Urban rail transit multi-signal real-time processing method and system

Abstract

The invention discloses an urban rail transit multi-signal real-time processing method and an urban rail transit multi-signal real-time processing system, which relate to the technical field of traffic signal processing, and are characterized in that: through operation state set driven signal segment processing, time-distance two-dimensional gridding and three-neighborhood interpolation, dimension raising mapping and state law-based form transformation, a multi-signal real-time processing result is obtained; high-precision alignment and fusion of multi-source signals are realized, synchronous processing of trackside, vehicle-mounted and central multi-signal streams can be completed at a millisecond level, and the signal detail recovery capability is remarkably improved while original additional parameters are reserved, so that time sequence drift and spatial deviation are effectively eliminated, the consistency of multi-dimensional signal data and an actual operation state is ensured, and the reliability of the system is improved. Based on the morphological transformation of the running state rule, the signal fusion result is highly matched with the modes such as actual acceleration and deceleration, stop and the like of the train, and the state recovery precision and the anomaly detection accuracy of the whole running process are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic signal processing, and in particular to a method and system for real-time processing of multiple signals in urban rail transit. Background Art

[0002] The processing system is designed for high-density operation scenarios such as subways and light rail. It needs to simultaneously process multi-source heterogeneous signals from trains, trackside equipment, and environmental monitoring, including key data such as speed, position, doors, power status, switches, track circuits, and safety monitoring, to ensure driving safety, efficient operations, and timely emergency response, while providing a high-precision data foundation for intelligent scheduling and predictive maintenance.

[0003] Existing urban rail transit signal processing technology still has the following drawbacks when dealing with multi-source heterogeneous signals collected by trackside equipment, onboard sensors, and dispatching centers during train operation: 1. Because traditional methods often use single signal stream processing or independent step-by-step processing, it is difficult to achieve strict synchronization of different signal sources on the time axis. There is obvious time drift and spatial deviation between signals, which affects the precise restoration of train operation status and the accuracy of safety monitoring. 2. There is a lack of an adaptive segmentation processing mechanism for the running status. When the train is in different states such as acceleration, deceleration, and stopping, the signal processing parameters cannot be dynamically adjusted, resulting in insufficient segmentation accuracy.

[0004] Based on this, the present invention proposes a real-time processing method and system for multiple signals in urban rail transit. By introducing signal segmentation processing driven by operating status sets, three-neighborhood interpolation based on time-distance two-dimensional gridding, dimensionality-increasing mapping and morphological transformation based on state change laws, it can achieve high-precision synchronization and fusion of multiple signal streams at the trackside, on-board and dispatching center within milliseconds, effectively improving the anomaly detection capability and response speed throughout the entire operation process. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and system for real-time processing of multiple signals in urban rail transit, so as to solve the problems in the background technology.

[0006] In order to achieve the above object, the present invention provides the following technical solution: a method for real-time processing of multiple signals in urban rail transit, the processing method comprising the following steps: S1: Continuously collect multiple types of signal data and corresponding operating status sets along the train running process; S2: Segmenting the multi-type signal data according to the operating state segments in the multiple operating state sets to obtain multiple signal units; S3: For each signal unit, calculate the first time window of adjacent sampling points in the time axis direction and the second time window in the running distance direction. Use the first time window and the second time window as the resolution to map the signal unit to a time-distance two-dimensional grid, obtain the two-dimensional coordinates of each original sampling point, and perform a three-neighborhood interpolation operation on each two-dimensional signal point to obtain an expanded two-dimensional signal coordinate set. S4: Based on the first time window and the second time window, the expanded two-dimensional signal coordinate set is mapped to the multidimensional signal space, and the numerical value of each additional parameter in the two-dimensional signal coordinate set is retained and mapped to the corresponding dimension; S5: Calculate the change rule of the running state set, and perform morphological transformation on multiple signal units in the multidimensional signal space according to the change rule; S6: Multiple morphologically transformed signal units are spliced ​​into a complete operating signal set in the order of state segments, realizing real-time restoration and alignment of multiple signals for the entire rail transit operation, and outputting the results to the train dispatching center in real time.

[0007] In a preferred embodiment, step S2: segmenting the multi-type signal data according to the multiple operating state segments to obtain multiple signal units includes the following steps: Calculating the total duration of the plurality of running state segments and the proportion of each running state segment in the total duration; Principal component analysis is performed on multi-type signal data to extract the main direction vector of signal changes; Constructing a first attitude matrix based on the main direction vector, and using the first attitude matrix to perform baseline alignment on the signal data, so that the change direction is aligned with the time axis, the secondary change direction is aligned with the running distance axis, and the remaining change components are aligned with the subsidiary parameter axis; Determine the enclosing interval of the aligned signal data, and determine the longest time dimension as the total length of the signal; Determine the time range of the signal unit corresponding to each state segment based on the ratio of the duration of each state segment and the total length of the signal; According to the time range corresponding to each state segment and the aligned signal time axis, the corresponding signal data is extracted and segmented to obtain multiple signal units.

[0008] In a preferred embodiment, the construction of the first posture matrix includes: aligning the direction vector corresponding to the maximum eigenvalue with the time axis, aligning the direction vector corresponding to the second largest eigenvalue with the running distance axis, aligning the remaining direction vectors with the subsidiary parameter axis, and generating a rotation matrix, i.e., the first posture matrix, by orthogonalizing the eigenvectors.

[0009] In a preferred embodiment, before the principal component analysis is performed on the multi-type signal data, it is constructed into an N×M data matrix according to a unified time reference, where N is the number of sampling points and M is the number of signal channels. The signal channels include speed, voltage, current, temperature, and vibration signal channels after time synchronization.

[0010] In a preferred embodiment, the signal unit includes a time sequence, a spatial position sequence, and speed, current, voltage, and temperature; Each original sampling point of the signal unit is mapped to a row and column index position of a two-dimensional grid according to its time value and distance value, and the corresponding speed, current, voltage, and temperature signal channel values ​​are bound to the row and column index positions.

[0011] In a preferred embodiment, performing a three-neighborhood interpolation operation on each two-dimensional signal point to obtain an expanded two-dimensional signal coordinate set includes the following steps: For the grid position not covered by the original sampling point, three known signal points in the two-dimensional grid that are closest to the grid position and located at different directions are selected; Calculate the time difference and distance difference between the grid position and each neighboring point, and combine them into a distance weight factor; Performing inverse distance weighted averaging on the signal channel values ​​of the three neighborhood points based on the weight factors to obtain interpolation results of each channel at the grid position; When there are less than three neighboring points, expand the search range until a neighboring point that meets the conditions is found.

