Support state evaluation method for railroad slabs
By employing time-frequency analysis and a learning data set to evaluate slab track support, the method addresses the inefficiencies of conventional devices, enabling accurate and efficient detection of voids in slab tracks with a reduced device size.
Patent Information
- Application Number
- JP2023081562
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-05-17
- Publication Date
- 2024-11-28
AI Technical Summary
Conventional hammering sound testing devices for slab tracks are large and inefficient due to the need for multiple striking and sound collection positions, making it difficult to accurately detect voids in the packed layer beneath the track slab.
A method involving time-frequency analysis of sound pressure waveforms, using a learning data set to create a spectrogram, calculate a threshold value, and estimate voids based on the ratio of absolute error, reducing the number of impact and sound collection positions while maintaining accuracy.
This approach allows for a more compact testing device and improved inspection efficiency by accurately detecting voids in the packed layer, reducing the number of impact positions and sound collection points without compromising evaluation accuracy.
Smart Images

Figure 2024165394000001_ABST
Abstract
Description
[Technical field]
[0001] The present invention relates to a method for evaluating the support condition of a track slab, and more particularly to a method for evaluating the support condition of a track slab in a slab track having a filling layer below the track slab. [Background technology]
[0002] Conventionally, known types of track for railways include ballasted track, in which a ballast track bed is laid on the roadbed, and sleepers and rails are laid on the track bed, and slab track, in which a track slab, a concrete plate-like member, is placed on a concrete track bed or viaduct, and rails are laid on the track slab. Slab track is less susceptible to track displacement, which reduces the effort required for maintenance.
[0003] In slab track, a layer of mortar or other materials is placed between the bottom surface of the track slab and the top surface of the concrete track bed. This layer of mortar is known to deteriorate due to long-term use and frost damage caused by repeated freezing and thawing of the interstitial water, and this deterioration can cause gaps to form between the bottom surface of the track slab and the top surface of the layer of mortar. These gaps need to be repaired, but because the layer of mortar is located under the track slab, it is difficult to see them directly from the slab track.
[0004] There are various known methods for detecting the state of such a filling layer. For example, a method using a hammering test device for slab track, as described in Patent Document 1, is known.
[0005] The hammering sound test device for slab track described in Patent Document 1 is attached to a vehicle that can run on the rails of a slab track, and comprises support beams extending in the width direction of the slab track, and a number of impact devices supported by the support beams lined up in the width direction of the slab track, and the impact devices drop weights, and a control device including a storage device that stores the sound pressure and load value of the impact sounds measured when the impact devices strike the upper surface of the track slab of the slab track, and the impact devices strike the upper surface of the track slab at a predetermined height from the upper surface of the track slab, and the control device evaluates the presence or absence of voids on the underside of the track slab based on the ratio of the sound pressure to the load value.
[0006] According to the method for evaluating the support condition of a track slab using such a slab track hammering test device, the height of the impact device that drops the weight from the top surface of the track slab is kept constant, so that a constant impact force can be applied to each point on the top surface of the track slab. This makes it possible to perform a highly accurate hammering test and accurately detect voids that have occurred on the underside of each point on the track slab. [Prior art documents] [Patent documents]
[0007] [Patent Document 1] JP 2022-80347 A Summary of the Invention [Problem to be solved by the invention]
[0008] However, in the conventional method of evaluating the support condition of a track slab using an impact test device for a slab track, the evaluation is performed using a time history response waveform of sound pressure or a frequency response spectrum. As a result, it is necessary to place the impact position where the weight is dropped and the sound collection position where the sound pressure and load value are measured close to each other. This necessitates the provision of multiple impact devices, making it difficult to reduce the size of the impact test device for a slab track.
[0009] In addition, as mentioned above, conventional hammering test equipment for slab track needs to strike about 150 to 200 points per slab track, and there has been a demand to reduce the number of striking points from the viewpoint of inspection efficiency. However, when multiple sound collection positions are set for one striking position, it is extremely difficult to capture the acoustic difference with the configuration of the conventional hammering test equipment for slab track.
[0010] Therefore, the present invention has been made in consideration of the above problems, and aims to provide a method for evaluating the support condition of track slabs, which can reduce the number of striking positions and sound collection positions to reduce the size of the test equipment, while also ensuring evaluation accuracy even with the reduction in the number of striking positions and sound collection positions. [Means for solving the problem]
[0011] The present invention relates to a method for evaluating the support condition of a track slab on a slab track having a filling layer below the track slab, the method comprising the steps of: striking an arbitrary striking position on the track slab; measuring a time history response waveform of sound pressure due to the striking at a plurality of sound collection positions at a predetermined distance from the striking position; performing time-frequency analysis on the time history response waveform of sound pressure to obtain a judgment spectrogram; calculating a mean absolute error using the judgment spectrogram and a mapping and projection generated from a previously prepared learning dataset; calculating a threshold from the learning dataset; calculating an anomaly degree from the ratio of the mean absolute error and the threshold; and estimating a gap range from the relationship between the anomaly degree and the gap depth.
[0012] In addition, in the method for evaluating the support condition of a track slab according to the present invention, it is preferable that the learning data set is a collection of learning spectrograms obtained by performing multiple impact tests on a slab track in which the track slab is completely supported by a filling layer, and performing time-frequency analysis on the time history response waveforms of sound pressure obtained in the impact tests.
[0013] In addition, in the method for evaluating the support condition of a track slab according to the present invention, the step of obtaining a threshold value from the learning data set includes a step of dividing the learning data set into data for mapping generation and data for threshold calculation, a step of compressing and restoring the data for mapping generation to generate a mapping, and a step of generating a projection of the data for threshold calculation onto the mapping, and it is preferable to obtain the threshold value by averaging the absolute values of pixel differences in a spectrogram from the projection and the data for threshold calculation over the total number of pixels.
