A rapid identification method for ballastless track defects based on wheel-rail noise and vibration waveforms

Abstract

This invention discloses a rapid identification method for ballastless track defects based on wheel-rail noise and vibration waveforms, comprising the following steps: S1 collecting vibration data in the ballastless track structure with frequencies between 0 and 2500 Hz, wherein vibration data not exceeding 400 Hz corresponds to the vibration frequency of the ballastless track bed, and the vibration data within this range is stored as vibration waveforms converted into images; vibration data exceeding 400 Hz corresponds to the audio frequency of wheel-rail noise, and the vibration data within this range is stored as noise audio characteristic data; S2 establishing a wheel-rail noise audio database and a ballastless track bed vibration waveform image database through long-term monitoring; S3 performing similarity analysis between the real-time collected train wheel-rail noise audio data and ballastless track bed vibration waveform data and the previously established databases to evaluate the service status of the track structure, and then determining whether it is a rail defect, fastener defect, or ballastless track bed defect, significantly improving the effectiveness, accuracy, and reliability of real-time identification of ballastless track structure defects.

Description

Technical Field

[0001] This invention belongs to the field of rail transit engineering monitoring technology, specifically relating to a rapid identification method for ballastless track defects based on wheel-rail noise and vibration waveforms. Background Technology

[0002] With the increase in railway operating mileage and the extension of service life, the workload of railway track maintenance is gradually increasing. To ensure the normal operation of railway tracks, the main inspection methods for rail transit engineering currently include: riding on trains, manual inspection, and comprehensive inspection vehicles. However, because the inspection methods mentioned above are cyclical, they cannot provide real-time updates on the track's service status.

[0003] Track structure defects are diverse and can be categorized structurally into rail defects, fastener defects, and ballast bed defects. Rail defects further include damage to turnouts and rail expansion joint components, rail corrugation, and dents in rail welded joints; fastener defects include broken fastener spring clips, loose fastener bolts, and missing fasteners; ballast bed defects include surface cracking, ballast bed damage, and gaps between ballast bed layers.

[0004] While traditional inspection methods can detect some defects in ballastless track structures, they are time-consuming and inefficient, and cannot identify hidden defects. Therefore, there is currently no effective monitoring method for the rapid identification of defects in ballastless track structures. Summary of the Invention

[0005] In response to one or more of the above-mentioned defects or improvement needs of the existing technology, the present invention provides a rapid identification method for ballastless track defects based on wheel-rail noise and vibration waveforms. This method can significantly improve the effectiveness, accuracy and reliability of ballastless track structural defect identification, enable timely early warning and corresponding maintenance, and reduce the labor intensity and labor costs of railway engineering departments, etc.

[0006] To achieve the above objectives, this invention provides a method for rapid identification of ballastless track defects based on wheel-rail noise and vibration waveforms, comprising the following steps:

[0007] S1 collects vibration data of the ballastless track structure with a frequency H between 0 and 2500 Hz. Vibration data within the range of 0 < H ≤ 400 Hz corresponds to the vibration frequency of the ballastless track bed. Vibration data within this range is stored as vibration waveforms converted into images. ＜ Vibration data with H≤2500Hz correspond to the audio frequency of wheel-rail noise, and vibration data within this range are stored as noise audio characteristic data.

[0008] S2 establishes a wheel-rail noise audio database and a ballastless track vibration waveform image database through long-term monitoring. The wheel-rail noise audio database includes a wheel-rail noise audio health database and a wheel-rail noise audio disease database. The ballastless track vibration waveform image database includes a healthy state vibration waveform database and a diseased state vibration waveform database.

[0009] S3 performs similarity analysis on the real-time collected wheel-rail noise audio data and ballastless track vibration waveform data of each train with the previously established database to evaluate the service status of the track structure and thus quickly identify ballastless track defects.

[0010] As a further improvement of the present invention, in step S3...

[0011] If both the wheel-rail noise audio data and the ballastless track vibration waveform data conform to the characteristics of health status data, then the track structure is in normal service condition.