[0012] In a preferred embodiment, step S4: based on the first time window and the second time window, up-dimensioning and mapping the expanded two-dimensional signal coordinate set to the multidimensional signal space, and retaining the numerical values ​​of each additional parameter in the two-dimensional signal coordinate set and mapping them to the corresponding dimensions, includes the following steps: Acquire a two-dimensional signal coordinate set that has been subjected to two-dimensional gridding and interpolation processing, wherein the two-dimensional signal coordinate set includes a time value, a distance value, and additional parameter values ​​of voltage, current, temperature, and vibration; Constructing a coordinate reference of a multidimensional signal space based on the first time window and the second time window, establishing a time index with the first time window as the minimum time resolution, establishing a distance index with the second time window as the minimum distance resolution, and assigning an independent parameter dimension index to each parameter according to the type of additional parameter value; Determine the corresponding time index and distance index based on the time value and distance value, assign each additional parameter value to the corresponding parameter dimension position, and establish a parameter vector containing each parameter value for each time-distance grid point; Normalize each parameter dimension in the mapped multidimensional signal space and convert the time, distance and parameter values ​​into floating-point numbers. The output is a multidimensional signal space dataset containing time dimension, distance dimension, and multiple parameter dimensions.

[0013] In a preferred embodiment, step S5: calculating the change rule of the operating state set and performing morphological transformation on multiple signal units in the multidimensional signal space according to the change rule, includes the following steps: Match the operating state segment and the corresponding state law vector for each signal unit; Calculate the deviation distribution between the actual change curve of the signal unit and the theoretical curve corresponding to the state law vector; Generate transformation rules based on the deviation distribution and perform parameter adjustment, introducing physical constraints such as non-negative speed and current not exceeding safety thresholds during the adjustment process; After the transformation is completed, the goodness of fit between the transformed signal curve and the theoretical curve is calculated, and the correlation changes between the parameter dimensions are verified.

[0014] In a preferred embodiment, step S6: splicing the multiple transformed signal units into a complete operating signal set in the order of the state segments to achieve real-time restoration and alignment of multiple signals for the entire operation of rail transit, includes the following steps: Assign a unique sequence index to each signal unit according to the time sequence of each state segment in the running state set, and align the start and end times of each signal unit with the global time reference, and align the spatial coordinate interval with the global distance axis; The signal units are spliced ​​in the multidimensional signal space in the order of sequence index, including: Matching sampling points at the time and space boundaries of adjacent signal units; The time-distance-parameter matrices of each signal unit are spliced ​​into a continuous multi-dimensional matrix structure according to the global time and distance order; Perform weighted sliding average or piecewise polynomial interpolation on speed and current parameters within the state segment boundary interval; The fused and aligned full-route operation signal set is encapsulated into segmented data packets according to timestamp, distance coordinate, parameter vector and state segment identifier, and sent to the train automatic control system and dispatching center in real time through the rail transit communication protocol.

[0015] The present application also provides a real-time processing system for multiple signals in urban rail transit, including a data acquisition and segmentation module, a data mapping module, a morphological transformation module, and an output module; Data acquisition and segmentation module: continuously collects multiple types of signal data and corresponding operating status sets along the train running process, and segments the multiple types of signal data according to the operating status segments in multiple operating status sets to obtain multiple signal units; Data mapping module: For each signal unit, the first time window of adjacent sampling points in the time axis direction and the second time window in the running distance direction are calculated. The signal unit is mapped to a time-distance two-dimensional grid with the first and second time windows as the resolution to obtain the two-dimensional coordinates of each original sampling point. A three-neighborhood interpolation operation is performed on each two-dimensional signal point to obtain the expanded two-dimensional signal coordinate set; Morphological transformation module: Based on the first time window and the second time window, the expanded two-dimensional signal coordinate set is mapped to the multidimensional signal space, and the numerical values ​​of each additional parameter in the two-dimensional signal coordinate set are retained and mapped to the corresponding dimension. The change pattern of the operating state set is calculated, and the morphological transformation of multiple high-precision signal units in the multidimensional signal space is performed according to the change pattern. Output module: Splices multiple morphologically transformed signal units into a complete operating signal set in the order of state segments, realizes real-time restoration and alignment of multiple signals for the entire operation of rail transit, and outputs the results to the train dispatching center in real time.

[0016] In the above technical solution, the technical effects and advantages provided by the present invention are: The present invention achieves high-precision alignment and fusion of multi-source signals through signal segmentation processing driven by a set of operating states, two-dimensional time-distance gridding and three-neighborhood interpolation, dimensionality-increasing mapping, and morphological transformation based on state laws. It can complete the synchronous processing of multiple signal streams at the trackside, onboard, and in the center at the millisecond level, and significantly improve the ability to recover signal details while retaining the original additional parameters, thereby effectively eliminating timing drift and spatial deviation, ensuring the consistency of multi-dimensional signal data with the actual operating state. Morphological transformation based on operating state laws ensures that the signal fusion results are highly consistent with the train's actual acceleration, deceleration, and stop patterns, significantly improving the state restoration accuracy and anomaly detection accuracy throughout the entire operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments described in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0018] Figure 1 Flowchart of the processing method of the present invention.

[0019] Figure 2 It is a timing diagram of the processing method of the present invention. DETAILED DESCRIPTION

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0021] Example 1: Please refer to Figure 1-Figure 2 As shown, this embodiment provides a method for real-time processing of multiple signals in urban rail transit, and the processing method includes the following steps: S1: Continuously collect multi-type signal data and corresponding operation status sets along the train operation process. The multi-type signal data includes multiple uncalibrated multi-type signal data streams, and the operation status set includes multiple operation status segments (such as interval operation, entering the station, stopping at the station, leaving the station, etc.).

[0022] S2: Segment the multi-type signal data according to the operating state segments in the multiple operating state sets to obtain multiple signal units, including: Calculating the total duration of the plurality of running state segments and the proportion of each running state segment in the total duration; Principal component analysis is performed on multi-type signal data to extract the main direction vector of signal changes; Constructing a first attitude matrix based on the main direction vector, and using the first attitude matrix to perform benchmark alignment on the signal data, so that the main change direction is aligned with the time axis, the secondary change direction is aligned with the running distance axis, and the remaining change components are aligned with the subsidiary parameter axis; Determine the enclosing interval of the aligned signal data, and determine the longest time dimension as the total length of the signal; Determine the time range of the signal unit corresponding to each state segment based on the ratio of the duration of each state segment and the total length of the signal; According to the time range corresponding to each state segment and the aligned signal time axis, the corresponding signal data is extracted and segmented to obtain multiple signal units.

[0023] S3: For each signal unit, calculate a first time window in the time axis direction and a second time window in the running distance direction of adjacent sampling points; map the signal unit to a time-distance two-dimensional grid using the first and second time windows as resolutions to obtain the two-dimensional coordinates of each original sampling point; perform a three-neighborhood interpolation operation on each two-dimensional signal point to obtain an expanded high-density two-dimensional signal coordinate set; S4: Based on the first time window and the second time window, the expanded two-dimensional signal coordinate set is mapped to a multidimensional signal space (time-distance-parameter), and the values ​​of the additional parameters (such as voltage, current, temperature, vibration, etc.) in the two-dimensional signal coordinate set are retained and mapped to the corresponding dimensions; S5: Calculate the change pattern of the operating state set (such as acceleration and deceleration curves, stop patterns, energy consumption change patterns, etc.), and perform morphological transformation on multiple high-precision signal units in the multidimensional signal space based on the change pattern, so that the change pattern of the transformed signal units is consistent with the pattern of the operating state set; S6: Multiple transformed signal units are spliced ​​into a complete operating signal set in the order of state segments, realizing real-time restoration and alignment of multiple signals throughout the entire rail transit operation. The results are then output to the train automatic control system and dispatching center in real time, providing a high-precision data foundation for safety monitoring, abnormal warning, and operation optimization.

