficiency of locating anomaly locations.

[0072] Step 230: Based on one or more first anomaly parameters, determine re-inspection parameters, and control the robot to detect one or more candidate anomaly locations according to the re-inspection parameters, thereby obtaining second detection data for one or more candidate anomaly locations.

[0073] Re-inspection parameters refer to the parameters that guide the robot in its inspection process. Each candidate anomaly location corresponds to a single re-inspection parameter. In some embodiments, re-inspection parameters may include the image resolution of the camera, the scanning duration of the laser scanner, and the ultrasonic emission frequency of the ultrasonic detector, etc.

[0074] In some embodiments, the management platform can determine the re-inspection parameters based on the first anomaly parameter of the candidate anomaly location by querying a re-inspection parameter table. For example, for each structural anomaly in the first anomaly parameter, the management platform queries the probability range corresponding to the structural anomaly in the re-inspection parameter table and determines the reference re-inspection parameter corresponding to the probability range as the re-inspection parameter.

[0075] The re-inspection parameter table is pre-set based on historical experience, including multiple probability ranges corresponding to each type of structural anomaly and reference re-inspection parameters for each probability range. For example, since cracks are viewed through a camera, the probability range for cracks in the re-inspection parameter table could be 0 to 10% or 10% to 20%, etc., with each probability range corresponding to an image resolution.

[0076] The second type of detection data refers to the data collected by the robot according to the re-inspection parameters, such as images or videos, laser point cloud data, and ultrasonic waveform data.

[0077] Step 240: Based on the second detection data, determine one or more target anomaly locations from one or more candidate anomaly locations.

[0078] The target anomaly location refers to the tunnel area where structural anomalies have been identified.

[0079] In some embodiments, the management platform can extract features of the second detection data for each candidate anomaly location, determine the comprehensive anomaly value for each candidate anomaly location based on the features and the judgment rules, and determine the candidate anomaly location whose comprehensive anomaly value is greater than the second anomaly threshold as the target anomaly location.

[0080] The decision rules are used to determine whether a structural anomaly exists at a candidate anomaly location. In some embodiments, there may be multiple decision rules, each corresponding to the determination of a type of structural anomaly. Multiple decision rules and a second anomaly threshold are preset based on historical experience.

[0081] For example, the management platform uses feature extraction algorithms such as edge detection to extract various image features, such as edge features and grayscale features, from image data at candidate anomaly locations. The management platform iterates through multiple judgment rules, matching the extracted image features with the judgment rules. The more judgment rules the image features match, the larger the overall anomaly value.

[0082] For example, the judgment rule can be that if there are lines in the image features whose length, width, and edge contrast are within the corresponding judgment range, then the candidate abnormal position is suspected to have a crack.

[0083] In some embodiments, the management platform can also determine whether each candidate anomaly location is the target anomaly location based on the updated first anomaly parameter. (See also: [link to related content]) Figure 4 And its related descriptions.

[0084] Step 250: Based on one or more first abnormal parameters corresponding to one or more target abnormal locations, generate speed control commands and early warning commands, and control the traction motor to drive the train through the tunnel based on the speed control commands and issue early warnings based on the early warning commands.

[0085] In some embodiments, speed control commands include an upper limit on the rotational speed of the traction motor. Warning commands include warning levels (such as high, medium, and low levels).

[0086] In some embodiments, the management platform can determine the speed limit and warning level by querying a preset instruction table based on the number of one or more target anomaly locations and the average of the highest anomaly probabilities corresponding to one or more first anomaly parameters at each of the one or more target anomaly locations. The highest anomaly probability refers to the probability value with the highest probability among the values ​​of one or more first anomaly parameters at each target anomaly location.

[0087] The preset instruction table is pre-set based on historical experience, including multiple groups of data ranges and the corresponding rotational speed upper limits and warning levels for each group. A group of data ranges includes the quantity range of target abnormal positions and the average value range of the highest abnormal probability. Among them, the more the quantity of target abnormal positions, the larger the average value of the highest abnormal probability, the smaller the rotational speed upper limit, and the higher the warning level, so as to reduce the impact of the train operation on structural anomalies.

[0088] In some embodiments, the management platform can send the generated speed control instruction to the control system of the train, and the control system controls the rotational speed of the traction motor according to the rotational speed upper limit in the speed control instruction.

