Rail transit state early warning method and system for multi-source perception terminal

Abstract

This invention discloses a method and system for early warning of rail transit status using multi-source sensing terminals. The method includes: real-time collection of rail transit status data via multi-source sensing terminals deployed along the rail transit line; preprocessing the rail transit status data to generate a status dataset; constructing a rail transit operation status feature matrix based on the status dataset; analyzing the status feature matrix using the isolated forest algorithm to identify abnormal states; classifying the identified abnormal states according to preset abnormality level rules to determine the abnormality level; and comparing the abnormality level with a risk threshold to generate status warning information. This invention achieves real-time and accurate identification of rail transit status data and refined risk level warnings by deploying multi-source sensing terminals and utilizing the isolated forest algorithm.

Description

Technical Field

[0001] This invention relates to the technical field of rail transit condition monitoring and intelligent early warning, and in particular to a rail transit condition early warning method and system for multi-source sensing terminals. Background Technology

[0002] The safe and stable operation of rail transit systems is directly related to the safety of people's lives and property and the harmonious development of society. Therefore, real-time and efficient rail transit condition monitoring and early warning technologies are receiving increasing attention. Currently, rail transit condition monitoring mainly relies on the deployment of various types of sensors along the line. Related technologies are constantly being upgraded and optimized to improve the accuracy, reliability and real-time performance of monitoring.

[0003] CN109862532A discloses a method for optimizing the layout of multi-sensor nodes in rail transit condition monitoring. This technology optimizes the sensor layout by establishing a node optimization layout weight function and a node utility function model, which effectively improves the monitoring and information transmission capabilities. Although this method optimizes the layout of sensor nodes, it does not involve the fusion of multi-source heterogeneous data in complex scenarios and the method for accurate identification of abnormal states, making it difficult to achieve refined risk level assessment.

[0004] CN115009324A discloses a method for early warning of unstable network accidents in rail transit based on cloud-edge collaboration. It improves the stability of network communication and the real-time performance of accident warnings through TCP communication. Although this method can improve communication stability, it does not make full use of data mining and intelligent algorithms for abnormal state identification, resulting in insufficient timeliness and accuracy of state warnings.

[0005] The aforementioned existing technologies still have shortcomings in real-time processing of multi-source heterogeneous data and in the accuracy of state recognition and classification under complex working conditions. For example, the data fusion accuracy is not high, the speed of abnormal state recognition is slow, and the level of precision of early warning information is insufficient.

[0006] Therefore, existing rail transit status early warning technologies suffer from insufficient accuracy in multi-source heterogeneous data fusion, low efficiency in real-time analysis of abnormal states, and low level of refinement in status early warning information. To address these technical deficiencies, this invention utilizes the isolated forest algorithm by deploying multi-source sensing terminals to achieve real-time and accurate identification of rail transit status data and refined risk level early warning. Summary of the Invention

[0007] The purpose of this section is to outline some aspects of the embodiments of the present invention and to briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section, as well as in the abstract and title of the present application, to avoid obscuring the purpose of this section, the abstract and title of the invention. Such simplifications or omissions shall not be used to limit the scope of the present invention.

[0008] In view of the aforementioned existing problems, the present invention is proposed.

[0009] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0010] Real-time rail transit status data is collected by multi-source sensing terminals deployed along the rail transit line, and the rail transit status data is preprocessed to generate a status dataset.

[0011] Based on the aforementioned state dataset, a rail transit operation state feature matrix is ​​constructed, and the isolated forest algorithm is used to analyze the state feature matrix to identify abnormal states.

[0012] The identified abnormal states are classified according to preset abnormality level rules to determine the abnormality level;

[0013] The anomaly level is compared and analyzed with the risk threshold to generate a status warning.

[0014] As a preferred embodiment of the rail transit status early warning method for multi-source sensing terminals described in this invention, the rail transit status data includes at least track structure status data, track vibration data, track temperature data, vehicle operation status data, surrounding environment monitoring data, and vehicle communication data.