[0012] If the wheel-rail noise audio data conforms to the characteristics of healthy state data, but the ballastless track vibration waveform data does not conform to the characteristics of healthy state data, then it is preliminarily determined that the ballastless track is in an abnormal state. Then, the ballastless track vibration waveform data is compared with the vibration waveform of the defect state to further determine the type and probability of the defect corresponding to the ballastless track.

[0013] If the vibration waveform data of the ballastless track bed conforms to the characteristics of healthy track data, but the wheel-rail noise audio data does not conform to the characteristics of healthy track data, it is initially determined that the rail or wheel is abnormal. The wheel-rail noise audio data of other passing trains is evaluated. If other trains do not have abnormal wheel-rail noise audio data, the abnormality is determined to be caused by abnormal wheel condition. If other trains also have abnormal wheel-rail noise audio data, the abnormality is determined to be caused by abnormal rail or fastener condition of that section of track structure. Finally, the wheel-rail noise audio data is compared with the defect audio characteristic data to determine the type and probability of the defect.

[0014] If the wheel-rail noise audio data and the ballastless track vibration waveform data do not meet the characteristics of healthy status data, the monitoring data of subsequent trains will be compared first. If the conclusions are consistent, the ballastless track will be judged to be abnormal; otherwise, the vehicle will be judged to be abnormal. Then, the real-time wheel-rail noise audio data and ballastless track vibration waveform data of the train will be compared and analyzed with the previously established defect database to determine the corresponding defects and their probability of occurrence.

[0015] As a further improvement of the present invention, the method for establishing the ballastless track bed vibration waveform image database includes the following steps:

[0016] Vibration data from 0 to 2500 Hz were collected from different track structure types. Vibration data with frequencies not exceeding 400 Hz were converted into images and stored in the corresponding ballastless track bed health status vibration waveform database and disease status vibration waveform database respectively.

[0017] Meanwhile, the characteristics of the vibration waveforms obtained from different ballastless track bed defects are stored in the vibration waveform database of ballastless track bed defect status.

[0018] As a further improvement of the present invention, the characteristics of the vibration waveform after analysis include spectrum, octave band and correlation characteristics.

[0019] As a further improvement of the present invention, the method for establishing the wheel-rail noise audio database includes the following steps:

[0020] Vibration data from 0 to 2500 Hz were collected from different track structure types, and vibration data with frequencies exceeding 400 Hz were stored as noise audio characteristic data.

[0021] Different types of noise audio feature data are classified and labeled, and audio data of normal driving in wheel-rail healthy sections, driving in track defect sections, and audio data generated by wheel abnormalities are extracted. The audio data is preprocessed by frame segmentation and windowing. By extracting features from the audio signal of each frame and merging these features, a new overall feature of the entire audio segment is obtained.

[0022] Extract the temporal and frequency domain feature parameters of the merged audio data to establish a wheel-rail noise audio database.

[0023] As a further improvement of the present invention, when classifying and labeling different types of noise audio feature data, it is necessary to remove train start audio data, train braking audio data, background white noise audio data when the train stops, station voice broadcast audio data, and alarm bell beeping audio data.

[0024] As a further improvement of the present invention, the time-domain characteristic parameters include short-time average energy, short-time average zero-crossing rate, peak-to-peak normal distribution statistical 95% confidence interval, and maximum vibration level; the frequency-domain characteristic parameters include 1 / 3 octave band spectral sound pressure, Mel frequency cepstral coefficients and their first or second order differences.

[0025] As a further improvement of the present invention, the wheel-rail noise audio defect database includes driving audio data of track defect sections and abnormal audio data generated by wheel abnormalities;

[0026] The audio data of the track defect section includes audio feature data corresponding to rail corrugation and rail joint damage; the abnormal audio data generated by wheel abnormalities includes audio feature data corresponding to wheel damage.

[0027] As a further improvement of the present invention, the vibration waveform image database of the ballastless track bed includes vibration waveforms corresponding to the following conditions: sleeper block damage, cracking or damage of the ballastless track bed, separation between the ballastless track bed and the sleeper, separation and detachment between the track bed slab and the base, broken spring of the steel spring floating slab track bed, settlement and deformation of the lower foundation, and arching of the lower foundation.

