Track circuit interference test analysis method, system and equipment
By using multi-channel synchronous acquisition and intelligent diagnosis based on Gaussian mixture models, the problem of accurately locating mixed interference sources and capturing intermittent faults in existing track circuit testing equipment has been solved. This has enabled accurate location and objective diagnosis of track circuit interference, improving fault investigation efficiency and the consistency of diagnostic results.
Patent Information
- Application Number
- CN202511938664.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-22
- Publication Date
- 2026-01-20
- Estimated Expiration
- 2045-12-22
AI Technical Summary
Existing track circuit testing equipment cannot accurately locate mixed interference sources, cannot capture intermittent faults, and diagnostic conclusions rely subjectively on experience and lack intelligent analysis capabilities.
The system employs multi-channel synchronous acquisition of voltage and current signals from the track circuit, performs real-time storage and frequency and time domain analysis, and combines Gaussian mixture models for intelligent diagnosis, generating real-time and diagnostic analysis reports.
It enables precise location and objective diagnosis of track circuit interference, improves fault diagnosis efficiency and consistency of diagnostic results, and enhances the ability to detect hidden faults.
Smart Images

Figure CN121364353A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application belongs to the field of rail transit signal technology, and particularly relates to a rail circuit interference test analysis method, system and device. BACKGROUND
[0002] The rail circuit is the basis of the railway signal system, and its reliability is directly related to train safety and transportation efficiency. In modern railway operation, the working environment of the rail circuit is increasingly complex, and it is inevitably affected by various electromagnetic interferences, such as traction return current and its harmonics generated when electric locomotives are running, signal crosstalk between adjacent rail circuits, and even radiation interference from other communication or power equipment along the line. These interference signals are superimposed with the working signals of the rail circuit itself (such as 25Hz, UM71, etc.), forming a complex composite signal, which is an important reason for abnormal operation of the rail circuit and causes faults such as "red light band" (i.e. false display of track occupation).
[0003] To troubleshoot such faults, the existing technology mainly provides two types of test means: one is a large-scale automated test bench in the maintenance base, and the other is a portable test instrument used by field maintenance personnel. However, both types of devices have obvious limitations in actual application, which directly leads to the following technical problems that remain unresolved:
[0004] Fault root cause is difficult to locate: the existing field test instrument has limited ability to process mixed signals, and when multiple interferences exist simultaneously, it cannot accurately decompose the composite signal into individual frequency components. This leads to the fact that even if the maintenance personnel detects an abnormality, they cannot determine the root cause of the fault - whether it is due to excessive 50Hz power frequency interference, specific frequency traction harmonics, or other unknown sources of interference. Ultimately, the "trial and error" troubleshooting method must be used, which is inefficient and cannot guarantee that the problem will be completely solved.
[0005] Occasional faults are difficult to capture: many serious faults that affect train operation are caused by transient and occasional strong interferences. Most existing test instruments require personnel to operate on site, and cannot perform long-term continuous unattended monitoring. This leads to the fact that when a fault occurs, the test instrument may not be connected to the system, thus missing the key fault moment, making such "come and go" faults a persistent problem in maintenance, posing a huge risk to train safety.
[0006] Diagnostic conclusions are subjective and dependent on experience: due to the lack of deep intelligent analysis capability, existing devices can usually only provide raw waveforms or simple effective value data. How to interpret fault information from these data highly depends on the personal experience and knowledge level of the maintenance personnel. For personnel with insufficient experience, it is difficult to form a clear diagnostic conclusion from these data, leading to different judgments of the same problem by different people, making the fault handling process lack objectivity and standardization.
[0007] Therefore, there is an urgent need in the art for a new technical solution that not only enables long-term continuous monitoring of the track circuit to capture all data, but also enables automatic, accurate analysis and intelligent diagnosis of the collected complex mixed signals, thereby objectively and efficiently locating the root cause of the fault. SUMMARY
[0008] To solve the above problems in the prior art, i.e., difficulty in accurately locating the mixed interference source, inability to capture occasional faults, and subjective diagnosis, the present application provides a track circuit interference test analysis method, system, and device.
[0009] In a first aspect, the present application provides a track circuit interference test analysis method, comprising:
[0010] Multi-channel synchronous acquisition of voltage signals and current signals of the track circuit to obtain original time-domain waveform signals;
[0011] Real-time storage of the original time-domain waveform signals to form historical data;
[0012] Real-time processing of the original time-domain waveform signals, the real-time processing including: frequency domain analysis of the original time-domain waveform signals to resolve different frequency components contained therein, and time domain analysis of the original time-domain waveform signals to calculate real-time values and change rates thereof; comparison of the resolved frequency components, real-time values, and change rates with preset alarm thresholds, and triggering of an alarm when the thresholds are exceeded, interception of waveform data before and after the alarm time, and generation of a real-time analysis report containing fault information;
[0013] Processing of the historical data, the processing including: reading the historical data and performing time domain and frequency domain analysis of the composite signals under normal operating conditions to establish a probability model of normal operating conditions; input of historical data at the time of fault occurrence and the time periods before and after the time of fault occurrence into the probability model for verification and probability calculation; identification of fault characteristics based on the calculation results, and generation of a diagnostic analysis report containing the cause of the fault;
[0014] Output of the real-time analysis report or the diagnostic analysis report.
[0015] Further, the multi-channel synchronous acquisition of voltage signals and current signals of the track circuit comprises:
[0016] Synchronous acquisition of voltage signals through a plurality of mutually electrically isolated voltage acquisition channels, and synchronous acquisition of current signals through a plurality of mutually electrically isolated current acquisition channels;
[0017] The signals obtained by each acquisition channel are subjected to analog-digital conversion to generate digitized original time-domain waveform signals, wherein the frequency band range of the acquisition covers from the normal working signal frequency of the track circuit to the potential interference signal frequency.
[0018] Further, the real-time storage of the original time-domain waveform signals comprises:
[0019] The original time-domain waveform signals are attached with time identifiers and cached in the form of data units with the time identifiers;
[0020] The cached data units are continuously written into a non-volatile storage medium to form a continuous historical data sequence with a preset time length, and when the storage capacity reaches a preset upper limit, the earliest stored historical data is overwritten in accordance with the time sequence.