[0024] This application achieves high-precision alignment and fusion of multi-source signals through signal segmentation processing driven by the operating state set, time-distance two-dimensional gridding and three-neighborhood interpolation, dimensionality-increasing mapping, and morphological transformation based on state laws. It can complete the synchronous processing of multiple signal streams at the trackside, onboard, and center within milliseconds, and significantly improve the ability to recover signal details while retaining the original additional parameters, thereby effectively eliminating timing drift and spatial deviation, ensuring the consistency of multi-dimensional signal data with the actual operating state. The morphological transformation based on the operating state law makes the signal fusion results highly consistent with the actual acceleration, deceleration, and stop modes of the train, significantly improving the state restoration accuracy and anomaly detection accuracy throughout the entire operation.

[0025] Example 2: This example provides a real-time processing system for multiple signals in urban rail transit, including a data acquisition and segmentation module, a data mapping module, a morphology transformation module, and an output module; Data acquisition and segmentation module: continuously collects multiple types of signal data and corresponding operating status sets along the train running process, segments the multiple types of signal data according to the operating status segments in multiple operating status sets, obtains multiple signal units, and sends the signal units to the data mapping module; Data mapping module: For each signal unit, the first time window of adjacent sampling points in the time axis direction and the second time window in the running distance direction are calculated. The signal unit is mapped to a time-distance two-dimensional grid with the first and second time windows as the resolution to obtain the two-dimensional coordinates of each original sampling point. A three-neighborhood interpolation operation is performed on each two-dimensional signal point to obtain an expanded two-dimensional signal coordinate set. The first time window, the second time window, and the two-dimensional signal coordinate set are sent to the morphological transformation module. Morphological transformation module: Based on the first time window and the second time window, the expanded two-dimensional signal coordinate set is mapped to the multidimensional signal space, and the numerical values ​​of each additional parameter in the two-dimensional signal coordinate set are retained and mapped to the corresponding dimensions. The change pattern of the operating state set is calculated, and the morphological transformation of multiple high-precision signal units in the multidimensional signal space is performed according to the change pattern. The signal units after morphological transformation are sent to the output module; Output module: Splices multiple morphologically transformed signal units into a complete operating signal set in the order of state segments, realizes real-time restoration and alignment of multiple signals for the entire operation of rail transit, and outputs the results to the train dispatching center in real time.

[0026] Example 3: S1: Continuously collect multi-type signal data and corresponding operation status sets along the train operation process. The multi-type signal data includes multiple uncalibrated multi-type signal data streams, and the operation status set includes multiple operation status segments (such as interval operation, entering the station, stopping at the station, leaving the station, etc.).

[0027] Throughout the train's operation, a comprehensive multi-signal acquisition network is constructed through trackside sensors deployed along the track, onboard acquisition devices installed on the train, and communication interfaces connected to the dispatching center. This enables the continuous acquisition of multiple types of signal data and operational status sets. These multiple types of signal data include, but are not limited to, uncalibrated raw velocity signals, acceleration signals, position coordinate signals, train voltage signals, current signals, braking status signals, door opening and closing status signals, trackside equipment status signals (such as switch position and track circuit conductivity), and environmental signals (temperature, humidity, smoke concentration, vibration intensity, etc.). These signals may originate from equipment from different manufacturers, have inconsistent sampling frequencies, and differ in accuracy levels. Furthermore, they lack unified timestamp calibration when the data is generated, resulting in timing drift and spatial reference differences in the initial state.

[0028] The running state set consists of multiple running state segments, each of which corresponds to a specific running mode of the train during operation, such as interval operation, station entry, station stop status, and station exit acceleration. There are two ways to obtain the running state set: Forward collection: Based on the operation plan issued by the dispatching center, the section entry and exit times recorded by the signal system, and the operation event identifiers uploaded in real time by the automatic train monitoring system (ATS), the expected operation status segment is generated in advance.

[0029] Reverse judgment: By real-time analysis of the train speed curve, braking signal, position coordinate change rate, etc., the train's current operating status segment is dynamically identified, and its time range and identification information are recorded in the operating status set.

[0030] In order to ensure the availability of collected data and the accuracy of subsequent processing, the signal is preliminarily synchronized and checked for data integrity during the acquisition process. The synchronization logic is as follows: Sampling time alignment: For signal streams with different sampling frequencies, first extract the sampling time series of each signal stream, calculate the statistical characteristics (maximum, minimum, and mean) of adjacent sampling time intervals, and use the sampling time interval with the minimum common multiple as a unified benchmark to interpolate or downsample all signal streams to this benchmark sampling frequency.

[0031] Packet timing sorting: The collected multi-signal packets are sorted according to a unified timestamp. If a signal stream is found to be missing or delayed beyond a threshold (for example, twice the unified sampling interval), the missing value compensation logic is triggered to fill the missing value through short-term prediction or delay caching.

[0032] Spatial position alignment: For position-dependent signals (such as wayside equipment trigger signals, track section status signals, etc.), their physical locations are mapped to the train mileage coordinate system to ensure that signals from different sources at the same location can be directly associated.

[0033] Ultimately, step S1 forms a signal dataset with unified three-dimensional indexing of time, space, and state, as well as a corresponding set of operating states. This not only provides a unified reference for subsequent segmentation processing, two-dimensional gridding, and interpolation operations, but also avoids subsequent processing errors caused by asynchronous and inconsistent multi-type signal data. For example, when a train enters the station entry phase, the speed signal, voltage signal, brake signal, and door opening and closing signal are correlated at the same time and position reference, allowing for precise analysis of the corresponding relationship between equipment response and train dynamics during the entry process.

[0034] S2: Segment the multi-type signal data according to the operating state segments in the multiple operating state sets to obtain multiple signal units, including: Calculating the total duration of the plurality of running state segments and the proportion of each running state segment in the total duration; Principal component analysis is performed on multi-type signal data to extract the main direction vector of signal changes; Constructing a first attitude matrix based on the main direction vector, and using the first attitude matrix to perform benchmark alignment on the signal data, so that the main change direction is aligned with the time axis, the secondary change direction is aligned with the running distance axis, and the remaining change components are aligned with the subsidiary parameter axis; Determine the enclosing interval of the aligned signal data, and determine the longest time dimension as the total length of the signal; Determine the time range of the signal unit corresponding to each state segment based on the ratio of the duration of each state segment and the total length of the signal; According to the time range corresponding to each state segment and the aligned signal time axis, the corresponding signal data is extracted and segmented to obtain multiple signal units.

