A freight rail fastener fault detection method based on distributed optical fiber vibration sensing

Abstract

The application discloses a freight rail fastener fault detection method based on distributed optical fiber vibration sensing, and specifically comprises the following steps: S1, a 4-core single-mode armored vibration optical cable is laid along one side of a freight rail, and the vibration optical cable captures fastener abnormal vibration and train wheel impact vibration signals; S2, a collection host is arranged at a track monitoring station along a track to collect track vibration signals; S3, the collected signals are transmitted to a processing host, and the processing host converts the received vibration signals; S4, a comprehensive monitoring platform pre-processes the received signals, and inputs the pre-processed signals into a customized model to identify abnormal fasteners; and S5, hierarchical alarm is given according to the identified abnormal fasteners, and alarm information is pushed to operation and maintenance personnel through the comprehensive monitoring platform. The application focuses on the core function of freight rail fastener abnormal detection, adopts a standardized optical cable laying scheme, does not need to additionally add hardware devices, simplifies the detection process, and reduces operation and maintenance costs by 40%.

Description

Technical Field

[0001] This invention belongs to the field of rail transit fault diagnosis and detection technology, and in particular relates to a method for detecting faults in freight rail fasteners based on distributed optical fiber vibration sensing. Background Technology

[0002] Freight track fasteners, as core connecting components of the track structure, are subjected to long-term heavy-load locomotive pressure and outdoor environmental erosion, making them prone to loosening and breakage, directly compromising track stability and threatening the safe operation of freight railways. Existing distributed fiber optic vibration sensing technology is mostly adapted to passenger tracks or other fields (such as pipelines and bridges) for monitoring, and has not been customized for heavy-load freight track conditions. Its core shortcomings lie in: the lack of an integrated core architecture of "customized Fourier calculation + waterfall plot visualization," the absence of a fastener anomaly judgment system based on impact parameters (impact number, frequency, energy), and the lack of LSTM neural network reconstruction modeling adapted to heavy-load freight scenarios. This results in the inability to accurately distinguish between abnormal fastener vibrations and normal locomotive vibrations, leading to frequent invalid alarms, low maintenance efficiency, and high maintenance costs. Therefore, a streamlined, efficient, and highly adaptable fault detection method specifically for freight track fasteners is urgently needed to address these technical pain points. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art. This invention provides a method for detecting faults in freight track fasteners based on distributed optical fiber vibration sensing. By focusing on the core function of detecting abnormalities in freight track fasteners and adopting a standardized optical cable laying scheme, this invention eliminates the need for additional hardware equipment, simplifies the detection process, reduces maintenance costs by 40%, and is suitable for large-scale industrial applications in freight tracks. It differs from the high cost of complex detection systems for ordinary passenger tracks.

[0004] To achieve the above-mentioned objectives, the technical solution provided by this invention patent is as follows: A method for detecting faults in freight rail fasteners based on distributed fiber optic vibration sensing, the method specifically includes the following steps: S1, optical cable laying: a 4-core single-mode armored vibrating optical cable is laid along one side of the freight track. The vibrating optical cable captures abnormal vibration signals of fasteners and impact vibration signals of train wheels. An optical cable splice box is set every 1km along the track. S2, the data acquisition host is deployed at the monitoring stations along the track, the core parameters are configured, and the track vibration signal is collected; S3 transmits the collected signals to the processing host, which converts the received vibration signals and transmits the converted vibration signals to the integrated monitoring platform. S4, the integrated monitoring platform preprocesses the received signals, builds a feature library, reconstructs a customized model based on the LSTM long short-term memory neural network, and inputs the preprocessed signals into the customized model for abnormal fastener identification; S5 generates tiered alarms based on identified abnormal fasteners and pushes alarm information to maintenance personnel through the integrated monitoring platform.

[0005] Furthermore, during the laying of the optical cable, the distance between the optical cable and the main body of the track is 0.5m-3m, and the laying tension of the optical cable is 80N-100N.

[0006] Furthermore, the core parameter configuration of the acquisition host specifically includes a frequency response range of 15Hz-800Hz, a sampling frequency of 1kHz, and an optical fiber transmission loss threshold of <12.5dB.

[0007] Furthermore, the processing host has a built-in Fourier calculation module, which adopts a customized 256-point calculation window and a 0.1V frequency domain data extraction threshold. The processing host performs fast Fourier transform calculation on the collected time-domain vibration raw signal, converts the time-domain vibration raw signal into a frequency domain signal, and extracts the impact frequency and impact energy parameters. The frequency domain data is then synchronously transmitted to the integrated monitoring platform.