[0089] In some embodiments, the management platform performs different warning operations according to the warning level in the warning instruction. For example, for a low warning level, the management platform can send a warning message to the maintenance personnel through the emergency supervision user platform and incorporate the relevant data and judgment results into the daily report automatically generated by the management platform. For another example, for a medium warning level, the management platform can send a warning message to the maintenance supervisor and issue a warning prompt sound in departments such as the train dispatching center. For still another example, for a high warning level, the management platform can send a warning message to departments such as the tunnel management person in charge and the train dispatching center and make a phone confirmation, etc. The management platform can also represent different warning levels through different color identifications on its own display page.

[0090] In some embodiments of this specification, by integrating the data collected by the distributed optical fiber sensors in the tunnel and the train operation data, the early and comprehensive monitoring of tunnel structure anomalies is achieved, the timeliness and coverage of warnings are improved, and the defect of low efficiency of traditional manual inspections is overcome.

[0091] It should be noted that the above description of the process 200 of the smart city tunnel crack emergency supervision method is only for illustration and explanation, and does not limit the scope of application of this specification. For those skilled in the art, various corrections and changes can be made to the process 200 of the smart city tunnel crack emergency supervision method under the guidance of this specification. However, these corrections and changes are still within the scope of this specification.

[0092] Figure 3 It is an exemplary schematic diagram for determining one or more candidate abnormal positions shown in some embodiments of this specification.

[0093] Such as Figure 3As shown, for each basic location, the management platform can further execute step 310 to determine the comprehensive difference parameter corresponding to the basic location based on the second anomaly parameter, and then execute step 320 to determine whether the comprehensive difference parameter corresponding to the basic location meets the difference condition. If the comprehensive difference parameter meets the difference condition, the management platform executes step 330 to determine the basic location as a candidate anomaly location. If the comprehensive difference parameter does not meet the difference condition, the management platform executes step 340 to determine the basic location as a non-candidate anomaly location. For an explanation of the second anomaly parameter, basic location, and candidate anomaly location, please refer to [link to relevant documentation]. Figure 2 And its related descriptions.

[0094] The comprehensive difference parameter characterizes the probability of structural anomalies occurring at fundamental locations. It differs from the second anomaly parameter in that the comprehensive difference parameter is a single numerical value and does not include the probability of each structural anomaly occurring.

[0095] In some embodiments, the management platform calculates a weighted sum of the occurrence probabilities of each structural anomaly based on a second anomaly parameter corresponding to the basic location, to obtain a comprehensive difference parameter corresponding to the basic location. The weight of each structural anomaly is preset.

[0096] The difference condition is used to determine whether the base location is a candidate anomaly location. In some embodiments, the difference condition may include a comprehensive difference parameter greater than a first difference threshold. The first difference threshold is preset.

[0097] In some embodiments, the difference conditions may be determined based on the current time period, the tunnel's geographical environment, and the tunnel's average daily traffic volume. For a description of the tunnel's geographical environment, see [link to relevant documentation]. Figure 2 And its related descriptions.

[0098] In some embodiments, the management platform can divide a day into multiple time periods, such as peak traffic hours, off-peak traffic hours, and nighttime hours. The management platform obtains the current time through its internal system clock and determines the current time period to which it belongs.

[0099] The average daily traffic volume of a tunnel refers to the average number of vehicles passing through the tunnel each day within a statistical period (such as the past year). In some embodiments, the management platform can obtain the average daily traffic volume of the tunnel from an external platform (such as a traffic management department).

[0100] In some embodiments, the management platform can determine a first difference threshold by querying a preset table based on the current time period, the tunnel's geographical environment, and the tunnel's average daily traffic flow. The preset table can be pre-set based on historical experience and includes multiple sets of data and a corresponding first difference threshold for each set. Each set of data includes the time period, geographical environment, and average daily traffic flow. The time period, geographical environment, and average daily traffic flow data reflect the tunnel's usage intensity and importance; the higher the usage intensity and importance, the lower the first difference threshold.

[0101] In some embodiments of this specification, the difference conditions are adjusted dynamically based on factors such as the current time period and the actual conditions of the tunnel, thereby improving the flexibility in determining candidate anomaly locations and effectively reducing unnecessary false alarms.