[0035] After obtaining the multi-type signal data set and operating status set formed by S1, the goal of this step is to accurately divide the signal flow of the entire operating process according to the operating status segments, forming multiple independently processable signal units, laying a structured data foundation for subsequent two-dimensional gridding and dimensionality-increasing mapping.

[0036] First, based on the start and end time of each state segment in the running state set, the total duration of all state segments is counted. The processing logic is: Calculate the total duration: traverse the running state set, extract the start time and end time of each state segment, calculate the duration of each segment and accumulate them to get the total duration.

[0037] Calculate the proportional coefficient: divide the duration of each state segment by the total duration to obtain the proportional coefficient of the state segment in the entire process. The proportional coefficient will be used in the subsequent steps to map the time range of the signal unit.

[0038] Secondly, principal component analysis (PCA) is performed on the original multi-type signal data to extract the main direction vector of the signal change. The specific processing logic is: Data matrix construction: Organize multiple types of signals into an N×M dimensional matrix based on a unified time base, where N is the number of sampling points and M is the number of signal channels.

[0039] Covariance calculation: Calculate the covariance matrix of the matrix between channels to reflect the correlation of signal changes in different dimensions.

[0040] Eigendecomposition: Perform eigenvalue decomposition on the covariance matrix, extract the eigenvector corresponding to the largest eigenvalue as the main direction vector, and arrange the rest in order of eigenvalue size.

[0041] Based on the above main direction vector, the first posture matrix is ​​constructed. The generation logic of this matrix is: Let the dimension corresponding to the main direction vector be the first reference axis and align it with the global time axis; The direction vector corresponding to the second largest eigenvalue is selected as the second reference axis and aligned with the running distance axis; The remaining direction vector serves as the third reference axis, aligned with the subsidiary parameter axis (such as voltage, current, temperature, vibration, etc.); These three sets of orthogonal direction vectors are combined into a rotation matrix, and matrix multiplication transformation is performed on the multi-type signal data matrix to achieve baseline alignment of the signals under a unified posture.

[0042] After the attitude matrix transformation is completed, the aligned signal data will present a standardized layout of three dimensions: time, distance, and auxiliary parameters. At this time, the bounding box of the aligned data is calculated, and the processing logic is as follows: Get the minimum and maximum values ​​for each dimension; Determines the extent of the time dimension to the total length of the signal, which will be used in the scaling map.

[0043] Then, the time range mapping of each state segment in the aligned signal is calculated based on the ratio value of each state segment and the total length of the signal: Multiply the state segment ratio by the total signal length to obtain the length of the segment on the signal time axis; The lengths of each segment are accumulated to obtain the start and end time range of each segment.

[0044] Finally, according to the calculated time range, the signal data corresponding to the interval in the aligned signal time axis is extracted to form multiple independent signal units {U1, U2, …, Un}. Each signal unit contains all signal channels and associated parameters within the time range and retains the mapping relationship with the operating status segment.

[0045] Resample and align multiple types of signal data under a unified time base to build The data matrix: in, Number of sampling points, Number of signal channels. Channels include: time-synchronized speed, voltage, current, temperature, vibration, etc. (can be expanded based on implementation scenarios). :No. The sampling point is The values ​​of the channels (after unified timestamp alignment, missing compensation / interpolation).

[0046] Suppose the state set of a running process contains state segment (such as interval operation, entry, stop, exit), part The start and end times are , single segment duration: , total duration: , proportionality coefficient: , parameter definition: :No. Status segment duration. Total duration of the entire process, No. The proportion of time in the whole process, .

[0047] To extract the main direction of signal change, Do mean centering: ,in, No. The sample mean of the channel. 1: Length is A vector of all 1s. Covariance matrix: right Perform eigendecomposition: , The eigenvector matrix is ​​sorted by eigenvalue from largest to smallest. : corresponding eigenvalue diagonal matrix, Main direction vector: First main direction: (maximum eigenvalue correspond).

[0048] Second main direction: .

[0049] Other directions: .

[0050] Align the direction vector corresponding to the largest eigenvalue with the time axis, the direction vector corresponding to the second largest eigenvalue with the running distance axis, and the remaining direction vectors with the subsidiary parameter axes. To ensure orthogonality, right-handed system and numerical stability, Orthogonalize and assemble the rotation matrix. Eigenvector orthogonalization (illustration): For example, the Gram-Schmidt idea is used to obtain the standard orthogonal basis : ,the remaining The same construction is carried out to form an orthogonal basis matrix: ,Will Specify the timeline. is consistent with the distance axis, and the rest are consistent with the subsidiary parameter axes, then This is the first posture matrix.

[0051] Datum alignment (coordinate transformation): . For the aligned data, the first column is along the “main direction of time variation”, the second column is along the “main direction of distance”, and the remaining columns correspond to the “subsidiary parameter directions”.

[0052] right Find the minimum / maximum values ​​for each column of , and the value range of the time corresponding column is defined as the total length of the signal: ,in, The total length in the time dimension (coverage of the internal time coordinates after alignment). Map each state segment to the aligned timeline interval ,in, , define the cumulative proportion of the state segment: , then The start and end points of the segments on the timeline after alignment are: in, , No. The start and end coordinates of the segment on the aligned timeline. To The time proportion accumulated up to the end of the period. The first column (time coordinate after alignment) is the index, and the All rows of , together with the remaining columns (distance and auxiliary parameters) constitute the signal unit .repeat ,get: , each Each carries the corresponding status segment identifier and the full channel data within the time range.

[0053] Through the above processing, this step not only achieves precise segmentation of signals based on operating status, but also eliminates differences in spatial reference and scale between different signals through PCA and attitude matrix transformation, ensuring that subsequent interpolation and dimensionality increase operations can be performed in a unified data coordinate system. For example, when a train is in the "entering the station" state, the corresponding signal unit will accurately contain information such as speed changes, braking status, current fluctuations, and door movements during this stage, facilitating independent analysis of the operating characteristics of this stage.

[0054] S3: For each signal unit, calculate the first time window of adjacent sampling points in the time axis direction and the second time window in the running distance direction; with the first time window and the second time window as the resolution, map the signal unit to the time-distance two-dimensional grid to obtain the two-dimensional coordinates of each original sampling point; perform a three-neighborhood interpolation operation on each two-dimensional signal point to obtain an expanded high-density two-dimensional signal coordinate set.

[0055] After obtaining multiple independent signal units from step S2, this step aims to map each signal unit onto a unified time-distance two-dimensional grid space for subsequent dimensionality increase and multi-signal fusion processing. Because different signal units may have inconsistencies in sampling intervals and spatial distribution, it is necessary to establish a unified spatial sampling benchmark by calculating time windows and distance windows. Interpolation algorithms are also used to supplement data missing at grid points to improve data density and continuity.