[0021] Further, the frequency-domain analysis of the original time-domain waveform signals to resolve different frequency components contained therein comprises:
[0022] A specific frequency band to be analyzed and a center frequency thereof are determined; and the original time-domain waveform signals are subjected to digital frequency shifting according to the center frequency and an original sampling frequency, so that the spectrum of the specific frequency band is shifted to a baseband to obtain a frequency-shifted signal;
[0023] The frequency-shifted signal is subjected to an anti-aliasing low-pass filtering, and the cutoff frequency of the low-pass filtering is determined according to a predetermined frequency resolution magnification and the original sampling frequency;
[0024] The low-pass filtered signal is subjected to a re-sampling processing with a reduced sampling rate, and the sampling interval of the re-sampling is equal to the frequency resolution magnification;
[0025] The re-sampling processed signal is subjected to a fast Fourier transform to calculate the spectrum thereof, so that a spectrum with an improved frequency resolution is obtained within the specific frequency band;
[0026] Different frequency signal components and their amplitude information are separated and resolved from the original time-domain waveform signals according to the spectrum.
[0027] Further, the preset alarm threshold comprises an amplitude upper threshold and an amplitude lower threshold independently set according to the types of the resolved different frequency signal components, a signal real-time value change rate threshold set based on historical data or theoretical values, and a frequency offset range threshold set for a specific frequency component;
[0028] The comparison includes: comparing the amplitude of the parsed specific frequency component with its corresponding upper amplitude threshold and lower amplitude threshold, comparing the calculated signal real-time value change rate with the change rate threshold, and / or comparing the parsed signal frequency with its corresponding frequency offset range threshold, wherein any comparison result exceeding the threshold triggers an alarm.
[0029] Further, the probability model of the normal working condition is established, including:
[0030] Select data of a plurality of non-fault periods from the historical data as training data;
[0031] Extract time domain features and frequency domain features from the composite signal corresponding to the training data;
[0032] Mathematically transform the extracted features to make them conform to Gaussian distribution;
[0033] Based on the transformed feature data, model the probability distribution of the composite signal features under normal working condition by a Gaussian mixture model, and the Gaussian mixture model is used to calculate the probability that the input features belong to the normal working condition.
[0034] Further, the historical data of the fault occurrence time and the time periods before and after the fault occurrence are input into the probability model for verification and probability calculation, and the fault features are identified based on the calculation results, including:
[0035] Input the historical data of the fault period into the Gaussian mixture model to calculate the probability value that the fault period data belong to the normal working condition;
[0036] When the probability value is lower than a preset fault determination threshold, determine that the fault period data is abnormal;
[0037] Using the data of the fault period and the calculated probability value thereof, adjust the parameters of the Gaussian mixture model by a parameter optimization algorithm to update the probability model;
[0038] Based on the updated probability model, recalculate the feature probability of the fault period data and identify at least one key feature that causes the probability value to decrease, the key feature corresponding to a specific signal frequency component or a time domain parameter.
[0039] Further, the method further includes:
[0040] Receive and respond to external control instructions, including configuration instructions for setting the signal type and range of each acquisition channel, control instructions for starting or stopping the signal acquisition process, and state query instructions for querying the current working state.
[0041] In a second aspect, the present application provides a track circuit interference test analysis system based on the track circuit interference test analysis method of the first aspect, which comprises:
[0042] a data acquisition module configured to synchronously collect voltage signals and current signals of the track circuit in multiple channels to obtain original time-domain waveform signals;
[0043] a data storage module configured to store the original time-domain waveform signals in real time to form historical data;
[0044] a real-time analysis report generation module configured to process the original time-domain waveform signals in real time, wherein the real-time processing comprises: performing frequency domain analysis on the original time-domain waveform signals to analyze different frequency components contained therein, and performing time domain analysis on the original time-domain waveform signals to calculate real-time values and change rates thereof; comparing the analyzed frequency components, real-time values and change rates with preset alarm thresholds, and triggering an alarm when the thresholds are exceeded, intercepting and saving waveform data before and after the alarm time, and generating a real-time analysis report containing fault information;
[0045] a diagnostic analysis report generation module configured to process the historical data, wherein the processing comprises: reading the historical data and performing time domain and frequency domain analysis on the composite signals under normal working conditions to establish a probability model of the normal working conditions; inputting historical data of the time when the fault occurs and the time periods before and after the time into the probability model for verification and probability calculation; identifying fault characteristics based on the calculation results, and generating a diagnostic analysis report containing fault causes;
[0046] an output module configured to output the real-time analysis report or the diagnostic analysis report.
[0047] In a third aspect, the present application provides an electronic device comprising:
[0048] at least one processor;
[0049] a memory in communication connection with the at least one processor; wherein,
[0050] the memory stores instructions executable by the processor, and the instructions are used to be executed by the processor to realize a track circuit interference test analysis method.
[0051] The present application has the following advantages:
[0052] The present application can clearly decompose the collected composite signal into track circuit working signal, power frequency interference, each harmonic and unknown interference and other multiple independent frequency components by using high-resolution spectrum analysis algorithm such as ZOOM-FFT, and quantize the amplitude. This enables the maintenance personnel to directly and objectively determine the root interference source causing the fault without guessing, fundamentally discarding the trial-and-error troubleshooting method, and greatly improving the troubleshooting efficiency.
[0053] The acquisition extension of the present application supports long-time uninterrupted recording of signal data, ensuring that any signal anomaly, no matter how short its duration, can be recorded completely. This provides complete and reliable first-hand data for subsequent fault analysis and reproduction, fills the gap in the field of transient fault monitoring in the prior art, and greatly enhances the ability to discover strong hidden safety hazards.
[0054] The present application introduces an intelligent diagnosis model (such as GMM) based on unsupervised learning. The model autonomously constructs a health state benchmark of the track circuit through learning of a large amount of normal data. When a fault occurs, the system can automatically analyze the deviation characteristics of the fault data from the model, thereby outputting objective and quantitative diagnosis conclusions and possible cause directions. This changes the fault diagnosis process from subjective judgment to scientific analysis based on data models, ensuring the consistency and reliability of the diagnosis results, and is particularly helpful in improving the fault handling ability of inexperienced personnel. BRIEF DESCRIPTION OF DRAWINGS
[0055] Other features, objects and advantages of the present application will become more apparent from the following detailed description of non-limiting embodiments made with reference to the accompanying drawings:
[0056] Figure 1 is a flowchart of a track circuit interference test analysis method of the present application;
[0057] Figure 2 is an acquisition extension architecture diagram of a track circuit interference test analysis method of the present application. DETAILED DESCRIPTION
[0058] The present application will be further described in detail below with reference to the accompanying drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the related application, and not to limit the application. In addition, it should be noted that only the parts related to the application are shown in the drawings for ease of description.