[0028] In summary, the technical solutions conceived by this invention have the following beneficial effects compared with the prior art:

[0029] This invention relates to a rapid identification method for ballastless track defects based on wheel-rail noise and vibration waveforms. Based on wheel-rail noise and vibration waveforms, and targeting potential defects in the later stages of ballastless track bed construction, vibration data from 0 to 2500 Hz is collected. This vibration data is stored according to different frequency ranges, presented as vibration waveform diagrams and audio data. Furthermore, the real-time collected data is compared with a database for similarity analysis to determine whether the defect is a rail defect, fastener defect, or ballastless track bed defect. This significantly improves the effectiveness, accuracy, and reliability of real-time identification of ballastless track structural defects, and enables timely early warning and corresponding maintenance, reducing the labor intensity and costs for railway maintenance departments. Attached Figure Description

[0030] Figure 1 This is a flowchart of a rapid identification method for ballastless track defects based on wheel-rail noise and vibration waveforms according to an embodiment of the present invention.

[0031] Figure 2 This is a flowchart illustrating the establishment of a vibration waveform image database for ballastless track bed based on vibration waveforms, according to an embodiment of the present invention.

[0032] Figure 3 The flowchart below illustrates the process of establishing a wheel-rail noise audio database based on wheel-rail noise, as described in this embodiment of the invention. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0034] This invention addresses potential defects in ballastless track during its later stages. By combining track vibration data acquired through vibration sensors, it proposes a rapid identification method for ballastless track defects based on wheel-rail noise and vibration waveforms. (See reference...) Figures 1 to 3 The present invention provides a rapid identification method for ballastless track defects based on wheel-rail noise and vibration waveforms, comprising the following steps:

[0035] (1) Vibration data of the ballastless track structure with frequencies H ranging from 0 to 2500 Hz were collected, where 0 ＜ Vibration data with H ≤ 400Hz are stored as images converted from vibration waveforms. ＜ Vibration data with H≤2500Hz are stored as noise audio characteristic data;

[0036] Ballastless track bed is a concrete structure, and its vibration characteristics change significantly after structural defects occur. Therefore, analyzing vibration waveform data provides the greatest discriminative power for differentiating between defects. The vibration characteristics of ballastless track bed are mainly low-frequency vibration data. Based on previously accumulated field test data and analysis results, 0 ＜ The frequency range of H≤400Hz can cover almost all the characteristic frequency points of the track bed.

[0037] Both wheels and rails are metal components, and different stages of rail damage and development are best distinguished by noise audio characteristics. Since wheel-rail noise is primarily high-frequency data, based on previously accumulated field test data and analysis results, 400... ＜ Vibration data with H≤2500Hz can almost cover all the characteristic frequency points of wheel-rail noise defects. 。

[0038] Preferably, the vibration data of the ballastless track structure is collected using sensors laid on the surface of the ballastless track bed.

[0039] (2) Through long-term monitoring, establish a wheel-rail noise audio database and a ballastless track vibration waveform image database. The wheel-rail noise audio database includes a wheel-rail noise audio health database and a wheel-rail noise audio disease database. The vibration waveform database includes a healthy state vibration waveform database and a diseased state vibration waveform database. Before establishing the audio database, it is necessary to classify and label different types of sounds.

[0040] Preferably, the wheel-rail noise audio defect database includes audio data of track defect sections and abnormal audio data caused by wheel abnormalities; the audio data of track defect sections includes audio feature data corresponding to rail corrugation and rail joint damage; the abnormal audio data caused by wheel abnormalities includes audio feature data corresponding to wheel damage.

[0041] (3) Based on big data learning, the real-time wheel-rail noise audio data and ballastless track vibration waveform data of each train are collected and compared with the database established in the early stage to conduct similarity analysis, to preliminarily evaluate the service status of the track structure, and then determine whether it is rail defect, fastener defect or ballastless track defect.

[0042] Specifically, if both the wheel-rail noise audio data and the ballastless track vibration waveform data meet the characteristics of health status data, then the track structure is in normal service condition.

[0043] If the wheel-rail noise audio data matches the characteristics of healthy status data, but the ballastless track vibration waveform data does not match the characteristics of healthy status data, then the ballastless track is preliminarily judged to be in abnormal condition. Then, the ballastless track vibration waveform data is compared with the vibration waveform of the defective condition to further determine the possible defect type and probability of the ballastless track.