[0056] For each signal unit, its first time window (Δt1) in the time axis direction and its second time window (Δd1) in the running distance direction are calculated respectively. The processing logic is as follows: Time axis interval calculation: Extract the time difference between adjacent sampling points from the time series of the signal unit, count the distribution characteristics of all time differences (maximum value, minimum value, mean value, mode, etc.), and select the representative value (usually mean value or mode value) that best reflects the sampling stability in the statistical results as the first time window Δt1.

[0057] Calculation of running distance direction interval: Extract the distance difference between adjacent sampling points from the spatial position sequence of the signal unit (usually obtained by odometer, trackside positioning points or GPS), and use the same statistical method as the time window calculation to obtain the second time window Δd1.

[0058] After obtaining Δt1 and Δd1, these two window values ​​are used as the resolution of the two-dimensional grid to construct a time-distance two-dimensional grid. The horizontal axis of the grid represents the train running time, and the vertical axis represents the running distance. Each grid point in the grid represents a fixed time-distance coordinate. The processing logic is: Traverse each original sampling point in the signal unit and calculate the row and column index of the point in the two-dimensional grid based on its time value and distance value; Bind each signal channel value (including speed, voltage, current, etc.) of the original sampling point to the corresponding grid coordinate position; Grid coordinates that are not covered by the original sampling points are marked as "null values" for subsequent interpolation filling.

[0059] Then, a three-neighborhood interpolation operation is performed on each valid signal point in the two-dimensional grid to obtain an expanded high-density two-dimensional signal coordinate set. The logic of the three-neighborhood interpolation is: For the target null point, find the three closest known signal points in the two-dimensional grid. These three points need to be distributed in different directions of the target point (for example, upper left, lower right, directly above, etc.) to ensure the spatial coverage of the interpolation. Calculate the time difference and distance difference between the target point and the three known points, and combine these two differences into a distance weight factor (the smaller the weight, the closer the position); Perform weighted averaging on the signal values ​​of the three neighboring points according to their weights, and use the result as the interpolation value of the target point; Perform the above weighted interpolation on multi-channel signals (such as speed, current, and voltage) separately to maintain the independence of each channel; If a null point cannot find three valid neighboring points, the search range is expanded outward until enough neighboring points are found for interpolation calculation.

[0060] In a two-dimensional grid, let the sampling intervals be the time steps and distance step length , the original sampling point set is: , No. The time coordinates of a known point, No. The distance coordinates of the known points, :The point Channel values ​​(such as speed, voltage, current, temperature, vibration, etc.). Discretize the entire operating range into a regular grid : , find out what has not been Covered location : .

[0061] For each position to be interpolated : Divide the neighborhood by orientation: Divide the plane into three 120° sectors (or equal angle areas) to As the center, according to the vector , The polar angles are grouped.

[0062] Select the nearest point: Select the nearest point in each sector. The Euclidean distance to the nearest known point has a maximum of three neighboring points in three directions.

[0063] Handling less than three points: If a sector has no points, the search range is expanded from smaller to larger radius until a point is found in that sector or the entire domain has been searched. If there are still less than three points, the search can be reduced to two-neighborhood linear interpolation or single-neighborhood nearest neighbor assignment.

[0064] Calculation of distance weight factor: For the selected neighborhood points , calculate the time difference and distance difference: , define the normalized distance: ,in, 、 is the normalization factor in the time and distance directions (it can be a global sampling interval or data statistic). Inverse distance weight: ,in is the power exponent (usually 1-3).

[0065] For each channel ,exist The interpolated value is: , The actual number of neighborhood points selected (ideally 3). For the The first neighboring point Channel value. As the multi-channel signal value of the grid position. Traverse all ,After the interpolation assignment is completed, the original known points and the interpolation points together form the expanded ,two-dimensional signal coordinate set: .

[0066] After interpolation, the original sparse signal unit data is converted into a two-dimensional coordinate set with uniform coverage density and higher data continuity. This high-density two-dimensional signal coordinate set not only ensures sampling consistency on the time axis and distance axis, but also effectively fills data gaps while preserving the trends of multiple types of signal data, reducing the accumulation of errors during subsequent dimensional mapping. For example, during the deceleration phase when the train enters the platform, the original data may be missing some distance points due to communication delays. Through the three-neighborhood interpolation in this step, the speed and braking status of these missing points can be accurately restored, thereby ensuring the continuity and accuracy of the operation status analysis.

[0067] S4: Based on the first time window and the second time window, the expanded two-dimensional signal coordinate set is mapped to the multidimensional signal space (time-distance-parameter), and the numerical values ​​of each additional parameter (such as voltage, current, temperature, vibration, etc.) in the two-dimensional signal coordinate set are retained and mapped to the corresponding dimension.

[0068] After obtaining the expanded, high-density 2D signal coordinate set in step S3, the goal of this step is to map the 2D signal data into a multidimensional signal space, enabling time, distance, and various additional parameters to be uniformly represented in the same coordinate system. This provides a complete data foundation for subsequent morphological transformations and full-process signal fusion. This up-dimensional mapping not only preserves the time-distance structure of the 2D signal but also incorporates the additional parameters into the multidimensional space as independent dimensions, ensuring the full integration of multi-source information.

[0069] First, based on the first time window Δt1 and the second time window Δd1, a coordinate reference of the multi-dimensional signal space is constructed. The processing logic is: Time dimension construction: Taking Δt1 as the minimum time resolution, the time axis of the two-dimensional signal set is divided into several continuous intervals, each of which corresponds to a time coordinate index in the multidimensional space; Distance dimension construction: Taking Δd1 as the minimum distance resolution, the distance axis of the two-dimensional signal set is divided into several continuous intervals, each of which corresponds to a distance coordinate index in the multidimensional space; Parameter dimension planning: Based on the types of additional parameters contained in the two-dimensional signal set (such as voltage, current, temperature, vibration, etc.), each parameter is assigned an independent dimension index to ensure that the parameter value can be accessed and processed separately in the multidimensional space.

[0070] Next, each data point of the two-dimensional signal coordinate set is mapped to the multi-dimensional signal space. The logic of the mapping process is: Index positioning: According to the time value and distance value of the data point, the corresponding time coordinate index and distance coordinate index are searched respectively to determine its two-dimensional position in the multidimensional space; Parameter mapping: For each additional parameter of the data point, directly assign its value to the corresponding parameter dimension coordinate position; Multi-channel synchronization: A parameter vector is established for each time-distance grid point in multidimensional space. Each component of the parameter vector corresponds to a value such as voltage, current, temperature, and vibration. Missing parameter filling: If a grid point has no original value in a specific parameter dimension, the previous valid value retention strategy or local neighborhood mean strategy is called to fill it in to avoid the influence of null values ​​in subsequent calculations.