[0102] In some embodiments, the difference condition is further determined based on the criticality of multiple basic locations, wherein a single basic location corresponds to a single difference condition. For an explanation of the criticality of basic locations, see step 220 and its related description.

[0103] In some embodiments, for each basic location, the management platform can determine a first difference threshold in the difference conditions corresponding to the basic location based on the criticality of the basic location and preset rules. The preset rules can be pre-set, including the correspondence between criticality and the first difference threshold; the higher the criticality, the lower the first difference threshold.

[0104] In some embodiments of this specification, specific differential conditions are determined for each basic location based on its criticality, enabling more rigorous monitoring of structurally more important or higher-risk tunnel areas and significantly improving the ability to detect early-stage tunnel hazards.

[0105] Some embodiments in this specification, by determining comprehensive difference parameters of basic locations, can quantify multiple potential structural anomalies into a unified risk indicator, thereby achieving a more scientific and accurate assessment of the overall risk level of basic locations. Furthermore, by setting difference conditions to screen basic locations, the efficiency and objectivity of the screening process can be significantly improved.

[0106] Figure 4 This is an exemplary schematic diagram of a parameter update model based on some embodiments of this specification.

[0107] In some embodiments, for each of the one or more candidate anomaly locations, the management platform can update the first anomaly parameter corresponding to each candidate anomaly location through the parameter update model 450 based on the first anomaly parameter 410 corresponding to each candidate anomaly location and the corresponding second detection data 420. The management platform can also determine whether each candidate anomaly location is a target anomaly location based on the updated first anomaly parameter 470.

[0108] In some embodiments, the parameter update model 450 can be a machine learning model, including a feature extraction layer 451 and an anomaly detection layer 452. For example, the feature extraction layer is a Long Short-Term Memory (LSTM) model or a Convolutional Neural Network (CNN) model, and the anomaly detection layer is a Support Vector Machine (SVM) model, etc.

[0109] In some embodiments, the input to the feature extraction layer includes second detection data 420 corresponding to each candidate anomaly location and a corresponding first anomaly parameter 410, and the output includes a feature vector 460 representing the second detection data 420 and the first anomaly parameter 410. The input to the anomaly judgment layer includes the feature vector 460, and the output includes the updated first anomaly parameter 470.

[0110] It is understandable that the first anomaly parameter is determined based on the first detection data and the train operation data. The parameter update model can update the first anomaly parameter based on the data detected by the robot in the field, thereby obtaining a first anomaly parameter that more accurately reflects the multiple probabilities corresponding to the occurrence of various structural anomalies (i.e., the updated first anomaly parameter 470).

[0111] In some embodiments, the parameter update model can be trained using multiple sets of second training samples with second labels, employing methods such as gradient descent. The second training samples include second detection data of the sample at the anomaly location and the sample's first anomaly parameter. The second label includes the updated first anomaly parameter corresponding to the anomaly location. The acquisition method and training process of the second training samples and second labels for the parameter update model are similar to those for the first training samples and first labels of the anomaly parameter model. For an explanation of the anomaly parameter model, please refer to [link to documentation]. Figure 2 And its related descriptions.

[0112] In some embodiments, the management platform can determine whether each candidate anomaly location is a target anomaly location based on the updated first anomaly parameter for each candidate anomaly location. For example, for each candidate anomaly location, the management platform can recalculate the comprehensive difference parameter (denoted as the first comprehensive difference parameter) based on the updated first anomaly parameter, and compare the first comprehensive difference parameter with a second difference threshold. If the first comprehensive difference parameter is greater than the second difference threshold, the candidate anomaly location is determined as the target anomaly location. The second difference threshold is preset to be less than the first difference threshold.

[0113] For an explanation of the first difference threshold and the calculation of the comprehensive difference parameter, please refer to [link / reference]. Figure 3And its related descriptions.

[0114] In some embodiments, the input to the feature extraction layer 451 includes multiple sets of first detection data 430 and train operation data 440 within a preset time period. The output of the anomaly detection layer includes multiple future anomaly parameters 480. For an explanation of the preset time period, see step 210 and its related description.

[0115] Multiple future anomaly parameters refer to the first anomaly parameter corresponding to multiple future time points. These multiple future time points are pre-set based on historical experience.