[0015] As a preferred embodiment of the rail transit status early warning method for multi-source sensing terminals described in this invention, the rail transit status data is preprocessed to generate a status dataset, including:

[0016] Data cleaning is performed on the real-time collected rail transit status data to remove invalid and redundant data;

[0017] Interpolation methods are used to fill in missing data points;

[0018] Then, wavelet denoising algorithm is used to reduce noise interference in the data;

[0019] Normalization is used to unify the denoised data to the same scale, generating a state dataset.

[0020] As a preferred embodiment of the rail transit status early warning method for multi-source sensing terminals described in this invention, a rail transit operation status feature matrix is ​​constructed based on the status dataset, including:

[0021] Feature parameters are extracted from the state dataset. The feature parameters include at least the track vibration spectrum characteristics, track temperature gradient, real-time vehicle speed and acceleration, track structure stability index, and environmental parameter fluctuation amplitude.

[0022] The feature parameters are arranged in a matrix according to the time window to form a feature matrix of rail transit operation status with the feature parameters as columns and the time series as rows.

[0023] As a preferred embodiment of the rail transit state early warning method for multi-source sensing terminals described in this invention, the state feature matrix is ​​analyzed using the isolated forest algorithm, including:

[0024] Define the state feature matrix as follows:

[0025]

[0026] Where, x i Let m represent the feature vector of the i-th time window sample, m be the total number of samples in the current feature matrix, and d be the feature dimension.

[0027] Randomly select features and randomly determine feature split points, recursively partition the sample space, and construct an isolated tree structure;

[0028] The formula for calculating anomaly scores is defined as follows:

[0029]

[0030] Where h(x) represents the path length of sample point x, E(h(x)) represents the expected value of the path length, and c(m) is the adjustment coefficient of the path length when the number of samples m is given;

[0031] When the abnormal score is greater than the abnormal threshold, it is considered an abnormal state.

[0032] As a preferred embodiment of the rail transit condition early warning method for multi-source sensing terminals described in this invention, the levels are divided according to the magnitude of the anomaly score s(x), wherein:

[0033] When the anomaly score satisfies 0.7 ≤ s(x) ≤ 1, it is considered a severe anomaly;

[0034] When the anomaly score satisfies 0.4 ≤ s(x) < 0.7, it is considered a moderate anomaly;

[0035] When the anomaly score satisfies 0.1 ≤ s(x) < 0.4, it is considered a minor anomaly.

[0036] When the abnormal score satisfies 0≤s(x)<0.1, it is considered a normal state.

[0037] As a preferred embodiment of the rail transit status early warning method for multi-source sensing terminals described in this invention, the generation of status early warning information includes:

[0038] Define risk thresholds;

[0039] When the anomaly level is severe, if the anomaly score s(x) > the risk threshold, a high-level status warning message is generated.

[0040] When the anomaly level is medium or slight, if the anomaly score s(x) ≤ the risk threshold, a low-level status warning message is generated.

[0041] The generated status warning information includes at least the location of the anomaly, timestamp, anomaly level, anomaly parameters, and suggested handling measures. The generated status warning information is then pushed to the security maintenance terminal and dispatch center for timely processing.

[0042] As a preferred embodiment of the rail transit status early warning system for multi-source sensing terminals described in this invention, it includes: one or more processors;

[0043] The memory stores operable instructions that, when executed by the one or more processors, cause the one or more processors to perform operations, including the flow of the rail transit condition early warning method for multi-source sensing terminals as described above.

[0044] As a preferred embodiment of the computer-readable medium for storing software according to the present invention, the software includes instructions executable by one or more computers, the instructions causing the one or more computers to perform operations, the operations including the flow of the rail transit condition early warning method for multi-source sensing terminals as described above.

[0045] The beneficial effects of this invention are:

[0046] 1. It fully utilizes the synergistic effect of multi-source heterogeneous terminals in the data acquisition process, ensuring the real-time, completeness and multi-dimensional coverage of rail transit operation data acquisition. It can effectively eliminate the problems of one-sided, delayed or excessive redundant data acquisition from single-type sensor data. Through data preprocessing, invalid noise and outliers in the data are removed, improving the accuracy and stability of the dataset and laying a solid data foundation for subsequent accurate analysis.