[0071] In order to ensure the consistency and comparability of the data after dimensionality increase, normalization and data type standardization are required during the mapping process: Normalization: Scaling or standard deviation normalization is performed on the values ​​of each parameter dimension to eliminate the impact of differences in dimensions and value ranges between different parameters. For example, voltage is normalized to the range of 0-1 and temperature is centered around the mean value. Data type standardization: Ensure that time, distance, and parameter values ​​are unified as high-precision floating-point types to avoid calculation errors caused by integer truncation or insufficient precision.

[0072] Ultimately, the two-dimensional signal set is mapped into a multidimensional signal space data structure of three dimensions or higher. The most basic three-dimensional coordinate system is time-distance-parameter, where the parameter dimension can be expanded into multiple sub-dimensions (for example, speed, voltage, current, temperature, and vibration each occupy a different dimension index). In this representation, each multidimensional grid point contains all available physical parameters at that moment and location, forming a complete slice of the operating status.

[0073] (1) Time dimension index Assume the first time window is (Unit: seconds), Create a time index for the minimum time resolution: in, The time value of the two-dimensional signal point, The start time of the full run; Time index (negative integers up to 15 years).

[0074] (2) Distance dimension index Let the second distance window be (Unit: meter), Create a distance index for the minimum distance resolution: , Distance value of two-dimensional signal point The starting distance of the full run; Distance index (non-negative integer).

[0075] Parameter dimension planning Let the additional parameter set be: Voltage , current ,temperature ,vibration Assign a separate dimension index to each parameter ,in and is the total number of parameters.

[0076] For each data point in the two-dimensional signal set Calculated according to the formula and Establish coordinate positions in multidimensional signal space: , each additional parameter value Stored to the corresponding parameter dimension position.

[0077] Data structure of multidimensional signal space: can be viewed as a tensor of three dimensions or more , where the first dimension is the time index , the second dimension is the distance index , the third and subsequent dimensions are the dimensions of each parameter. In order to ensure the comparability of the values ​​of different parameters, for each parameter dimension Perform normalization: ,in, No. The mean of the parameters; No. The standard deviation of the parameters; is the normalized parameter value. After normalization, the time, distance, and parameter values ​​are uniformly converted to high-precision floating-point numbers (such as IEEE754 double precision) to avoid precision loss caused by integer truncation.

[0078] The advantage of this dimensional mapping is that it can store and process multiple types of signals in a unified structure while ensuring temporal and spatial synchronization. For example, when a train reaches a distance of 2.5 km and a travel time of 180 seconds, the grid point may contain data such as speed 65 km / h, voltage 750 V, current 320 A, temperature 28°C, and vibration 0.3 g. This data has a fixed index position in the multidimensional space, making it easy for subsequent morphological transformation or feature extraction algorithms to directly access and calculate it.

[0079] S5: Calculate the changing rules of the operating state set (such as acceleration and deceleration curves, stop patterns, energy consumption change patterns, etc.), and perform morphological transformation on multiple high-precision signal units in the multidimensional signal space based on the changing rules, so that the changing rules of the transformed signal units are consistent with the rules of the operating state set.

[0080] After completing the multidimensional signal space construction in S4, this step aims to perform morphological transformations on each high-precision signal unit in the multidimensional signal space by analyzing the changing patterns of the operating state set (such as acceleration and deceleration curves, stop patterns, and energy consumption variation patterns). This process not only eliminates local noise and abnormal fluctuations during the acquisition process but also improves the fit between signal variation patterns and the actual train operation patterns across the entire journey, thereby enhancing the accuracy of subsequent fusion and anomaly detection.

[0081] First, the change rules are calculated based on the running state set. The processing logic is: Data extraction: Select time, distance and key operating parameters (such as speed, current, braking status, voltage, etc.) from the multi-dimensional signal space and group them according to operating status segments; Curve fitting: For each operating state (such as acceleration, deceleration, constant speed, and stop), the key parameter change curve of the corresponding section is fitted. The fitting method can use piecewise polynomial fitting, piecewise spline interpolation, or trend curve fitting based on weighted least squares method to ensure that the fitting curve can smoothly reflect the changing trend within the state segment; Pattern extraction: Parameterize the fitting curves of each operating state. For example, the acceleration segment is characterized by average acceleration, maximum acceleration, and acceleration change rate; the stop segment is characterized by stop time, power consumption level, and zero speed duration; and the energy consumption pattern is characterized by the slope of the power consumption curve and the peak power position. Regularity modeling: The above parameters are used as the state regularity vector R={R1, R2,…, Rn}, where each Ri corresponds to a typical characteristic indicator of a certain operating state, which is used to guide subsequent morphological transformation.

[0082] Then, the calculated operating state rules are applied to the morphological transformation of the multi-dimensional signal unit. The processing logic is: State mapping: Match each multidimensional signal unit to its operating state segment and obtain the regular vector Ri corresponding to the state segment; Deviation calculation: Compare the actual change curve of the signal unit in the time-distance-parameter space with the theoretical curve corresponding to the regular vector, and calculate the deviation distribution between the two (including mean square error, maximum deviation, change trend difference, etc.); Transformation rule generation: Generate morphological transformation rules based on the deviation distribution. For example, when the acceleration of the velocity curve is lower than the theoretical value, the acceleration component is amplified proportionally; when the current fluctuation amplitude is too high, the current dimension is smoothed; Transformation execution: The multidimensional parameter vector of the signal unit is adjusted point by point according to the transformation rules to ensure that the overall trend of the adjusted signal is consistent with the regular vector, while trying to keep the detailed features of the original data from being over-smoothed; Constraint check: Physical constraints are introduced during the transformation process, such as speed cannot be negative, current cannot exceed the safety threshold, and temperature change rate cannot exceed the reasonable range, to prevent distortion of the adjustment results or the introduction of unreasonable values.

[0083] To ensure the quality of the signal units after morphological transformation, consistency verification is performed after the transformation is completed: Trend consistency verification: Calculate the goodness of fit between the transformed signal curve and the theoretical regularity curve (e.g. R 2 value or dynamic time warping (DTW) matching score), confirming that it reaches a preset threshold; Multi-dimensional coordination verification: Check whether the correlation changes between different parameter dimensions conform to the normal pattern under the operating state. For example, the increase in speed in the acceleration section should be accompanied by an increase in current, and the speed in the stop section should be zero and the power consumption should decrease.

[0084] Assume that the running status set is: ,in Indicates the There are three operating state segments (acceleration, constant speed, deceleration, and stop). Select key parameters (speed , current ,Voltage etc.), and fitting its time series to obtain the theoretical curve function: ,in, Parameter dimension number, Time variables; Status segment Medium parameters Theoretical value of Fitting function (can be a polynomial, spline curve or weighted least squares fitting result). Characterize the fitting curve and obtain the state law vector: ,in, Status segment average speed; Average acceleration; Peak current; other components may include power slope, humidity change rate, etc.

[0085] For the state segment in the multidimensional signal space The signal unit is used to calculate the deviation between the actual curve and the theoretical curve: ,in, Signal unit in parameter dimension The actual value of Deviation value. The deviation distribution characteristics can be further statistically analyzed: ,in, Mean square error; :Maximum deviation; Trend slope deviation.