[0044] The vibration waveform data of the ballastless track bed conforms to the characteristics of healthy state data, but the wheel-rail noise audio data does not conform to the characteristics of healthy state data. Therefore, it is initially determined that the abnormality is in the rail or wheel. Further, the wheel-rail noise audio data of other trains passing by are evaluated. If other trains do not have abnormal wheel-rail noise audio data, it means that the abnormality at that time was caused by abnormal wheel condition. If other trains also have abnormal wheel-rail noise audio data, it means that the abnormality is caused by the rail or fastener of the track structure in that section. Then, the wheel-rail noise audio defect database is used to determine the type and probability of the defect through the audio characteristics of the noise data.

[0045] If the wheel-rail noise audio data and the ballastless track vibration waveform data do not conform to the characteristics of healthy track data, then the monitoring data of subsequent trains should be compared first. If the conclusions are consistent, it can be determined that the ballastless track is abnormal; otherwise, it is a vehicle abnormality. Then, the real-time wheel-rail noise audio data and ballastless track vibration waveform data of this train should be compared and analyzed with the previously established defect database to determine the possible defects and their probability of occurrence.

[0046] Furthermore, track defect types can be identified using the defect database provided by this invention, and operation and maintenance suggestions can be provided, which can then be inspected and verified by maintenance personnel.

[0047] This invention, based on wheel-rail noise and vibration waveforms, targets potential defects in ballastless track beds in the later stages. It collects vibration data from 0 to 2500 Hz and stores the vibration data according to different frequency ranges, presenting both vibration waveforms and audio data. Furthermore, it performs similarity analysis between the real-time collected data and a database to determine whether the defect is a rail defect, fastener defect, or ballastless track bed defect. This significantly improves the effectiveness, accuracy, and reliability of real-time identification of defects in ballastless track structures, enabling timely early warning and appropriate maintenance, thus reducing the labor intensity and costs for railway maintenance departments.

[0048] Specifically, the process of establishing the ballastless track bed vibration waveform image database of the present invention is as follows: Figure 2 As shown, it includes the following steps:

[0049] 1) Collect vibration data from 0 to 2500 Hz in different track structure types, and convert the vibration data in the frequency domain range of 0 to 400 Hz (i.e., not exceeding 400 Hz) into images and store them in the corresponding ballastless track bed health state vibration waveform database and disease state vibration waveform database respectively.

[0050] Before establishing a vibration waveform database, it is necessary to classify and label different types of vibration waveforms. The vibration waveform database includes healthy vibration waveforms and defective vibration waveforms. Among them, the defective vibration waveforms of ballastless track include, but are not limited to, vibration waveforms corresponding to sleeper block damage, cracking or damage of ballastless track, separation between ballastless track and sleeper, separation and detachment between track slab and base, broken spring of steel spring floating slab track, settlement and deformation of the lower foundation, and arching of the lower foundation.

[0051] The measured vibration waveforms of the ballastless track bed were converted into photo format and stored in the corresponding vibration waveform databases for healthy and damaged ballastless track beds.

[0052] 2) The spectrum, overtones, coherence, and other characteristics of the vibration waveforms obtained from different ballastless track defects are stored in the ballastless track defect state vibration waveform database.

[0053] The vibration waveforms obtained from different ballastless track defects have specific spectral, octave band, and coherence characteristics after analysis.

[0054] Optionally, the vibration waveform image data of the ballastless track bed of the present invention can be stored in two ways: either by converting the images into binary data streams and storing them in a database in key-value pair format, or by storing the images on a server, with the database storing the storage path of the images on the server and establishing the corresponding attributes of the vibration waveform data. Both methods require comprehensive consideration of the database and server storage capabilities. Given the large volume of vibration waveform images, the method of storing the image storage path on the server in the database is preferred.

[0055] Furthermore, the process of establishing the wheel-rail noise audio database of the present invention is as follows: Figure 3 As shown, the specific steps include the following:

[0056] 1) Collect vibration data from 0 to 2500 Hz in different track structure types, and store the vibration data in the frequency domain range of 400 to 2500 Hz (i.e., above 400 Hz) as noise audio characteristic data.