[0008] Furthermore, the integrated monitoring platform preprocesses the received signals by: employing a wavelet transform 5-layer multi-scale decomposition and reconstruction algorithm to denoise the original time-domain vibration signal, filtering out outdoor environmental clutter and mechanical interference signals; then, normalizing the amplitude of the denoised signal and the frequency domain parameters obtained by Fourier calculation, unifying data analysis standards, classifying abnormal vibration of fasteners from normal vibration of locomotives, improving subsequent identification accuracy, and adapting to the complex interference environment of freight tracks.

[0009] Furthermore, the feature library is constructed as follows: different types are classified based on the impact frequency and impact energy of the track fasteners, including slight loosening, more severe loosening, and breakage. When the impact frequency of the track fastener is 20-29Hz and the impact energy is 0.1-0.49J, it is classified as slight loosening; when the impact frequency of the track fastener is 30-49Hz and the impact energy is 0.5-0.9J, it is classified as more severe loosening; and when the impact frequency of the track fastener is ≥50Hz and the impact energy is ≥1.0J, it is classified as breakage.

[0010] Furthermore, the customized model reconstructed based on the LSTM long short-term memory neural network is specifically constructed as follows: 10,000 sets of abnormal vibration samples and normal vibration samples of freight rail fasteners are used for training, with a learning rate of 0.001 and 100 iterations to complete the training and construct the customized model; the network structure of the customized model is optimized to have 24 neurons in the input layer, 4 hidden layers, and 3 neurons in the output layer, focusing on the temporal analysis of impact parameters to adapt to the temporal characteristics of freight rail impact signals.

[0011] Furthermore, the customized model identifies the input signals and simultaneously counts the number of impacts, impact frequency, and impact energy for each signal; when the number of impacts is ≥5 or the impact parameters exceed the corresponding threshold, it is determined that the fastener is abnormal; the impact parameters include impact frequency and impact energy.

[0012] Furthermore, the abnormal fasteners are classified into three levels of alarms: Level 1: breakage or high-frequency, high-energy loosening, requiring immediate action; Level 2: relatively serious loosening, requiring action within a specified time limit; Level 3: minor loosening, requiring action in conjunction with routine inspections; Alarm information including the location of the abnormal fastener, the type of abnormality, the core impact parameters, and action suggestions are pushed to maintenance personnel.

[0013] Based on the above technical solution, the fault detection method for freight rail fasteners based on distributed optical fiber vibration sensing, as proposed in this invention, has achieved the following technical effects through practical application: 1. The present invention provides a method for detecting fastener faults in freight rails based on distributed fiber optic vibration sensing. Through a core architecture of "customized Fourier calculation and waterfall plot visualization" combined with a customized LSTM model, it can accurately distinguish between fastener abnormalities and normal locomotive vibrations. The abnormality identification accuracy is ≥95%, the invalid alarm rate is reduced by 80%, and the ±5m positioning accuracy meets the requirements for precise handling of freight rails. This method is different from the low anti-interference defects of existing machine vision detection solutions.

[0014] 2. The present invention provides a method for detecting faults in freight rail fasteners based on distributed optical fiber vibration sensing. By focusing on the core function of detecting abnormalities in freight rail fasteners and adopting a standardized optical cable laying scheme, it eliminates the need for additional hardware equipment, simplifies the detection process, reduces maintenance costs by 40%, and is suitable for large-scale industrial applications in freight rail, unlike the high cost of complex detection systems for ordinary passenger rail. Attached Figure Description

[0015] Figure 1 This is a data processing flowchart of a freight rail fastener fault detection method based on distributed optical fiber vibration sensing according to the present invention.

[0016] Figure 2This is a schematic diagram of track fastener fault detection in a freight track fastener fault detection method based on distributed optical fiber vibration sensing according to the present invention.

[0017] Figure 3 This is a waterfall diagram showing the transformation of an embodiment of the freight rail fastener fault detection method based on distributed optical fiber vibration sensing according to the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and effects of this invention clearer, specific examples are provided below. However, it should be understood that these descriptions are merely exemplary and not intended to limit the scope of the invention. Furthermore, descriptions of well-known structures and technologies are omitted in the following description to avoid unnecessarily obscuring the concepts of this invention.