[0086] Generate adjustment factors based on the deviation distribution: ,in, Adjust the sensitivity coefficient (0-1); parameter When the deviation is mainly manifested as insufficient amplitude, proportional amplification can be used: When the deviation is mainly manifested in excessive fluctuation, smoothing can be used: , Smoothing coefficient (0-1); .

[0087] Ultimately, the morphologically transformed signal units output by this step exhibit a highly consistent variation pattern across the multidimensional space of time, distance, and parameters, consistent with the set of operating states. For example, during a deceleration section approaching a station, the speed curve, braking status, and current changes will all conform to typical station entry patterns, eliminating abnormal fluctuations in multi-type signal data caused by sensor delays or noise.

[0088] S6: Multiple transformed signal units are spliced ​​into a complete operating signal set in the order of state segments, realizing real-time restoration and alignment of multiple signals throughout the entire rail transit operation. The results are then output to the train automatic control system and dispatching center in real time, providing a high-precision data foundation for safety monitoring, abnormal warning, and operation optimization.

[0089] After completing the morphological transformation of each signal unit in step S5, the goal of this step is to splice these corrected high-precision signal units in the order of their corresponding operating status segments, construct a complete multi-signal dataset covering the entire train operation, and achieve global alignment in time, space, and parameter dimensions. Ultimately, the results are transmitted in real time to the train automatic control system (ATO) and dispatching center for safety monitoring, abnormality warning, and operation optimization.

[0090] First, establish the sequence mapping relationship between signal units and running status segments. The processing logic is: Sequence index generation: assign a unique sequence index ID to each state segment according to the time sequence of each state segment in the running state set, and record the index in the signal unit metadata; Time base unification: Extract the start and end times of all signal units, determine the unified start time T0 and end time Tn for the entire process, and perform time offset adjustment on each signal unit so that its start and end times are strictly aligned with the global timeline of the running status segment; Unified spatial reference: Using the train's cumulative mileage or interval position as the spatial reference, the spatial coordinate interval of each signal unit is mapped to a unified distance axis for the entire journey, ensuring that the positions of different state segments after splicing are continuous without overlap or gaps.

[0091] Secondly, the signal units are spliced ​​and fused. The processing logic is as follows: Boundary matching: Matches the sampling points of adjacent signal units on the time and space boundaries. If there are overlapping intervals, weighted averaging or priority signal source strategy is used for fusion. The weights can be dynamically adjusted according to sampling reliability, data freshness or signal quality indicators. Parameter vector splicing: In the multidimensional signal space, the time-distance-parameter matrices of each unit are arranged in sequence according to the global time and distance order to form a continuous multidimensional matrix structure; Boundary smoothing: To eliminate sudden changes in parameters such as speed and current between adjacent state segments, a smooth transition algorithm (such as weighted sliding average or piecewise polynomial interpolation) is used within the boundary interval to ensure that the curve is continuous and physically reasonable.

[0092] Next, perform global alignment and consistency check on the entire run signal set: Global time alignment: Confirm that the sampling interval of all parameters on the entire time axis is consistent with Δ Maintain consistency. If a local interval anomaly is found, perform interpolation or supplementary sampling correction; Global space alignment: Confirm that the sampling interval of all parameters on the global distance axis is consistent with Δ Maintain consistency and perform distance interpolation or coordinate mapping corrections when necessary; Multi-dimensional correlation verification: Check whether the physical correlation of parameters such as speed, current, voltage, and braking status in each state segment is reasonable. For example, the current should increase in the acceleration segment, and the speed should be zero and accompanied by low power consumption in the stop segment.

[0093] Finally, the fused and aligned full-route operation signal set is output to the train automatic control system and dispatching center in real time. The processing logic is as follows: Data packaging: encapsulate the entire operation signal set into segmented data packets, each of which contains a timestamp, distance coordinates, parameter vector, and state segment identifier; Real-time transmission: Data packets are sent in sequence via dedicated rail transit communication protocols (such as LTE-M, GSM-R, or Industrial Ethernet), ensuring they reach control and monitoring terminals within millisecond latency. Receiving-end applications: The ATO system can use real-time data for automatic speed control and section safety monitoring. The dispatching center can visualize the operating status, issue warnings for abnormal conditions, and conduct energy consumption optimization analysis based on the full signal set.

[0094] For example, after a train completes a run from the starting station to the terminal station, the full-route operation signal set generated in this step can fully present the changes in speed, voltage, current, temperature, vibration and other parameters in all state segments such as acceleration, cruising, entering the station, stopping, and leaving the station, and is strictly continuous in time and distance, without omissions or jumps, ensuring that the monitoring system and scheduling decisions are based on high-precision and reliable data.

[0095] Throughout this specification, references to terms such as "one embodiment," "example," or "specific example" indicate that the specific features, structures, materials, or characteristics described in conjunction with that embodiment or example are included in at least one embodiment or example of the present invention. In this specification, schematic representations of these terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0096] The preferred embodiments of the present invention disclosed above are intended only to help illustrate the present invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the present invention to specific embodiments. Obviously, many modifications and variations are possible based on the contents of this specification. These embodiments are selected and described in detail in this specification to better explain the principles and practical applications of the present invention, thereby enabling those skilled in the art to better understand and utilize the present invention. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for real-time processing of multiple signals in urban rail transit, characterized by: The processing method comprises the following steps: S1: Continuously collect multiple types of signal data and corresponding operating status sets along the train running process; S2: Segmenting the multi-type signal data according to the operating state segments in the multiple operating state sets to obtain multiple signal units; S3: For each signal unit, calculate the first time window of adjacent sampling points in the time axis direction and the second time window in the running distance direction. Use the first time window and the second time window as the resolution to map the signal unit to a time-distance two-dimensional grid, obtain the two-dimensional coordinates of each original sampling point, and perform a three-neighborhood interpolation operation on each two-dimensional signal point to obtain an expanded two-dimensional signal coordinate set. S4: Based on the first time window and the second time window, the expanded two-dimensional signal coordinate set is mapped to the multidimensional signal space, and the numerical value of each additional parameter in the two-dimensional signal coordinate set is retained and mapped to the corresponding dimension; S5: Calculate the change rule of the running state set, and perform morphological transformation on multiple signal units in the multidimensional signal space according to the change rule; S6: Multiple morphologically transformed signal units are spliced ​​into a complete operating signal set in the order of state segments, realizing real-time restoration and alignment of multiple signals for the entire rail transit operation, and outputting the results to the train dispatching center in real time.