[0047] 2. By efficiently organizing data in a matrix manner and introducing the isolated forest algorithm, the algorithm leverages its fast, efficient and high-dimensional data anomaly identification characteristics to achieve real-time and accurate anomaly monitoring and diagnosis of rail transit status. This solves the problems of low efficiency and high false judgment rate of traditional anomaly detection algorithms when processing high-dimensional data, thereby significantly improving the accuracy and real-time performance of rail transit anomaly analysis.

[0048] 3. By establishing a systematic and refined anomaly classification mechanism, multi-level classification is achieved based on the specific characteristics and impact of the anomaly state, thereby meeting the needs of different rail transit operation scenarios for accurate early warning. This effectively makes up for the problem that the existing technology is too coarse and not clear and specific in the classification of anomalies, and achieves the beneficial effect of more clear and specific anomaly early warning levels and more accurate risk assessment.

[0049] 4. By directly comparing the anomaly level with the risk threshold quantitatively, the problem of early warning delays or false alarms caused by subjective human judgment is effectively avoided, enabling precise control of rail transit operation risks. This ensures that rail transit operation risks are detected and communicated to management personnel in a timely manner, thereby effectively improving the safety and stability of rail transit operations. Attached Figure Description

[0050] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0051] Figure 1 This is a flowchart illustrating the rail transit status early warning method for multi-source sensing terminals as shown in this invention. Detailed Implementation

[0052] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0053] Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort should fall within the scope of protection of this invention.

[0054] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0055] According to an embodiment of the present invention, in combination Figure 1 The flowchart shown illustrates a method for early warning of rail transit conditions using multi-source sensing terminals, which specifically includes the following steps:

[0056] S1. Real-time collection of rail transit status data is achieved through multi-source sensing terminals deployed along the rail transit line, and the rail transit status data is preprocessed to generate a status dataset. Note that the following points should be noted in this step:

[0057] Track structure monitoring terminals, triaxial vibration accelerometers, fiber optic temperature sensors, vehicle operation monitoring units, environmental meteorological stations, and vehicle-to-ground communication acquisition modules are deployed in a grid pattern along the track line and key sections (bridges, tunnels, turnouts, and station sections). Each terminal is connected to the edge computing node through the 5G-FRMCS private network to achieve time synchronization and data reporting.

[0058] A UTC timestamp is added to each real-time reported data. Edge nodes immediately discard out-of-order data packets and record them in the log to ensure the integrity and timing consistency of subsequent processing.

[0059] Perform field validity checks, physical interval threshold checks, and duplicate sample removal on track structure status data, vibration data, temperature data, vehicle speed / acceleration, environmental monitoring data, and vehicle communication KPIs to remove invalid and redundant data.

[0060] For the cleaned multi-field time series, a hierarchical interpolation strategy is adopted according to the missing segment length Δt, where linear interpolation is used when Δt ≤ 3 sampling periods;

[0061] Cubic spline interpolation is used when 3 < Δt ≤ 10;

[0062] When Δt>10, it is marked as unrecoverable and enters the data quality alarm queue;

[0063] Furthermore, the vibration and temperature sequences were decomposed into three levels using Daubechies-4 wavelets, and the signals were reconstructed after removing high-frequency noise using an adaptive soft thresholding function. For vehicle operation and communication KPIs, Coiflet-3 two-level decomposition was used for noise reduction to preserve step edge information.

[0064] For different physical quantities, interval normalization Min-Max is uniformly used to map to the [0,1] interval. When the features have a long tail distribution, Z-Score standardization is used to ensure the comparability of features of the same dimension in the subsequent feature matrix.

[0065] Using 1 second as the basic time window, the preprocessed data is aggregated into a structured record R(t) through a strategy of "timestamp alignment - field merging - primary key deduplication". <ID sensor Type, Value norm ,t>, N consecutive time windows are spliced ​​together to form a sliding window data block D N ={R(t) iWrite the data into a distributed columnar storage to obtain the state dataset, which provides the basis for constructing the running state feature matrix in step S2.

[0066] As an example, rail transit status data includes at least track structure status data, track vibration data, track temperature data, vehicle operation status data, surrounding environment monitoring data, and on-board communication data.