[0059] It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict. The present application will be described in detail below with reference to the accompanying drawings and embodiments.
[0060] The first embodiment of the present application provides a track circuit interference test analysis method, comprising:
[0061] Step S10, multi-channel synchronous acquisition of voltage signals and current signals of the track circuit is performed to obtain original time domain waveform signals;
[0062] Step S20, the original time domain waveform signals are stored in real time to form historical data;
[0063] Step S30, the original time domain waveform signals are processed in real time, and the real-time processing includes: frequency domain analysis of the original time domain waveform signals to analyze different frequency components contained therein, and time domain analysis of the original time domain waveform signals to calculate real-time values and change rates thereof; the analyzed frequency components, real-time values and change rates are compared with preset alarm thresholds, and an alarm is triggered when the thresholds are exceeded, waveform data before and after the alarm time is intercepted and saved, and a real-time analysis report containing fault information is generated;
[0064] Step S40, the historical data are processed, and the processing includes: reading the historical data and performing time domain and frequency domain analysis on composite signals under normal working conditions to establish a probability model of the normal working conditions; historical data at the time of fault occurrence and time periods before and after the time are input into the probability model for verification and probability calculation; fault characteristics are identified based on the calculation results, and a diagnostic analysis report containing fault causes is generated;
[0065] Step S50, the real-time analysis report or the diagnostic analysis report is output.
[0066] In order to more clearly describe the track circuit interference test analysis method of the present application, the following will be combined with Figure 1 The steps in the embodiment of the present application will be described in detail, including steps S10-S40, and the detailed description of each step is as follows:
[0067] Step S10, multi-channel synchronous acquisition of voltage signals and current signals of the track circuit is performed to obtain original time domain waveform signals;
[0068] In this embodiment, the multi-channel synchronous acquisition of voltage signals and current signals of the track circuit includes:
[0069] Step S11, voltage signals are synchronously acquired through a plurality of mutually electrically isolated voltage acquisition channels, and current signals are synchronously acquired through a plurality of mutually electrically isolated current acquisition channels;
[0070] Step S12, analog-to-digital conversion is performed on the signals obtained by each acquisition channel to generate digitized original time domain waveform signals, wherein the frequency band range of the acquisition covers from the normal working signal frequency of the track circuit to the potential interference signal frequency.
[0071] As Figure 2 shown in the embodiment, the multi-channel synchronous acquisition is realized by a dedicated acquisition extension. The acquisition extension adopts modular hardware design, and the core is a data acquisition module including 8 independent analog-digital conversion channels and a digital signal processor. Specifically, the first to fourth channels are designed as voltage acquisition channels, and the fifth to eighth channels are designed as current acquisition channels. Each channel can acquire and process up to 16834 data points per second, and adopts electrical isolation design with an isolation strength of not less than 5000V, ensuring the independence and safety of the signals of each channel.
[0072] The acquisition extension integrates signal sensing, processing, storage and communication as a whole. In the signal acquisition and conditioning part, the voltage channel adopts a high-impedance differential input circuit, which can be directly connected to the track circuit voltage test point. The current channel is equipped with a through-type open-close current transformer, which supports non-intrusive measurement of rail return current or traction current. After amplification and filtering, all channel signals are sent to the core processing module for high-fidelity analog-digital conversion and real-time processing.
[0073] The data processing and storage part takes a digital signal processor as the core, which is responsible for adding accurate time stamps to the acquired data, packaging and caching, and realizing long-time continuous waveform recording through the built-in large-capacity non-volatile memory, supporting cyclic coverage storage. The extension also integrates WIFI and USB communication interfaces, supporting wireless data transmission and remote command interaction. The power management module is built-in rechargeable lithium battery, with state indication and lightning protection grounding design, combined with industrial-grade protective shell, to ensure stable and reliable operation of the equipment in complex railway field environment.
[0074] In order to realize safe and reliable field measurement, especially in complex electromagnetic environment and possible high potential difference of railway signal field, strict electrical isolation design is adopted between all channels, with an isolation strength of 5000V. This isolation measure effectively prevents signal crosstalk, measurement error and even equipment damage caused by common ground or potential difference between different measurement points, ensuring the safety and data independence of each channel and the back-end processing unit when measuring large current signals such as traction return current or voltage signals with common-mode voltage. For the analog signals obtained by each acquisition channel, a high-performance analog-digital converter is used for analog-digital conversion to generate high-fidelity digital original time-domain waveform signals.
[0075] The collected frequency band range covers 0 Hz to 8000 Hz, and this wide frequency band design is considered to cover the normal working signal frequency of the track circuit to all potential interference signal frequencies. Specifically, it can completely collect the basic working signal of the 25Hz phase-sensitive track circuit, the 50Hz power frequency traction current and its harmonic interference, and the center frequency and side frequency signals of the widely used UM71, WG-21, ZPW-2000 series track circuits, and the standard carrier frequency includes 1700Hz, 2000Hz, 2300Hz and 2600Hz. In addition, this range also reserves the collection capability for other possible high-frequency interference. The collection of current signals is usually realized by means of a through-opening current transformer, which does not need to disconnect the measured conductor, and can be directly clamped on the rail lead-in line or return line, facilitating on-site rapid deployment and measurement. Through the above-mentioned mode, the acquisition extension can obtain the multi-channel voltage and current original waveform of the key monitoring point of the track circuit at one time and synchronously, and provides a comprehensive, synchronous and high-fidelity data basis for subsequent comprehensive analysis of interference components and positioning of fault sources. The present application overcomes the defects that the existing test equipment cannot simultaneously process multiple mixed signals or lacks synchronization, and lays a solid data acquisition foundation for accurately analyzing fault phenomena under the coupling of multiple interference sources.
[0076] Step S20, the original time domain waveform signal is stored in real time to form historical data, specifically including:
[0077] Step S21, a time identifier is attached to the original time domain waveform signal, and the data unit with the time identifier is cached;
[0078] Step S22, the cached data unit is continuously written into a non-volatile storage medium to form a continuous historical data sequence with a preset time length, and when the storage capacity reaches the preset upper limit, the earliest stored historical data is overwritten according to the time sequence.