[0116] In some embodiments, the second training samples include multiple sets of sample first detection data and sample train operation data corresponding to the sample anomaly locations. The second label includes the occurrence status of structural anomalies at each sample time point corresponding to the sample anomaly location; the label for occurring structural anomalies is set to 1, and the label for non-occurring structural anomalies is set to 0. The sample time points are preset based on historical experience and are later than the time points when the sample first data were collected.

[0117] In some embodiments, the management platform can also determine whether each candidate anomaly location is the target anomaly location based on the updated first anomaly parameter and multiple future anomaly parameters.

[0118] In some embodiments, for each candidate anomaly location, the management platform can calculate a comprehensive difference parameter (denoted as a second comprehensive difference parameter) corresponding to each future anomaly parameter based on multiple future anomaly parameters corresponding to the candidate anomaly location. The management platform performs a weighted summation of the first comprehensive difference parameter and multiple second comprehensive difference parameters. If the sum is greater than a second difference threshold, the candidate anomaly location is determined as the target anomaly location.

[0119] In some embodiments, the weights of the first comprehensive difference parameter and each of the second comprehensive difference parameters can be preset based on historical experience. Specifically, the weight of the second comprehensive difference parameter is smaller the further back in time it is from the current moment.

[0120] In some embodiments of this specification, by comprehensively considering the updated first anomaly parameter and multiple future anomaly parameters, a "trend" assessment of possible structural anomalies can be performed, potential risk points that may deteriorate rapidly can be identified in advance, and the predictability and proactivity of tunnel safety management can be improved.

[0121] In some embodiments of this specification, by adding more comprehensive inputs to the parameter update model, the output of the parameter update model can be comprehensively judged by combining the detection data of the abnormal location and the background environmental load, thereby significantly improving the robustness and accuracy of the abnormal situation judgment.

[0122] In some embodiments of this specification, by introducing a parameter update model, a secondary refined diagnosis of candidate anomaly locations is performed, which significantly improves the accuracy and reliability of anomaly location judgment, greatly reduces the false alarms and false negatives that may occur if only preliminary screening is relied upon, and makes the finally determined target anomaly location more reliable.

[0123] This specification provides one or more embodiments of a computer-readable storage medium that stores computer instructions. When a computer reads the computer instructions from the storage medium, the computer executes the smart city tunnel crack emergency monitoring method described in the above embodiments.

[0124] Furthermore, certain features, structures, or characteristics in one or more embodiments of this specification may be appropriately combined.

[0125] In some embodiments, numbers describing the quantity of components and attributes are used. It should be understood that such numbers used in the description of embodiments are modified in some examples with the terms "approximately," "approximately," or "generally." Unless otherwise stated, "approximately," "approximately," or "generally" indicates that the numbers are allowed to vary by ±20%. Accordingly, in some embodiments, the numerical parameters used in the specification and claims are approximate values, which may be changed depending on the characteristics required by individual embodiments. In some embodiments, numerical parameters should take into account specified significant digits and employ a general method of digit reservation. Although the numerical ranges and parameters used to confirm their breadth of range in some embodiments of this specification are approximate values, in specific embodiments, such values ​​are set as precisely as feasible.

[0126] If there is any inconsistency or conflict between the descriptions, definitions, and / or terms used in the materials referenced in this specification and the content described in this specification, the descriptions, definitions, and / or terms used in this specification shall prevail.

Claims

1. A smart city tunnel crack emergency monitoring IoT system, characterized in that, The Internet of Things system includes an emergency monitoring and management platform, which is configured as follows: Acquire train operation data inside the tunnel and multiple sets of first detection data corresponding to the tunnel; Based on the multiple sets of first detection data and the train operation data, one or more candidate anomaly locations and one or more corresponding first anomaly parameters are determined within the tunnel. Based on the one or more first anomaly parameters, re-inspection parameters are determined, and the robot is controlled to detect the one or more candidate anomaly locations according to the re-inspection parameters to obtain second detection data of the one or more candidate anomaly locations. The re-inspection parameters include the resolution of the camera, the duration of the laser scanner, and the emission frequency of the ultrasonic detector. Based on the second detection data, one or more target anomaly locations are determined from the one or more candidate anomaly locations; as well as Based on one or more first abnormal parameters corresponding to the one or more target abnormal locations, a speed control command and a warning command are generated. Based on the speed control command, the traction motor is controlled to drive the train through the tunnel, and a warning is issued based on the warning command.