[0067] For example, taking a 10km section of an urban subway as an example, a set of integrated track structure and vibration terminal is deployed every 50m, and a set of fiber optic temperature sensing chain is deployed every 200m. The onboard communication KPIs are reported by the train's ATP equipment at a frequency of 10Hz, with a uniform sampling period of 100ms, generating approximately 8×10 KPIs per day. 8 1 record.

[0068] It should be noted that the above implementation process ensures that the multi-source heterogeneous rail transit status data meets the requirements of high reliability and high consistency after noise suppression, missing data repair and scale unification, providing a data foundation for the isolated forest algorithm and improving the accuracy of subsequent anomaly identification and the timeliness of early warning.

[0069] S2. Based on the state dataset, construct a feature matrix of rail transit operation state, and use the isolated forest algorithm to analyze the state feature matrix to identify abnormal states. Note that the following points should be noted in this step:

[0070] Extract feature parameters from the state dataset. The feature parameters include at least the track vibration spectrum characteristics, track temperature gradient, real-time vehicle speed and acceleration, track structure stability index, and environmental parameter fluctuation amplitude.

[0071] The feature parameters are arranged in a matrix according to the time window to form a feature matrix of rail transit operation status with the feature parameters as columns and the time series as rows;

[0072] Define the state feature matrix as follows:

[0073]

[0074] Where, x i Let m represent the feature vector of the i-th time window sample, m be the total number of samples in the current feature matrix, and d be the feature dimension.

[0075] Furthermore, m are randomly selected from X without replacement. s Samples (e.g., m) s =256) to construct a single isolated tree;

[0076] Recursively execute on the sampling set:

[0077] ① Randomly select a one-dimensional feature q from the feature set {1, …, d};

[0078] ② Randomly select a splitting point p within the value range [min(X q ), max(X q )] of this feature;

[0079] ③ Divide the sample space according to X q < p and X q ≥ p and recurse until the depth limit or the number of samples in the node is 1;

[0080] Generate n t isolated trees (e.g., n t = 100) to form a forest;

[0081] Furthermore, for each sample x ∈ X, calculate its path length h j (x) in the j-th tree, and find the average path length and the expected path length;

[0082] Define the abnormal score calculation formula as:

[0083]

[0084] where h(x) represents the path length of the sample point x, E(h(x)) represents the expected value of the path length, and c(m) is the adjustment coefficient of the path length for a given number of samples m;

[0085] When the abnormal score is greater than the abnormal threshold, it is in an abnormal state:

[0086] When s(x) approaches 1, x is easily isolated in the tree structure and is more likely to be abnormal;

[0087] When s(x) approaches 0, x belongs to the normal dense area in the tree structure.

[0088] S3. Classify the identified abnormal states according to the preset abnormal level rules to determine the abnormal level. It should be noted in this step that the level is divided according to the size of the abnormal score s(x):

[0089] Perform a rolling statistics on all the sample abnormal scores output in step S2, and calculate their mean μ s and standard deviation σ s ;

[0090] Set three levels of adaptive thresholds according to the historical normal sample distribution:

[0091] τ L = min(μ s+1.5σ s ,0.60)

[0092] τ M =min(μ s +2.5σ s ,0.80)

[0093] τ H =min(μ s +3.5σ s ,0.95)

[0094] And ensure 0 < τ L <τ M <τ H <1;

[0095] For any sample x, the anomaly score s(x):

[0096]

[0097] Among them, s i Let m be the anomaly score for the sample in the i-th time window, m be the total number of samples in the current statistical window, and τ be the anomaly score. L τ is the dynamic threshold for minor anomalies. M τ is the dynamic threshold for moderate anomalies. H The dynamic threshold for severe anomalies is Level(x), which is the anomaly level determination result for sample x.

[0098] Will <t,ID sensor ,s(x),Level> is written into the abnormal event stream to provide input for the risk threshold comparison and early warning information generation in the subsequent step S4.

[0099] It should be noted that this embodiment, through the combination strategy of the above-mentioned adaptive statistical threshold and four-level classification, not only avoids the failure of fixed thresholds in different operating scenarios, but also maintains the simplicity and clarity of the level division, realizing rapid, precise and interpretable classification of rail transit operation anomalies, and providing a reliable risk classification basis for the traffic control center.