[0079] In this embodiment, the real-time storage function is completed by the storage module in the acquisition extension. Specifically, after generating the original time domain waveform signal, the digital signal processor will immediately attach a high-precision time identifier to each data frame, and the time identifier is derived from the real-time clock inside the device, ensuring that all collected data have accurate time stamps. These data with time identifiers are organized into fixed data units and first temporarily stored in the digital signal processor or special cache. Then, the storage control logic continuously and sequentially writes the data units in the cache into a non-volatile storage medium, such as a large-capacity FLASH memory, thereby forming a continuous historical data sequence with strict time sequence relationship.
[0080] The storage module is designed to accommodate no less than 72 hours of continuously collected data to meet the needs of long-term monitoring. When the capacity of the storage medium reaches the preset upper limit, the system will automatically start the cyclic coverage mechanism, i.e., the latest data is used to cover the earliest historical data in chronological order, so that long-term and uninterrupted automatic recording is realized without manual intervention to replace the storage medium. The present application ensures that the device can work continuously without human intervention, completely captures all data in the key time window before and after the fault occurs, provides a complete and reliable data archive for subsequent in-depth fault backtracking and intelligent analysis based on historical data, and effectively solves the problem that existing devices cannot record for a long time.
[0081] Step S30, real-time processing of the original time-domain waveform signal, the real-time processing including: frequency domain analysis of the original time-domain waveform signal to analyze different frequency components contained therein, while time domain analysis of the original time-domain waveform signal to calculate its signal real-time value and change rate; comparing the analyzed frequency components, real-time value and change rate with preset alarm thresholds, and triggering an alarm when the thresholds are exceeded, intercepting and saving waveform data before and after the alarm time, and generating a real-time analysis report containing fault information;
[0082] Among them, the frequency domain analysis of the original time-domain waveform signal to analyze different frequency components contained therein includes:
[0083] Step S31, determining a specific frequency band to be analyzed and a center frequency thereof; according to the center frequency and the original sampling frequency, digitally shifting the frequency of the original time-domain waveform signal to shift the spectrum of the specific frequency band to the baseband to obtain a frequency-shifted signal;
[0084] Step S32, anti-aliasing low-pass filtering of the frequency-shifted signal, the cutoff frequency of the low-pass filtering being determined according to a predetermined frequency resolution magnification and the original sampling frequency;
[0085] Step S33, re-sampling processing of the low-pass filtered signal to reduce the sampling rate, the sampling interval of the re-sampling being equal to the frequency resolution magnification;
[0086] Step S34, performing fast Fourier transform on the re-sampling processed signal to calculate its spectrum, thereby obtaining a spectrum with improved frequency resolution in the specific frequency band;
[0087] Step S35, separating and analyzing different frequency signal components and their amplitude information from the original time-domain waveform signal according to the spectrum.
[0088] In the embodiment, the preset alarm threshold includes an upper amplitude threshold and a lower amplitude threshold independently set according to the type of the parsed different frequency signal components, a signal real-time value change rate threshold set based on historical data or a theoretical value, and a frequency offset range threshold set for a specific frequency component;
[0089] The comparison includes comparing the amplitude of the parsed specific frequency component with the upper amplitude threshold and the lower amplitude threshold corresponding thereto, comparing the calculated signal real-time value change rate with the change rate threshold, and / or comparing the parsed signal frequency with the frequency offset range threshold corresponding thereto, wherein any comparison result exceeding the threshold triggers an alarm.
[0090] In the embodiment, the real-time processing is performed by an analysis host. The analysis host receives a real-time data stream of the original time-domain waveform signal from the acquisition extension through a wireless network or a USB connection. The real-time processing performs two core tasks in parallel:
[0091] High-resolution frequency domain analysis and continuous time domain monitoring. In terms of frequency domain analysis, in order to accurately separate different frequency components such as 25Hz track circuit signals, 50Hz traction currents and ZPW-2000 series signals from mixed signals and accurately obtain their amplitudes, the system uses a ZOOM-FFT algorithm. Specifically, first, according to a specific target frequency band to be analyzed, for example, 50Hz power frequency interference and its nearby harmonics need to be observed in detail, the center frequency of the frequency band is determined . Then, according to the formula , the original time-domain waveform signal is subjected to digital frequency shift processing, and the center frequency of the target frequency band is shifted to the vicinity of zero frequency (baseband), thereby obtaining a frequency-shifted signal , where is the original sampling frequency. In order to prevent spectral aliasing when subsequent downsampling, the frequency-shifted signal needs to be subjected to anti-aliasing low-pass filtering.
[0092] The device preferably uses an infinite impulse response (IIR) filter to implement this low-pass filtering, which is more computationally efficient when achieving the same filtering performance and more suitable for real-time processing requirements. It should be emphasized that the low-pass filtering here is a digital processing step for the frequency-shifted signal inside the ZOOM-FFT algorithm, and the cutoff frequency is dynamically determined according to the current analysis target narrowband, with the purpose of ensuring the quality of the frequency spectrum after downsampling. This processing does not affect the complete acquisition capability of the system front end for the original wideband signal. The system can set different center frequencies to realize high-resolution analysis of each working frequency band of the track circuit (such as 1700Hz, 2000Hz, etc.) in turn. The specific implementation of the IIR filter is defined by the following difference equation: Wherein, is the order of the filter or the length of the time domain data involved, and are filter coefficients determined in advance according to the filter performance (such as the cutoff frequency, ripple) design, is the current output of the filter, is the past output value, is the past input value. By recursively calculating the equation, low-pass filtering can be efficiently completed, with the cutoff frequency strictly set according to Shannon's theorem, not greater than , where D is the predetermined frequency resolution magnification, i.e., the decimation multiple. The filtered signal is obtained. After that, the filtered signal is resampled according to the magnification D, i.e., one sample is extracted from every D samples, so as to reduce the effective sampling rate to . In order to facilitate the use of standard point number (such as N1 points) FFT algorithm in subsequent fast Fourier transform, it is usually necessary to pad the data sequence after extraction to a point number, for example, by using the method of zero padding at the end of the sequence, the data length is restored or adjusted to N1 points for easy calculation. Then, the N1-point sequence after padding is executed for fast Fourier transform. The FFT calculation itself can be realized by using the optimized butterfly algorithm, and the calculation process can be expressed as: . Wherein, and represent the data at the even and odd index positions of the padded sequence, and are rotation factors.