2. The Internet of Things system according to claim 1, characterized in that, The emergency monitoring and management platform is further configured as follows: Based on the multiple sets of first detection data and the train operation data, second anomaly parameters corresponding to multiple basic locations in the tunnel are determined by an anomaly parameter model, wherein the anomaly parameter model is a machine learning model. as well as Based on the second anomaly parameter, the one or more candidate anomaly locations and the one or more first anomaly parameters are determined.

3. The Internet of Things system according to claim 2, characterized in that, The emergency monitoring and management platform is further configured as follows: Based on the second anomaly parameter, determine the comprehensive difference parameters corresponding to the plurality of basic positions respectively; and Based on the comprehensive difference parameters and difference conditions, one or more candidate anomaly locations are determined from the plurality of basic locations.

4. The Internet of Things system according to claim 1, characterized in that, The emergency monitoring and management platform is further configured as follows: For each of the one or more candidate anomaly locations: Based on the second detection data and the corresponding first anomaly parameter corresponding to each candidate anomaly location, the first anomaly parameter corresponding to each candidate anomaly location is updated through a parameter update model, wherein the parameter update model is a machine learning model. as well as Based on the updated first anomaly parameter, determine whether each candidate anomaly location is the target anomaly location.

5. The Internet of Things system according to claim 4, characterized in that, The emergency monitoring and management platform is further configured as follows: Based on the updated first anomaly parameter and multiple future anomaly parameters, it is determined whether each candidate anomaly location is the target anomaly location.

6. A method for emergency monitoring of tunnel cracks in smart cities, characterized in that, The method is executed by the emergency monitoring and management platform, and the method includes: Acquire train operation data inside the tunnel and multiple sets of first detection data corresponding to the tunnel; Based on the multiple sets of first detection data and the train operation data, one or more candidate anomaly locations and one or more corresponding first anomaly parameters are determined within the tunnel. Based on the one or more first anomaly parameters, re-inspection parameters are determined, and the robot is controlled to detect the one or more candidate anomaly locations according to the re-inspection parameters to obtain second detection data of the one or more candidate anomaly locations. The re-inspection parameters include the resolution of the camera, the duration of the laser scanner, and the emission frequency of the ultrasonic detector. Based on the second detection data, one or more target anomaly locations are determined from the one or more candidate anomaly locations; and Based on one or more first abnormal parameters corresponding to the one or more target abnormal locations, a speed control command and a warning command are generated. Based on the speed control command, the traction motor is controlled to drive the train through the tunnel, and a warning is issued based on the warning command.

7. The method according to claim 6, characterized in that, The step of determining one or more candidate anomaly locations and corresponding one or more first anomaly parameters within the tunnel based on the multiple sets of first detection data and the train operation data includes: Based on the multiple sets of first detection data and the train operation data, second anomaly parameters corresponding to multiple basic locations within the tunnel are determined using an anomaly parameter model, which is a machine learning model; and Based on the second anomaly parameter, the one or more candidate anomaly locations and the one or more first anomaly parameters are determined.

8. The method according to claim 7, characterized in that, Determining the one or more candidate anomaly locations and the one or more first anomaly parameters based on the second anomaly parameter includes: Based on the second anomaly parameter, determine the comprehensive difference parameters corresponding to the plurality of basic positions respectively; and Based on the comprehensive difference parameters and difference conditions, one or more candidate anomaly locations are determined from the plurality of basic locations.

9. The method according to claim 6, characterized in that, The method further includes: For each of the one or more candidate anomaly locations: Based on the second detection data and the corresponding first anomaly parameter corresponding to each candidate anomaly location, the first anomaly parameter corresponding to each candidate anomaly location is updated through a parameter update model, wherein the parameter update model is a machine learning model; and Based on the updated first anomaly parameter, determine whether each candidate anomaly location is the target anomaly location.

10. A computer-readable storage medium, characterized in that, The storage medium stores computer instructions. When the computer reads the computer instructions from the storage medium, the computer executes the method as described in any one of claims 6-9.

Images