[0019] Example 1, such as Figure 1-3 As shown, a method for detecting faults in freight rail fasteners based on distributed optical fiber vibration sensing is presented. This method specifically includes the following steps: S1, fiber optic cable laying: A 4-core single-mode armored vibrating fiber optic cable is laid along one side of the freight track. The vibrating fiber optic cable captures abnormal vibration signals of fasteners and impact vibration signals of train wheels. A fiber optic splice box is set every 1km along the track line. Special clamps are used to strengthen the fixation of the fiber optic cable in key areas where fasteners are prone to abnormality, such as turnouts and bridges, to prevent the fiber optic cable from shifting due to locomotive heavy load vibration, ensure the stability of signal acquisition, and adapt to the long-term heavy load operation requirements of freight tracks. S2, the data acquisition host is deployed at the monitoring stations along the track, the core parameters are configured, and the track vibration signal is collected; The distributed vibration fiber optic sensor host collects raw vibration signals across the entire track area at a sampling frequency of 1kHz and an interval of 5m. It focuses on capturing abnormal vibration signals of fasteners and impact vibration signals generated by train wheels passing over abnormal fasteners, providing high-quality raw data for subsequent processing and adapting to the heavy-load impact signal collection needs of freight tracks. S3 transmits the collected signals to the processing host, which converts the received vibration signals and transmits the converted vibration signals to the integrated monitoring platform. S4, the integrated monitoring platform preprocesses the received signals, builds a feature library, reconstructs a customized model based on the LSTM long short-term memory neural network, and inputs the preprocessed signals into the customized model for abnormal fastener identification; S5 generates tiered alarms based on identified abnormal fasteners and pushes alarm information to maintenance personnel through the integrated monitoring platform.

[0020] When laying the optical cable, the distance between the optical cable and the main body of the track is 0.5m-3m, and the laying tension of the optical cable is 80N-100N.

[0021] The core parameters of the acquisition host are specifically configured as follows: frequency response range of 15Hz-800Hz, sampling frequency of 1kHz, and fiber optic transmission loss threshold of <12.5dB.

[0022] The processing host has a built-in Fourier calculation module, which adopts a customized 256-point calculation window and a 0.1V frequency domain data extraction threshold. The processing host performs fast Fourier transform calculation on the collected time-domain vibration raw signal, converts the time-domain vibration raw signal into a frequency domain signal, and extracts the impact frequency and impact energy parameters. The frequency domain data is then synchronously transmitted to the integrated monitoring platform.

[0023] The integrated monitoring platform preprocesses the received signals by employing a wavelet transform 5-layer multi-scale decomposition and reconstruction algorithm to denoise the original time-domain vibration signal, filtering out outdoor environmental clutter and mechanical interference signals. Then, it performs amplitude normalization processing on the denoised signal and the frequency domain parameters obtained by Fourier calculation, unifies the data analysis standards, classifies abnormal vibration of fasteners from normal vibration of locomotives, improves subsequent identification accuracy, and adapts to the complex interference environment of freight tracks.

[0024] The integrated monitoring platform is equipped with a waterfall chart visualization module, which generates a waterfall chart with track distance on the horizontal axis, time on the vertical axis, and vibration intensity represented by color depth. It clearly defines the quantitative correspondence between color depth and impact energy (the darker the color, the greater the impact energy). By simulating fastener loosening, breakage, and normal locomotive operation vibration signals using standard vibration sources, a system benchmark signal library is constructed, and the communication link is debugged to ensure that the signal transmission delay is ≤500ms, meeting the real-time detection requirements of freight tracks. The waterfall chart visualization module generates a clear waterfall chart, which intuitively distinguishes between normal locomotive operation vibration and abnormal fastener impact vibration, providing visual support for subsequent identification and solving the pain point that existing technologies cannot accurately distinguish between the two types of vibration.

[0025] The feature library is constructed as follows: different types of track fasteners are classified based on the impact frequency and impact energy, including slight loosening, more severe loosening, and breakage. When the impact frequency of the track fastener is 20-29Hz and the impact energy is 0.1-0.49J, it is classified as slight loosening; when the impact frequency of the track fastener is 30-49Hz and the impact energy is 0.5-0.9J, it is classified as more severe loosening; and when the impact frequency of the track fastener is ≥50Hz and the impact energy is ≥1.0J, it is classified as breakage.