[0093] Through this series of operations, the frequency resolution in the target frequency band can be significantly improved from the original to without increasing the FFT operation point number, so as to obtain extremely high spectral analysis accuracy in the narrow band of interest, and accurately separate and quantify the amplitude and phase information of each frequency component from the composite signal according to the high-resolution spectrum. At the same time, in terms of time domain analysis, the system calculates the instantaneous value, true effective value of each channel signal in real time, and calculates the change rate of the signal based on the time sequence. A set of alarm threshold library that can be flexibly configured is pre-stored in the analysis host. These thresholds include amplitude upper and lower threshold values set independently according to the types of different frequency signal components, signal real-time value change rate threshold values set based on historical normal data statistical analysis or theoretical values, and allowed frequency deviation range threshold values set for specific frequency components such as track circuit working signals.
[0094] The comparison process is parallel and comprehensive: the system will continuously compare the amplitude of a specific frequency component (such as the 50Hz component) resolved from the frequency domain analysis with its independently configured upper and lower amplitude limits; compare the rate of change of the signal real-time value (such as the voltage effective value) calculated from the time domain analysis with the corresponding rate of change threshold; at the same time, the offset of the resolved key signal frequency from its nominal frequency is also compared with the allowed frequency offset threshold. If any of the above comparison results exceeds its corresponding preset threshold, the system determines that an anomaly has occurred and triggers an alarm. Once the alarm is triggered, the system will automatically intercept and permanently save the complete raw waveform data and related analysis results within a preset time window before and after the alarm time, and immediately generate a structured real-time analysis report. The report records the time, channel, specific over-limit parameter (such as "50Hz current amplitude over upper limit"), over-limit value, and associated waveform segment identifier, etc.
[0095] For example, in a certain railway field, the analysis host processes a segment of real-time signal from the 5th channel (current channel) of the acquisition extension. The frequency domain analysis module uses the ZOOM-FFT algorithm to perform high-resolution analysis in a narrow band around 50Hz, and calculates the amplitude of the current 50Hz current component in real time as 120A. The time domain analysis module synchronously calculates the rate of change of the current effective value of the channel as 15A per second. At the same time, the frequency analysis module measures the frequency of the current track circuit working signal as 1698.5Hz.
[0096] In the system's preset alarm threshold library, the upper limit of the "50Hz traction current" component is set to 100A; the rate of change threshold of the "current effective value" is set to 10A per second; and the allowed frequency offset range of the "ZPW-2000 track circuit signal" is set to the nominal 1700Hz ± 1Hz.
[0097] In parallel comparison, the system finds that: 1) the 50Hz current component amplitude 120A > upper limit 100A; 2) the current effective value rate of change 15A / s > threshold 10A / s; 3) the signal frequency 1698.5Hz is still within the allowed frequency offset range (1699Hz to 1701Hz) and has not exceeded the limit. According to the rule that "any comparison result exceeding the limit triggers an alarm", since the first two conditions are met simultaneously, the system immediately determines an anomaly.
[0098] The alarm is triggered at 14:30:05 on October 26, 2023. The system automatically intercepts and saves the complete original waveform data of the 5th channel, the real-time calculated spectrum and trend data in the time window from 14:29:55 to 14:30:15 (20 seconds) immediately after the alarm is triggered. At the same time, a real-time analysis report is generated, which structurally records the alarm triggering time as "2023-10-26 14:30:05", the alarm channel as "CH5-current", the over-limit parameters and values as "50Hz current component amplitude exceeds the upper limit, current value 120A, threshold value 100A" and "current effective value change rate exceeds the threshold value, current value 15A / s, threshold value 10A / s", and the report also associates the unique identifier of the saved waveform data segment.
[0099] The present application realizes accurate analysis of the frequency and amplitude of the interference signal through the ZOOM-FFT algorithm, significantly improves the accuracy and real-time performance of fault identification by combining multi-dimensional and customizable threshold comparison, and can timely capture transient interference and slowly changing faults and automatically retain key evidence, providing direct and reliable basis for on-site emergency disposal.
[0100] Step S40, processing the historical data, the processing including: reading the historical data and performing time domain and frequency domain analysis on the composite signal under normal working conditions to establish a probability model of normal working conditions; inputting the historical data of the fault occurrence time and the time period before and after the fault occurrence into the probability model for verification and probability calculation; identifying fault features based on the calculation results, and generating a diagnostic analysis report containing fault causes;
[0101] In this embodiment, the probability model of normal working conditions is established, including:
[0102] Step S41, selecting a plurality of data of time periods without faults from the historical data as training data;
[0103] Step S42, extracting time domain features and frequency domain features from the composite signal corresponding to the training data;
[0104] Step S43, performing mathematical transformation on the extracted features to make them conform to Gaussian distribution;
[0105] Step S44, modeling the probability distribution of the composite signal features under normal working conditions based on the transformed feature data through a Gaussian mixture model, and the Gaussian mixture model is used to calculate the probability that the input features belong to normal working conditions.
[0106] In this embodiment, inputting the historical data of the fault occurrence time and the time period before and after the fault occurrence into the probability model for verification and probability calculation, and identifying fault features based on the calculation results, include:
[0107] Step S45, input the historical data of the fault period into the Gaussian mixture model, and calculate the probability value thereof belonging to the normal working condition;
[0108] Step S46, when the probability value is lower than a preset fault determination threshold, determine that the fault period data is abnormal;
[0109] Step S47, use the data of the fault period and the calculated probability value thereof, and adjust the parameters of the Gaussian mixture model through a parameter optimization algorithm, to update the probability model;
[0110] Step S48, based on the updated probability model, recalculate the feature probability of the fault period data, and identify at least one key feature causing the probability value to decrease, the key feature corresponding to a specific signal frequency component or a time domain parameter.