[0100] S4. Compare and analyze the anomaly level with the risk threshold to generate a status warning. Note that the following points should be noted in this step:

[0101] Statistical analysis of historical anomaly score sequences in the same area over the past M days {s h}, thus obtaining the mean μ h With standard deviation σ h ;

[0102] The dynamic risk threshold R is defined based on the operator's risk tolerance coefficient λ∈[1,3]. th =μ h +λσh ;

[0103] Given the anomaly score s(x) and anomaly level L(x) of the current sample x:

[0104]

[0105] Among them, High indicates a high-level warning, Low indicates a low-level warning, and Info indicates an information prompt;

[0106] Generate structured early warning records<t,Pos,L,s,Alert,θ> ;

[0107] Where t is the UTC timestamp, Pos is the orbital mileage / GPS coordinates, and θ is the set of anomaly parameters (such as vibration RMS, temperature gradient, etc.).

[0108] Based on Alert and L, query the knowledge base K to obtain the corresponding action Action(Alert,L) and append it to the record;

[0109] JSON messages are pushed to security maintenance terminals along the route via the MQTT interface;

[0110] Push a WebSocket message to the SCADA screen at the traffic control center and trigger an audible and visual alarm;

[0111] All early warnings are synchronously written to a historical event database (such as Kafka+HBase) for post-event analysis;

[0112] Among them, the sending delay is ≤1s and the historical retention period is ≥3 years.

[0113] For example, taking the interval K12+000—K12+500 as an example, the historical data for M=7 days yields μ. h =0.64,σ h =0.08, the operations and maintenance department sets λ=2, then R th =0.64 + 2 × 0.08 = 0.80,

[0114] If sample x is detected at position K12+230 at 14:32:07 on the same day, its anomaly score s(x) = 0.87, and its grade L(x) = A, since s(x) > R th The result is Alert(x) = High;

[0115] The system immediately generates a JSON alert, as shown in the example below:

[0116]

[0117] And simultaneously push it to the traffic control center;

[0118] It should be noted that, through the implementation of the above-described methods, this dynamic comparison-push mechanism effectively reduces the delay of manual identification and ensures the safety and controllability of rail transit operations.

[0119] The aforementioned methods for preprocessing rail transit status data and for extracting features from the status dataset can be implemented using existing technologies and methods, and will not be elaborated upon in this example.

[0120] In addition to the above embodiments, other aspects of the present invention also propose a rail transit status early warning system for multi-source sensing terminals, including: one or more processors and a memory.

[0121] The memory is used to store operable instructions that, when executed by the one or more processors, cause the one or more processors to perform operations, including the flow of a rail transit condition early warning method for a multi-source sensing terminal according to the foregoing embodiments, particularly... Figure 1 The flowchart of the method is shown.

[0122] Other aspects disclosed in the embodiments of the present invention also propose a computer-readable medium for storing software including instructions executable by one or more computers, which, upon execution, cause the one or more computers to perform operations including the flow of a rail transit condition early warning method for a multi-source sensing terminal according to the foregoing embodiments, particularly... Figure 1 The flowchart of the method is shown.

[0123] It should be recognized that embodiments of the present invention may be implemented or carried out by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer-readable storage medium.

[0124] The method can be implemented using standard programming techniques, including a non-transitory computer-readable storage medium configured with a computer program in the computer program, wherein the storage medium is configured such that the computer operates in a specific and predefined manner.

[0125] Each program can be implemented in a high-level procedural or object-oriented programming language to communicate with the computer system; however, if required, the program can be implemented in assembly or machine language.

[0126] In any case, the language can be either compiled or interpreted.

[0127] Furthermore, for this purpose, the program can run on programmed application-specific integrated circuits.

[0128] The processes described herein (or variations and / or combinations thereof) can be executed under the control of one or more computer systems configured with executable instructions, and can be implemented by hardware or a combination thereof as code (e.g., executable instructions, one or more computer programs, or one or more applications) that commonly executes on one or more processors. The computer program includes a plurality of instructions executable by one or more processors.

[0129] Furthermore, the method can be implemented in any suitable computing platform, including but not limited to personal computers, minicomputers, mainframes, workstations, networked or distributed computing environments, standalone or integrated computer platforms, or in communication with charged particle tools or other imaging devices.