2. The method for real-time processing of multiple signals in urban rail transit according to claim 1, characterized in that: Step S2: Segmenting the multi-type signal data according to the multiple operating state segments to obtain multiple signal units, including the following steps: Calculating the total duration of the plurality of running state segments and the proportion of each running state segment in the total duration; Principal component analysis is performed on multi-type signal data to extract the main direction vector of signal changes; Constructing a first attitude matrix based on the main direction vector, and using the first attitude matrix to perform baseline alignment on the signal data, so that the change direction is aligned with the time axis, the secondary change direction is aligned with the running distance axis, and the remaining change components are aligned with the subsidiary parameter axis; Determine the enclosing interval of the aligned signal data, and determine the longest time dimension as the total length of the signal; Determine the time range of the signal unit corresponding to each state segment based on the ratio of the duration of each state segment and the total length of the signal; According to the time range corresponding to each state segment and the aligned signal time axis, the corresponding signal data is extracted and segmented to obtain multiple signal units.

3. The method for real-time processing of multiple signals in urban rail transit according to claim 2, characterized in that: The construction of the first posture matrix includes: aligning the direction vector corresponding to the largest eigenvalue with the time axis, aligning the direction vector corresponding to the second largest eigenvalue with the running distance axis, aligning the remaining direction vectors with the subsidiary parameter axis, and generating a rotation matrix, i.e., the first posture matrix, by orthogonalizing the eigenvector.

4. The method for real-time processing of multiple signals in urban rail transit according to claim 2, characterized in that: Before principal component analysis is performed on the multi-type signal data, an N×M data matrix is ​​constructed according to a unified time reference, where N is the number of sampling points and M is the number of signal channels. The signal channels include speed, voltage, current, temperature, and vibration signal channels after time synchronization.

5. The method for real-time processing of multiple signals in urban rail transit according to claim 4, characterized in that: The signal units include time series, spatial position series, speed, current, voltage and temperature; Each original sampling point of the signal unit is mapped to a row and column index position of a two-dimensional grid according to its time value and distance value, and the corresponding speed, current, voltage, and temperature signal channel values ​​are bound to the row and column index positions.

6. The method for real-time processing of multiple signals in urban rail transit according to claim 5, characterized in that: Step S3: performing a three-neighborhood interpolation operation on each two-dimensional signal point to obtain an expanded two-dimensional signal coordinate set, including the following steps: For the grid position not covered by the original sampling point, three known signal points in the two-dimensional grid that are closest to the grid position and located at different directions are selected; Calculate the time difference and distance difference between the grid position and each neighboring point, and combine them into a distance weight factor; Performing inverse distance weighted averaging on the signal channel values ​​of the three neighborhood points based on the weight factors to obtain interpolation results of each channel at the grid position; When there are less than three neighboring points, expand the search range until a neighboring point that meets the conditions is found.

7. The method for real-time processing of multiple signals in urban rail transit according to claim 1, characterized in that: Step S4: Based on the first time window and the second time window, the expanded two-dimensional signal coordinate set is mapped to the multidimensional signal space, and the numerical values ​​of each additional parameter in the two-dimensional signal coordinate set are retained and mapped to the corresponding dimension, including the following steps: Acquire a two-dimensional signal coordinate set that has been subjected to two-dimensional gridding and interpolation processing, wherein the two-dimensional signal coordinate set includes a time value, a distance value, and additional parameter values ​​of voltage, current, temperature, and vibration; Constructing a coordinate reference of a multidimensional signal space based on the first time window and the second time window, establishing a time index with the first time window as the minimum time resolution, establishing a distance index with the second time window as the minimum distance resolution, and assigning an independent parameter dimension index to each parameter according to the type of additional parameter value; Determine the corresponding time index and distance index based on the time value and distance value, assign each additional parameter value to the corresponding parameter dimension position, and establish a parameter vector containing each parameter value for each time-distance grid point; Normalize each parameter dimension in the mapped multidimensional signal space and convert the time, distance and parameter values ​​into floating-point numbers. The output is a multidimensional signal space dataset containing time dimension, distance dimension, and multiple parameter dimensions.

8. The method for real-time processing of multiple signals in urban rail transit according to claim 1, characterized in that: Step S5: Calculating the change rule of the running state set and performing morphological transformation on multiple signal units in the multidimensional signal space according to the change rule, including the following steps: Match the operating state segment and the corresponding state law vector for each signal unit; Calculate the deviation distribution between the actual change curve of the signal unit and the theoretical curve corresponding to the state law vector; Generate transformation rules based on the deviation distribution and perform parameter adjustment, introducing physical constraints such as non-negative speed and current not exceeding safety thresholds during the adjustment process; After the transformation is completed, the goodness of fit between the transformed signal curve and the theoretical curve is calculated, and the correlation changes between the parameter dimensions are verified.

9. The method for real-time processing of multiple signals in urban rail transit according to claim 1, characterized in that: Step S6: Splicing multiple transformed signal units into a complete operating signal set in the order of state segments to achieve real-time restoration and alignment of multiple signals for the entire operation of rail transit, including the following steps: Assign a unique sequence index to each signal unit according to the time sequence of each state segment in the running state set, and align the start and end times of each signal unit with the global time reference, and align the spatial coordinate interval with the global distance axis; The signal units are spliced ​​in the multidimensional signal space in the order of sequence index, including: Matching sampling points at the time and space boundaries of adjacent signal units; The time-distance-parameter matrices of each signal unit are spliced ​​into a continuous multi-dimensional matrix structure according to the global time and distance order; Perform weighted sliding average or piecewise polynomial interpolation on speed and current parameters within the state segment boundary interval; The fused and aligned full-route operation signal set is encapsulated into segmented data packets according to timestamp, distance coordinate, parameter vector and state segment identifier, and sent to the train automatic control system and dispatching center in real time through the rail transit communication protocol.

10. A real-time multi-signal processing system for urban rail transit, used to implement the processing method according to any one of claims 1 to 9, characterized in that: It includes data acquisition and segmentation module, data mapping module, morphological transformation module and output module; Data acquisition and segmentation module: continuously collects multiple types of signal data and corresponding operating status sets along the train running process, and segments the multiple types of signal data according to the operating status segments in multiple operating status sets to obtain multiple signal units; Data mapping module: For each signal unit, the first time window of adjacent sampling points in the time axis direction and the second time window in the running distance direction are calculated. The signal unit is mapped to a time-distance two-dimensional grid with the first and second time windows as the resolution to obtain the two-dimensional coordinates of each original sampling point. A three-neighborhood interpolation operation is performed on each two-dimensional signal point to obtain the expanded two-dimensional signal coordinate set; Morphological transformation module: Based on the first time window and the second time window, the expanded two-dimensional signal coordinate set is mapped to the multidimensional signal space, and the numerical values ​​of each additional parameter in the two-dimensional signal coordinate set are retained and mapped to the corresponding dimension. The change pattern of the operating state set is calculated, and the morphological transformation of multiple high-precision signal units in the multidimensional signal space is performed according to the change pattern. Output module: Splices multiple morphologically transformed signal units into a complete operating signal set in the order of state segments, realizes real-time restoration and alignment of multiple signals for the entire operation of rail transit, and outputs the results to the train dispatching center in real time.

Images