[0111] In the embodiment, the intelligent analysis and processing of the historical data is mainly completed by relying on the data intelligent analysis and processing logic module in the analysis host. The system reads the historical data stored by the collection extension, and selects a plurality of time period data not having faults as training data from the massive data in combination with the actual situation of railway line operation, which should be able to represent the normal operation state of the line under various working conditions. An unsupervised learning fault detection algorithm is used to systematically extract features capable of representing the signal state from the composite signals corresponding to the training data. These features include time domain features, such as the effective value, peak value, mean value and change trend of the signal in a time window, and frequency domain features, which are obtained by performing high-resolution spectrum analysis on the signal by using the aforementioned ZOOM-FFT algorithm, such as the amplitude of the 50Hz traction current, 25Hz track circuit signal and ZPW-2000 series (about 1700Hz) signal and the specific frequency component, specific frequency band energy or harmonic component ratio.
[0112] In order to adapt to the subsequent probability model, necessary mathematical transformation needs to be performed on the extracted features. Specifically, features conforming to the Gaussian distribution are selected as much as possible, and for those features not conforming to the Gaussian distribution, linear or nonlinear transformation methods are used to make their distribution closer to the Gaussian distribution. Then, based on the transformed feature data, the Gaussian mixture model is used to model the probability distribution of the feature vector of the composite signal under the normal working condition. The model parameters are estimated by using an algorithm such as expectation maximization, and finally a model capable of calculating the probability of any input feature vector belonging to the normal working condition distribution is obtained.
[0113] After the probability model is established, historical data of the time period before and after the failure occurrence is input into the probability model for verification and probability calculation, and the failure features are identified based on the calculation results. The specific steps include: first, according to the failure record or alarm time point, the data analysis time period is determined, and each item of collected data in the time period before and after the failure point is screened out. The data is input into the established Gaussian mixture model for calculation to obtain the probability value of belonging to the normal working condition. When the probability value is lower than a preset failure determination threshold, the system determines that the data in the failure period is abnormal. In order to continuously improve the diagnosis accuracy, the system adjusts the parameters of the Gaussian mixture model by using the data in the failure period through the optimization algorithm, so as to update the probability model.
[0114] Based on the updated model, the feature probability of the failure period data is recalculated, and it is analyzed in depth which specific features significantly deviate to cause the decrease of the overall probability value. These identified key features correspond to specific signal frequency components (for example, an abnormally increased 50Hz current component) or time domain parameters (for example, a sudden drop of track circuit voltage effective value). Based on this analysis, the system automatically generates a structured diagnostic analysis report which will clearly explain the failure causes analyzed, for example, "the failure is mainly caused by the abnormal increase of 50Hz traction current to 120A, accompanied by a 15% decrease of 1700Hz track circuit signal amplitude". The present application realizes intelligent and automatic root cause analysis of historical failure data, and improves the accuracy and efficiency of failure diagnosis.
[0115] Step S50, outputting the real-time analysis report or the diagnostic analysis report.
[0116] In the embodiment, outputting the analysis report is the final step of the system for information presentation and decision support. Whether it is the real-time analysis report generated by real-time processing or the diagnostic analysis report generated by deep mining of historical data, the output is realized through the man-machine interaction interface module of the analysis host and related data export functions.
[0117] After the alarm is triggered and the real-time analysis report is generated in the real-time processing process, the report will be dynamically displayed in a prominent area of the main window of the man-machine interaction interface of the analysis host, such as a special alarm information bar or a pop-up window. The report clearly lists the key information in a structured text form, usually including the timestamp of the alarm trigger accurate to seconds, the specific channel number of the alarm occurrence such as CH5, the signal type or frequency component judged as abnormal such as 50Hz traction current component, the specific parameter of the overrun such as amplitude and its value such as the current value 120A, the preset threshold value violated such as threshold value 100A, and the index or storage location of the associated original waveform data segment.
[0118] The interface synchronously highlights or automatically switches to the waveform and spectrum near the alarm time point, enabling the operation and maintenance personnel to intuitively see the specific form changes before and after the signal anomaly occurs. In addition, the real-time analysis report is automatically saved as a separate data file in the storage directory specified by the analysis host, and its format can be JSON, XML or a specific format log file. The file name usually contains information such as time and channel to facilitate tracing. In some embodiments, the system can also be configured to send alarm information and report summaries to remote servers or mobile terminals in real time through a network interface, enabling remote monitoring.
[0119] For diagnostic analysis reports, the output is usually completed after the user initiates an intelligent analysis instruction for a specific historical fault time period. After the analysis host completes fault feature recognition and root cause analysis based on the Gaussian mixture model, a more detailed report view is generated in the interface.
[0120] The report not only contains basic information such as fault time, location, and involved channels, but more importantly, it contains a fault cause analysis section. This section lists the analysis conclusions in clear text combined with key data, such as the analysis of this fault determining that the main cause is 50Hz traction current interference, with an amplitude higher than the normal range by up to 125A during the fault period; and the 1700Hz track circuit signal amplitude is suppressed and decreased by about 20%. The report also embeds or provides convenient links to key evidence used to reach this conclusion, such as 50Hz current amplitude comparison curves and 1700Hz signal spectrum comparison charts during the fault period and normal period. The complete diagnostic analysis report supports export and saving in standard document formats such as PDF or Word format, facilitating the inclusion of formal maintenance records or technical archives. The system can also update the diagnostic rules or feature patterns confirmed by repeated verification to the knowledge base or adjust the threshold parameters of real-time alarms, achieving self-learning and optimization of the system.
[0121] The present application converts complex signal processing and data analysis results into standardized and structured information that operation and maintenance personnel can directly understand and use for decision-making and guiding maintenance actions. This greatly shortens the time from discovering abnormalities to locating fault causes, improves the pertinence and efficiency of electric service maintenance operations, and also accumulates valuable standardized data for fault statistics, regularity research and preventive maintenance.
[0122] In the embodiment, the method further includes:
[0123] Receiving and responding to external control instructions, the control instructions including configuration instructions for setting the signal type and range of each acquisition channel, control instructions for starting or stopping the signal acquisition process, and state query instructions for querying the current working state.
[0124] In this embodiment, the method has the ability to interact with the operator and accept his control during operation, which is realized by receiving and responding to a series of external control instructions. These instructions are mainly generated by the analysis host and delivered to the acquisition extension through wireless WiFi connection or wired USB connection.