[0130] Various aspects of the present invention can be implemented in machine-readable code stored on a non-transitory storage medium or device, whether portable or integrated into a computing platform, such as a hard disk, optical read and / or write storage medium, RAM, ROM, etc., such that it can be read by a programmable computer, and when the storage medium or device is read by the computer, it can be used to configure and operate the computer to perform the processes described herein.

[0131] Furthermore, machine-readable code, or parts thereof, can be transmitted via wired or wireless networks.

[0132] When such media includes instructions or programs that combine with a microprocessor or other data processor to implement the steps described above, the invention described herein includes these and other different types of non-transitory computer-readable storage media.

[0133] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A rail transit state early warning method for multi-source perception terminals, characterized in that, include: Real-time rail transit status data is collected by multi-source sensing terminals deployed along the rail transit line, and the rail transit status data is preprocessed to generate a status dataset. Based on the aforementioned state dataset, a rail transit operation state feature matrix is ​​constructed, including: extracting feature parameters from the state dataset, wherein the feature parameters include at least track vibration spectrum features, track temperature gradient, real-time vehicle speed and acceleration, track structure stability index, and environmental parameter fluctuation amplitude; and arranging each feature parameter in a matrix according to a time window to form a rail transit operation state feature matrix with feature parameters as columns and time series as rows. The isolated forest algorithm is then used to analyze the state feature matrix and identify abnormal states; including: Define the state feature matrix as follows: in, Indicates the first Feature vectors of samples within a time window d represents the total number of samples in the current feature matrix, and d represents the feature dimension. Features are randomly selected and feature splitting points are randomly determined. The sample space is recursively partitioned to construct an isolated tree structure. The formula for calculating anomaly scores is defined as follows: in, Represents sample points Path length, This represents the expected value of the path length. For a given number of samples Adjustment factor for path length; When the abnormal score is greater than the abnormal threshold, it is considered an abnormal state. The identified abnormal state is classified according to preset abnormal level rules to determine an abnormal level; and the abnormal level is divided according to the size of the abnormal score wherein: When the abnormality score satisfies a severe abnormality; When the abnormality score satisfies a medium abnormality; When the abnormality score satisfies a slight abnormality; When the abnormal score satisfies normal state; The anomaly level is compared and analyzed with the risk threshold to generate a status warning message; specifically including: Define risk threshold: Statistically analyze the historical anomaly score sequence of the same area over the past M days. to obtain the mean with standard deviation Based on the operator's risk tolerance coefficient Define dynamic risk thresholds ; When the anomaly level is severe, the anomaly score is... When the risk threshold is reached, a high-level status warning message is generated; When the anomaly level is moderate or slight, the anomaly score is... If the risk threshold is ≤, a low-level status warning message will be generated; The generated status warning information includes at least the location of the anomaly, timestamp, anomaly level, anomaly parameters, and suggested handling measures. The generated status warning information is then pushed to the security maintenance terminal and traffic control center for timely processing. 2.The rail transit state early warning method for multi-source perception terminals according to claim 1, wherein, The rail transit status data includes at least track structure status data, track vibration data, track temperature data, vehicle operation status data, surrounding environment monitoring data, and on-board communication data. 3.The rail transit state early warning method for multi-source perception terminals according to claim 1 or 2, characterized in that, The rail transit status data is preprocessed to generate a status dataset, including: Data cleaning is performed on the real-time collected rail transit status data to remove invalid and redundant data; Interpolation methods are used to fill in missing data points; Then, wavelet denoising algorithm is used to reduce noise interference in the data; Normalization is used to unify the denoised data to the same scale, generating a state dataset.

4. A rail transit state early warning system for multi-source perception terminals, characterized in that, include: One or more processors; The memory stores operable instructions that, when executed by the one or more processors, cause the one or more processors to perform operations, including the flow of the rail transit condition early warning method for multi-source sensing terminals as described in any one of claims 1 to 3.

5. A computer readable medium storing software, characterized in that: The software includes instructions executable by one or more computers, which cause the one or more computers to perform operations, including the flow of the rail transit condition early warning method for multi-source sensing terminals as described in any one of claims 1 to 3.

Images