[0026] The customized model reconstructed based on the LSTM (Long Short-Term Memory) neural network is specifically constructed as follows: 10,000 sets of abnormal vibration samples and normal vibration samples of freight rail fasteners are used for training, with a learning rate of 0.001 and 100 iterations to complete the model training and construct the customized model. The network structure of the customized model is optimized to have 24 neurons in the input layer, 4 hidden layers, and 3 neurons in the output layer, focusing on the temporal analysis of impact parameters to adapt to the temporal characteristics of freight rail impact signals. After training, the model's recognition accuracy is ≥98%.

[0027] Model optimization: The distributed vibration fiber optic sensing system is automatically self-calibrated every 30 days to correct detection errors; combined with historical data on freight rail maintenance and signal acquisition, the LSTM neural network reconstruction modeling parameters are iteratively optimized to continuously improve the accuracy of fastener anomaly identification and impact parameter statistics, adapting to the dynamic changes in freight rail operating conditions.

[0028] The customized model identifies the input signals and simultaneously counts the number of impacts, impact frequency, and impact energy for each signal. When the number of impacts is ≥5 or the impact parameters exceed the corresponding threshold, it is determined that the fastener is abnormal. The impact parameters include impact frequency and impact energy.

[0029] The abnormal fastener alarm system is tiered as follows: Anomaly registration is divided into a three-level alarm mechanism, including: Level 1: Breakage or high-frequency, high-energy loosening, immediate action required; Level 2: More serious loosening, time-limited action required; Level 3: Minor loosening, handled in conjunction with routine inspections. Alarm information, including the location of the abnormal fastener, anomaly type, core impact parameters, and handling suggestions, is pushed to maintenance personnel. High adaptability: The customized design is specifically adapted to the heavy-load working conditions of freight rail. The sensor host has strong anti-interference ability (signal-to-noise ratio ≥35dB after noise reduction), supports long-distance full-area monitoring of 50km / channel, and can cover key areas such as turnouts and bridges, meeting the routine operation and maintenance needs of freight rail. It is different from the existing fiber optic sensing and detection technology adapted to passenger rail or other fields.

[0030] Example 2 uses a 50km port-connecting freight rail line as a specific implementation example. This line is a single-track heavy-load freight rail line with a locomotive axle load of 30t. It includes 4 turnouts, 2 bridges, and a total of 12,000 sets of fasteners. Frequent mechanical interference from the surrounding area aligns with the heavy-load freight scenario detection requirements of this invention. The implementation equipment includes: 50km of 4-core single-mode armored vibration optical cable, 1 set of distributed vibration fiber optic sensor host, 1 integrated monitoring platform, 50 optical cable splice boxes, and several dedicated fixing clamps.

[0031] Implementation process: Step S1 completes standardized fiber optic cable laying (tension 80N~100N) and system debugging, sets customized Fourier calculation parameters (256-point window, 0.1V threshold), waterfall plot visualization parameters, and constructs a reference signal library; Step S2 completes track vibration signal acquisition (1kHz sampling frequency, 5m interval), Fourier calculation and preprocessing (wavelet transform 5-layer noise reduction, db4 wavelet basis), and filters out port machinery interference signals; Step S3 constructs a feature library and customized LSTM modeling (24-4-3 neuron structure), completes fastener anomaly identification, location and impact parameter statistics; Step S4 implements graded alarms and model optimization (self-calibration every 30 days).

[0032] Actual test results: This method can accurately identify fastener loosening (impact frequency 20-49Hz) and breakage (impact frequency ≥50Hz), with a positioning error ≤5m. The statistical accuracy of impact parameters meets the needs of operation and maintenance. The invalid alarm rate is reduced by 85% compared with existing technologies, and the operation and maintenance efficiency is improved by 60%. It effectively reduces the workload and cost of operation and maintenance of fasteners for heavy-duty freight rail. It is suitable for the industrial application of distributed fiber optic sensing technology in the field of heavy-duty rail transit and is significantly different from the low anti-interference and short-distance detection defects of existing machine vision inspection solutions.

[0033] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications can still be made to the specific implementation of the invention or equivalent substitutions can be made to some technical features without departing from the spirit of the technical solutions of the present invention, and all such modifications and substitutions should be covered within the scope of the technical solutions claimed in the present invention.