[0125] Specifically, the configuration instruction is used to remotely set the hardware parameters of the acquisition extension. The operator can independently select the signal type of each of the 8 acquisition channels on the human-computer interaction interface of the analysis host, i.e., specify whether the channel is used to measure a voltage signal or a current signal. At the same time, the appropriate range can be set for each channel, for example, a certain voltage channel is set to 0-10V range, or a certain current channel is set to 0-5A range, to ensure the accuracy and safety of signal acquisition. After the digital signal processor of the acquisition extension receives such configuration instructions through the transmission module, it will parse the instruction content and configure the analog-to-digital converter parameters and signal conditioning circuit of the corresponding channel accordingly, and return the configuration confirmation information to the analysis host after completion.
[0126] The control instruction is used to directly manage the life cycle of the signal acquisition process. The most important control instructions include "start acquisition" and "stop acquisition". When the analysis host issues the "start acquisition" instruction, the digital signal processor of the acquisition extension will start the synchronous data acquisition and storage process of all configured channels. Correspondingly, the "stop acquisition" instruction will make the acquisition extension orderly terminate data acquisition and enter the standby state. This realizes the precise remote control of the test process without manual operation by personnel near the device.
[0127] The state query instruction is used to obtain the working state information of the acquisition extension in real time. Such instructions can include "time query" instructions to query the real-time clock inside the acquisition extension, "remaining power query" instructions to evaluate the remaining power, and "maximum test duration query" instructions to calculate how long the current storage space can support the test time. After receiving the query instruction, the acquisition extension will read the current state data from the corresponding sensor or storage management, and feedback to the analysis host through the communication link, and display it clearly to the operator on the human-computer interaction interface.
[0128] The present application gives the operation and maintenance personnel the flexibility and remote control ability of the test device. Users can adjust the measurement parameters, start and stop the acquisition task and monitor the device status at any time according to the on-site test requirements, thereby significantly improving the efficiency and adaptability of test deployment, and truly realizing the intelligent and convenient management of on-site test, overcoming the limitations of the existing device that requires personnel to operate and configure continuously on site.
[0129] Although the steps are described in the above-mentioned order in the above-mentioned embodiments, it can be understood by those skilled in the art that, in order to achieve the effects of the embodiments, the different steps do not have to be executed in such an order, and can be executed simultaneously (in parallel) or in a reversed order, and these simple changes are within the protection scope of the present application.
[0130] The track circuit interference test analysis system of the second embodiment of the present application is based on a track circuit interference test analysis method, and the system comprises:
[0131] a data acquisition module configured to perform multi-channel synchronous acquisition on voltage signals and current signals of the track circuit to obtain original time-domain waveform signals;
[0132] a data storage module configured to store the original time-domain waveform signals in real time to form historical data;
[0133] a real-time analysis report generation module configured to perform real-time processing on the original time-domain waveform signals, wherein the real-time processing comprises: performing frequency domain analysis on the original time-domain waveform signals to analyze different frequency components contained therein, and simultaneously performing time domain analysis on the original time-domain waveform signals to calculate real-time values and change rates of the signals; comparing the analyzed frequency components, real-time values and change rates with preset alarm thresholds, and triggering an alarm when the thresholds are exceeded, intercepting and saving waveform data before and after the alarm time, and generating a real-time analysis report containing fault information;
[0134] a diagnostic analysis report generation module configured to process the historical data, wherein the processing comprises: reading the historical data and performing time domain and frequency domain analysis on composite signals in normal working conditions to establish a probability model of the normal working conditions; inputting historical data of a fault occurrence time and time periods before and after the fault occurrence time into the probability model for verification and probability calculation; identifying fault characteristics based on the calculation results, and generating a diagnostic analysis report containing fault causes;
[0135] an output module configured to output the real-time analysis report or the diagnostic analysis report.
[0136] It should be noted that the above embodiment provides a track circuit interference test analysis system, and only the above-mentioned division of each functional module is exemplified, and in actual application, the above-mentioned functions can be completed by different functional modules according to needs, that is, the modules or steps in the embodiment of the application are further decomposed or combined, for example, the modules of the above-mentioned embodiment can be combined into one module, or can be further split into multiple sub-modules to complete all or part of the functions described above. The names of the modules and steps involved in the embodiments of the application are only for distinguishing various modules and steps, and are not considered as improper limitation of the application.
[0137] The third embodiment of the application is a device, comprising:
[0138] at least one processor;
[0139] and a memory in communication connection with the at least one processor;
[0140] wherein the memory stores instructions executable by the processor, and the instructions are used to be executed by the processor to realize the above-mentioned track circuit interference test analysis method.
[0141] The fourth embodiment of the application is a computer readable storage medium, and the computer readable storage medium stores computer instructions, and the computer instructions are used to be executed by the computer to realize the above-mentioned track circuit interference test analysis method.
[0142] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the storage device and the processing device described above and the related description can refer to the corresponding process in the foregoing method embodiment, and will not be repeated here.
[0143] The terms "first", "second", and the like are used to distinguish similar objects, not to describe or indicate a particular order or sequence.
[0144] The term "comprising" or any other similar term is intended to cover non-exclusive inclusion, so that the process, method, article or equipment / device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or further includes the elements inherent to the process, method, article or equipment / device.
[0145] The technical scheme of the present application has been described in combination with the preferred embodiments shown in the drawings, but it is easy for those skilled in the art to understand that the protection scope of the present application is obviously not limited to these specific embodiments. Those skilled in the art can make equivalent changes or replacements to the related technical features without departing from the principles of the present application, and the technical schemes after the changes or replacements will all fall within the protection scope of the present application.
Claims
1. A method of analyzing track circuit interference tests, characterized by, The method comprises the following steps: Multi-channel synchronous acquisition of voltage signals and current signals of the track circuit to obtain original time-domain waveform signals; Real-time storage of the original time-domain waveform signals to form historical data; Real-time processing of the original time-domain waveform signals, which comprises: frequency domain analysis of the original time-domain waveform signals to analyze different frequency components contained therein, and time domain analysis of the original time-domain waveform signals to calculate real-time values and change rates thereof; comparison of the analyzed frequency components, real-time values and change rates with preset alarm thresholds, and triggering of an alarm when the thresholds are exceeded, interception and storage of waveform data before and after the alarm time, and generation of a real-time analysis report containing fault information; Processing of the historical data, which comprises: reading of the historical data and time domain and frequency domain analysis of composite signals in normal working conditions to establish a probability model of normal working conditions; input of historical data at the time of fault occurrence and time periods before and after the time into the probability model for verification and probability calculation; identification of fault characteristics based on the calculation results, and generation of a diagnostic analysis report containing fault causes; Output of the real-time analysis report or the diagnostic analysis report.