[0057] 2) Classify and label different types of noise audio feature data, extract normal driving audio data of healthy track sections, driving audio data of track defect sections, and abnormal audio data caused by wheel abnormalities, perform frame-by-frame windowing preprocessing on the audio data; extract features from the audio signal of each frame, and merge these features to obtain new overall features of the entire audio segment;

[0058] Framing is the process of dividing a long segment of information (preferably 1 second) into several short-time signals, each of which can be considered stable. Signal framing can be understood as sampling the signal, grouping a certain number of continuous signal points into a frame. Each frame can be treated as a locally stationary audio signal. Then, by extracting features from the audio signal of each frame and merging these features, a new overall feature of the entire audio segment is obtained.

[0059] Preferably, when classifying and labeling different types of noise audio feature data, at least the following should be removed: train start audio data, train braking audio data, background white noise audio data when the train stops, station voice broadcast audio data, and alarm bell audio data.

[0060] 3) Extract the temporal and frequency domain feature parameters of the merged audio data to establish a wheel-rail noise audio database;

[0061] The wheel-rail noise audio database includes a wheel-rail noise audio health database and a wheel-rail noise audio disease database; wherein the wheel-rail noise audio disease database includes audio data of track disease sections and audio data generated by wheel anomalies; preferably, the audio data of track disease sections includes audio feature data corresponding to rail corrugation and rail joint damage; the abnormal audio data generated by wheel anomalies includes audio feature data corresponding to wheel damage.

[0062] Preferably, the time-domain characteristic parameters include short-time average energy, short-time average zero-crossing rate, peak-to-peak normal distribution statistical 95% confidence interval, and maximum vibration level; the frequency-domain characteristic parameters include 1 / 3 octave band spectral sound pressure, Mel frequency cepstral coefficients (400Hz~2500Hz) and their first or second order differences.

[0063] The wheel-rail noise audio database of the present invention is mainly established by storing the audio data format on a server, storing the storage path of the audio data format on the server in the database, and establishing various attributes of the corresponding audio data, thereby establishing a corresponding wheel-rail noise audio health database and wheel-rail noise audio disease database.

[0064] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for rapid identification of defects in ballastless track based on wheel-rail noise and vibration waveform, characterized in that, Includes the following steps: S1 collects vibration data in the ballastless track structure with a frequency H between 0 and 2500 Hz. Vibration data with a frequency H between 0 and 400 Hz corresponds to the vibration frequency of the ballastless track bed. Vibration data in this range is stored as vibration waveforms converted into images. Vibration data with a frequency H between 400 and 2500 Hz corresponds to the audio frequency of wheel-rail noise. Vibration data in this range is stored as noise audio characteristic data. S2 establishes a wheel-rail noise audio database and a ballastless track vibration waveform image database through long-term monitoring. The wheel-rail noise audio database includes a wheel-rail noise audio health database and a wheel-rail noise audio disease database. The ballastless track vibration waveform image database includes a healthy state vibration waveform database and a diseased state vibration waveform database. S3 performs similarity analysis on the real-time collected wheel-rail noise audio data and ballastless track vibration waveform data of each train with the previously established database to evaluate the service status of the track structure and thus quickly identify ballastless track defects.

2. The method for rapid identification of ballastless track defects based on wheel-rail noise and vibration waveforms according to claim 1, characterized in that, In step S3, If both the wheel-rail noise audio data and the ballastless track vibration waveform data conform to the characteristics of health status data, then the track structure is in normal service condition. If the wheel-rail noise audio data conforms to the characteristics of healthy state data, but the ballastless track vibration waveform data does not conform to the characteristics of healthy state data, then it is preliminarily determined that the ballastless track is in an abnormal state. Then, the ballastless track vibration waveform data is compared with the vibration waveform of the defect state in the defect state vibration waveform database to further determine the defect type and probability of the ballastless track. If the vibration waveform data of the ballastless track bed conforms to the characteristics of healthy state data, but the wheel-rail noise audio data does not conform to the characteristics of healthy state data, it is initially determined that the rail or wheel is abnormal. The wheel-rail noise audio data of other passing trains is evaluated. If other trains do not have abnormal wheel-rail noise audio data, the abnormality is determined to be caused by abnormal wheel condition. If other trains also have abnormal wheel-rail noise audio data, the abnormality is determined to be caused by abnormal rail or fastener condition of that section of track structure. Finally, the wheel-rail noise audio data is compared with the disease status audio characteristic data in the wheel-rail noise audio defect database to determine the type and probability of the defect. If the wheel-rail noise audio data and the ballastless track vibration waveform data do not meet the characteristics of healthy status data, the monitoring data of subsequent trains will be compared first. If the conclusions are consistent, the ballastless track will be judged to be abnormal; otherwise, the vehicle will be judged to be abnormal. Then, the real-time wheel-rail noise audio data and ballastless track vibration waveform data of the train will be compared and analyzed with the previously established defect database to determine the corresponding defects and their probability of occurrence.