Claims

1. A method for detecting faults in freight rail fasteners based on distributed fiber optic vibration sensing, characterized in that, The method specifically includes the following steps: S1, optical cable laying: a 4-core single-mode armored vibrating optical cable is laid along one side of the freight track. The vibrating optical cable captures abnormal vibration signals of fasteners and impact vibration signals of train wheels. An optical cable splice box is set every 1km along the track. S2, the data acquisition host is deployed at the monitoring stations along the track, the core parameters are configured, and the track vibration signal is collected; S3 transmits the collected signals to the processing host, which converts the received vibration signals and transmits the converted vibration signals to the integrated monitoring platform. S4, the integrated monitoring platform preprocesses the received signals, builds a feature library, reconstructs a customized model based on the LSTM long short-term memory neural network, and inputs the preprocessed signals into the customized model for abnormal fastener identification; S5 generates tiered alarms based on identified abnormal fasteners and pushes alarm information to maintenance personnel through the integrated monitoring platform.

2. The method for detecting faults in freight rail fasteners based on distributed optical fiber vibration sensing according to claim 1, characterized in that, When laying the optical cable, the distance between the optical cable and the main body of the track is 0.5m-3m, and the laying tension of the optical cable is 80N-100N.

3. The method for detecting faults in freight rail fasteners based on distributed optical fiber vibration sensing according to claim 1, characterized in that, The core parameters of the acquisition host are specifically configured as follows: frequency response range of 15Hz-800Hz, sampling frequency of 1kHz, and fiber optic transmission loss threshold of <12.5dB.

4. The method for detecting faults in freight rail fasteners based on distributed optical fiber vibration sensing according to claim 1, characterized in that, The processing host has a built-in Fourier calculation module, which adopts a customized 256-point calculation window and a 0.1V frequency domain data extraction threshold. The processing host performs fast Fourier transform calculation on the collected time-domain vibration raw signal, converts the time-domain vibration raw signal into a frequency domain signal, and extracts the impact frequency and impact energy parameters. The frequency domain data is then synchronously transmitted to the integrated monitoring platform.

5. The method for detecting faults in freight rail fasteners based on distributed optical fiber vibration sensing according to claim 1, characterized in that, The integrated monitoring platform preprocesses the received signals by employing a wavelet transform 5-layer multi-scale decomposition and reconstruction algorithm to denoise the original time-domain vibration signal, filtering out outdoor environmental clutter and mechanical interference signals. Then, it performs amplitude normalization processing on the denoised signal and the frequency domain parameters obtained by Fourier calculation, unifies the data analysis standards, classifies abnormal vibration of fasteners from normal vibration of locomotives, improves subsequent identification accuracy, and adapts to the complex interference environment of freight tracks.

6. The method for detecting faults in freight rail fasteners based on distributed optical fiber vibration sensing according to claim 5, characterized in that, The feature library is constructed as follows: different types of track fasteners are classified based on the impact frequency and impact energy, including slight loosening, more severe loosening, and breakage. When the impact frequency of the track fastener is 20-29Hz and the impact energy is 0.1-0.49J, it is classified as slight loosening; when the impact frequency of the track fastener is 30-49Hz and the impact energy is 0.5-0.9J, it is classified as more severe loosening; and when the impact frequency of the track fastener is ≥50Hz and the impact energy is ≥1.0J, it is classified as breakage.

7. The method for detecting faults in freight rail fasteners based on distributed optical fiber vibration sensing according to claim 5, characterized in that, The customized model reconstructed based on LSTM (Long Short-Term Memory) neural network is specifically constructed as follows: 10,000 sets of abnormal vibration samples and normal vibration samples of freight rail fasteners are used for training, with a learning rate of 0.001 and 100 iterations to complete the training and construct the customized model. The network structure of the customized model is optimized to have 24 neurons in the input layer, 4 in the hidden layer, and 3 in the output layer, focusing on the temporal analysis of impact parameters to adapt to the temporal characteristics of freight rail impact signals.

8. A method for detecting faults in freight rail fasteners based on distributed optical fiber vibration sensing according to claim 5, characterized in that, The customized model identifies the input signals and simultaneously counts the number of impacts, impact frequency, and impact energy for each signal. When the number of impacts is ≥5 or the impact parameters exceed the corresponding threshold, it is determined that the fastener is abnormal. The impact parameters include impact frequency and impact energy.

9. The method for detecting faults in freight rail fasteners based on distributed optical fiber vibration sensing according to claim 1, characterized in that, The abnormal fastener alarm system is classified into three levels: Level 1: breakage or high-frequency, high-energy loosening, requiring immediate action; Level 2: relatively serious loosening, requiring action within a specified time; Level 3: minor loosening, requiring action in conjunction with routine inspections. Alarm information, including the location of the abnormal fastener, the type of abnormality, the core impact parameters, and action recommendations, is pushed to maintenance personnel.

Images