2. The method of claim 1, wherein, The multi-channel synchronous acquisition of voltage signals and current signals of the track circuit comprises: Synchronous acquisition of voltage signals through a plurality of mutually electrically isolated voltage acquisition channels, and synchronous acquisition of current signals through a plurality of mutually electrically isolated current acquisition channels; Analog-to-digital conversion of signals obtained by each acquisition channel to generate digitized original time-domain waveform signals, wherein the frequency band range of acquisition covers from the normal working signal frequency of the track circuit to the frequency of potential interference signals.
3. The method of claim 1, wherein, The real-time storage of the original time-domain waveform signals comprises: Attaching time identifiers to the original time-domain waveform signals, and buffering the data units with the time identifiers; Continuous writing of the buffered data units into a non-volatile storage medium to form a continuous historical data sequence with a preset time length, and covering the earliest stored historical data in accordance with the time sequence when the storage capacity reaches the preset upper limit.
4. The method of claim 1, wherein, The frequency domain analysis of the original time-domain waveform signals to analyze different frequency components contained therein comprises: Determining a specific frequency band to be analyzed and a center frequency thereof; digital frequency shifting of the original time-domain waveform signals according to the center frequency and the original sampling frequency to shift the spectrum of the specific frequency band to the baseband to obtain a frequency-shifted signal; Anti-aliasing low-pass filtering of the frequency-shifted signal, wherein the cutoff frequency of the low-pass filtering is determined according to a predetermined frequency resolution magnification and the original sampling frequency; Re-sampling processing of the low-pass filtered signal at a reduced sampling rate, wherein the sampling interval of the re-sampling is equal to the frequency resolution magnification; Fast Fourier transform of the re-sampling processed signal to calculate its spectrum, thereby obtaining a spectrum with improved frequency resolution in the specific frequency band; Separation and analysis of signal components and their amplitude information at different frequencies from the original time-domain waveform signals according to the spectrum.
5. The method of claim 1, wherein, The preset alarm threshold includes an upper amplitude threshold and a lower amplitude threshold independently set according to the types of the different frequency signal components, a signal real-time value change rate threshold set based on historical data or theoretical values, and a frequency offset range threshold set for a specific frequency component; The comparison includes: comparing the amplitude of the specific frequency component with the upper amplitude threshold and the lower amplitude threshold corresponding thereto, comparing the calculated signal real-time value change rate with the change rate threshold, and / or comparing the analyzed signal frequency with the frequency offset range threshold corresponding thereto, wherein any comparison result exceeding the threshold triggers an alarm.
6. The method of claim 1, wherein, The probability model of the normal working condition is established by: selecting data of multiple periods without faults from the historical data as training data; extracting time domain features and frequency domain features from the composite signals corresponding to the training data; performing mathematical transformation on the extracted features to make them conform to Gaussian distribution; modeling the probability distribution of the composite signal features under the normal working condition based on the transformed feature data by using a Gaussian mixture model, the Gaussian mixture model being used to calculate the probability that the input features belong to the normal working condition.
7. The method of claim 6, wherein, The historical data of the fault occurrence time and the time periods before and after the fault occurrence are input into the probability model for verification and probability calculation, and the fault features are identified based on the calculation results, including: inputting the historical data of the fault period into the Gaussian mixture model to calculate the probability value that the data belong to the normal working condition; determining that the data of the fault period are abnormal when the probability value is lower than a preset fault determination threshold; adjusting the parameters of the Gaussian mixture model by using the data of the fault period and the calculated probability value to update the probability model by using a parameter optimization algorithm; recomputing the feature probability of the data of the fault period based on the updated probability model, and identifying at least one key feature that causes the probability value to decrease, the key feature corresponding to a specific signal frequency component or a time domain parameter.
8. The method of claim 1, wherein, The method further includes: receiving and responding to external control instructions, the control instructions including configuration instructions for setting the signal types and ranges of each acquisition channel, control instructions for starting or stopping the signal acquisition process, and state query instructions for querying the current working state.
9. A track circuit interference test analysis system based on the track circuit interference test analysis method according to any one of claims 1 to 8, characterized by The system includes: a data acquisition module configured to perform multi-channel synchronous acquisition on the voltage signal and the current signal of the track circuit to obtain original time domain waveform signals; a data storage module configured to store the original time domain waveform signals in real time to form historical data; a real-time analysis report generation module configured to perform real-time processing on the original time domain waveform signals, the real-time processing including: performing frequency domain analysis on the original time domain waveform signals to analyze different frequency components contained therein, and performing time domain analysis on the original time domain waveform signals to calculate signal real-time values and change rates thereof; comparing the analyzed frequency components, real-time values and change rates with preset alarm thresholds, and triggering an alarm when the thresholds are exceeded, intercepting and saving waveform data before and after the alarm time, and generating a real-time analysis report containing fault information; a diagnostic analysis report generation module configured to process the historical data, the processing including: reading the historical data and performing time domain and frequency domain analysis on the composite signals under normal working conditions to establish a probability model of normal working conditions; inputting the historical data of the time period before and after the fault occurrence time into the probability model for verification and probability calculation; identifying fault features based on the calculation results, and generating a diagnostic analysis report containing the fault causes; an output module configured to output the real-time analysis report or the diagnostic analysis report.
10. An electronic device, comprising: comprising: at least one processor; a memory in communication connection with the at least one processor; wherein, the memory stores instructions executable by the processor, and the instructions are executed by the processor to implement the track circuit interference test analysis method of any one of claims 1-8.
Citation Information
Patent Citations
Ground train transient impulse interference field intensity analysis system
CN104714108A
Rail traffic real-time fault diagnosis method and system based on data comparative analysis
CN105045256A
Multi-frequency signal rapid frequency measurement method based on continuous segmentation secondary identification
CN116125137A
Fault detection and root cause analysis system for multi-source energy system
CN119760607A
GIL equipment fault detection method, system and equipment based on voiceprint dynamic coupling model, and storage medium
CN120779154A
Cited By
Interference source analysis method and system for partial discharge measurement of large oil-filled equipment
CN122090877A