3. The rapid identification method for ballastless track defects based on wheel-rail noise and vibration waveforms according to claim 1 or 2, characterized in that, The method for establishing the ballastless track bed vibration waveform image database includes the following steps: Vibration data from 0 to 2500 Hz were collected from different track structure types. Vibration data with frequencies not exceeding 400 Hz were converted into images and stored in the corresponding ballastless track bed health status vibration waveform database and disease status vibration waveform database respectively. Meanwhile, the characteristics of the vibration waveforms obtained from different ballastless track bed defects are stored in the vibration waveform database of ballastless track bed defect status.

4. The method for rapid identification of ballastless track defects based on wheel-rail noise and vibration waveforms according to claim 3, characterized in that, The characteristics of the vibration waveform after analysis include the spectrum, octave band, and correlation characteristics.

5. The method for rapid identification of ballastless track defects based on wheel-rail noise and vibration waveforms according to claim 1 or 2, characterized in that, The method for establishing the wheel-rail noise audio database includes the following steps: Vibration data from 0 to 2500 Hz were collected from different track structure types, and vibration data with frequencies exceeding 400 Hz were stored as noise audio characteristic data. Different types of noise audio feature data are classified and labeled, and audio data of normal driving in wheel-rail healthy sections, driving in track defect sections, and audio data generated by wheel abnormalities are extracted. The audio data is preprocessed by frame segmentation and windowing. By extracting features from the audio signal of each frame and merging these features, a new overall feature of the entire audio segment is obtained. Extract the temporal and frequency domain feature parameters of the merged audio data to establish a wheel-rail noise audio database.

6. The method for rapid identification of ballastless track defects based on wheel-rail noise and vibration waveforms according to claim 5, characterized in that, When classifying and labeling different types of noise audio feature data, it is necessary to remove train start audio data, train braking audio data, background white noise audio data when the train stops, station voice broadcast audio data, and alarm bell audio data.

7. The method for rapid identification of ballastless track defects based on wheel-rail noise and vibration waveforms according to claim 5, characterized in that, The time-domain characteristic parameters include short-time average energy, short-time average zero-crossing rate, peak-to-peak normal distribution statistical 95% confidence interval, and maximum vibration level; the frequency-domain characteristic parameters include 1 / 3 octave band spectral sound pressure, Mel frequency cepstral coefficients and their first or second order differences.

8. The rapid identification method for ballastless track defects based on wheel-rail noise and vibration waveforms according to any one of claims 1, 2, 6 or 7, characterized in that, The wheel-rail noise audio defect database includes audio data of track defect sections and abnormal audio data generated by wheel abnormalities. The audio data of the track defect section includes audio feature data corresponding to rail corrugation and rail joint damage; the abnormal audio data generated by wheel abnormalities includes audio feature data corresponding to wheel damage.

9. The method for rapid identification of ballastless track defects based on wheel-rail noise and vibration waveforms according to any one of claims 1, 2, 6 or 7, characterized in that, The vibration waveform image database of the ballastless track bed includes vibration waveforms corresponding to the following conditions: sleeper block damage, cracking or damage of the ballastless track bed, separation between the ballastless track bed and the sleeper, separation and detachment between the track bed slab and the base, broken spring of the steel spring floating slab track bed, settlement and deformation of the lower foundation, and arching of the lower foundation.

Images