[0014] In addition, in the method for evaluating the support condition of a track slab according to the present invention, it is preferable that the impact position is an edge of the track slab in a width direction, and the sound collection position is arranged at a predetermined interval inward in the width direction from the impact position.
[0015] In the method for evaluating the support condition of a track slab according to the present invention, it is preferable that the estimation of the gap range is performed by creating a contour of the degree of abnormality in the track slab.
[0016] In the track slab support condition evaluation method according to the present invention, it is preferable that the judgment spectrogram is three-dimensional image data consisting of three axes of time, frequency and spectral intensity.
[0017] The above summary of the invention does not list all of the necessary features of the present invention, and any subcombination of these features may also constitute an invention. Effect of the Invention
[0018] The method for evaluating the support condition of a track slab according to the present invention includes the steps of: obtaining a spectrogram for judgment by performing time-frequency analysis on the time history response waveform of sound pressure; calculating a mean absolute error using the spectrogram for judgment and a mapping and projection generated from a previously prepared learning data set; calculating a threshold value from the learning data set; calculating the degree of abnormality from the ratio of the mean absolute error and the threshold value; and estimating the gap range from the relationship between the degree of abnormality and the gap depth. Therefore, it is possible to reduce the number of impact positions and sound collection positions and rationalize their arrangement, thereby enabling the miniaturization of the measurement device while maintaining the measurement accuracy. [Brief description of the drawings]
[0019] [Figure 1] 1 is a schematic diagram of a slab track for carrying out a method for evaluating the support condition of a track slab according to an embodiment of the present invention; [Diagram 2] FIG. 2 is a flow diagram for creating a learning data set used in the method for evaluating the support condition of a track slab according to an embodiment of the present invention. [Diagram 3] FIG. 2 is a flow diagram of a method for evaluating the support condition of a track slab according to an embodiment of the present invention. [Figure 4] FIG. 2 is a diagram showing an example of a spectrogram used in the method for evaluating the support condition of a track slab according to the present embodiment. [Diagram 5] FIG. 2 is a diagram showing an example of a spectrogram of judgment data used in the method for evaluating the support condition of a track slab according to the present embodiment. [Figure 6] 5 is a graph showing the relationship between gap depth and abnormality degree used in the method for evaluating the support condition of a track slab according to the present embodiment. [Figure 7] FIG. 4 is a diagram showing an example of an abnormality contour used in the method for evaluating the support condition of a track slab according to the present embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0020] Preferred embodiments for carrying out the present invention will be described below with reference to the drawings. Note that the following embodiments do not limit the inventions according to the claims, and all of the combinations of features described in the embodiments are not necessarily essential to the solution of the invention.
[0021] FIG. 1 is a schematic diagram of a slab track for carrying out a method for evaluating the support condition of a track slab according to an embodiment of the present invention, FIG. 2 is a flow diagram for creating a learning dataset for use in the method for evaluating the support condition of a track slab according to an embodiment of the present invention, FIG. 3 is a flow diagram for the method for evaluating the support condition of a track slab according to an embodiment of the present invention, FIG. 4 is a diagram showing an example of a spectrogram for use in the method for evaluating the support condition of a track slab according to the embodiment, FIG. 5 is a diagram showing an example of a spectrogram of judgment data for use in the method for evaluating the support condition of a track slab according to the embodiment, FIG. 6 is a graph showing the relationship between gap depth and anomaly degree for use in the method for evaluating the support condition of a track slab according to the embodiment, and FIG. 7 is a diagram showing an example of anomaly degree contour for use in the method for evaluating the support condition of a track slab according to the embodiment.
[0022] [Slab Rail Configuration] First, an overview of a slab track 1 that is the subject of evaluation in the method for evaluating the support state of a track slab according to this embodiment will be described.
[0023] As shown in Figure 1, slab track 1, which is the subject of evaluation in the method for evaluating the support condition of track slabs according to this embodiment, has a track structure in which rails 5 are fixed to a track slab 4, which is a precast concrete slab, by rail fastening devices 6 arranged at regular intervals. The bottom surface of the track slab 4 is supported by a filling layer 3 formed between the track slab 4 and a concrete trackbed 2, and transmits the train load to the structure. The filling layer 3 is made of cement asphalt mortar made of cement, asphalt emulsion, etc.
[0024] It is known that such slab track 1 develops voids S due to gaps and defects thought to be caused by freezing and thawing after long-term use. Depending on the degree of deterioration of the voids S, repairs may be necessary, so railway operators and others carry out regular inspections. It is known that deterioration of the voids S progresses from the edges of the filled layer 3 toward the center.
[0025] [Embodiment of the method for evaluating the support condition of track slabs] Next, the method for evaluating the support condition of a track slab according to this embodiment will be described with reference to Figs.
[0026] In the method for evaluating the support state of a track slab according to this embodiment, a learning data set is created according to the procedure shown in Fig. 2. To create the learning data set, a learning slab track is prepared in which the track slab is completely supported by the filling layer, i.e., the filling layer has no voids, and a hammering test is performed using the learning slab track to measure the time history response waveform of sound pressure (Step S101).
[0027] At this time, the impact positions and sound collection positions of the impact test are as shown in FIG. 1, an impact position H1 approximately near one edge of the slab track, a sound collection position M1 50 mm from the edge in the width direction, a sound collection position M2 200 mm away, and a sound collection position M3 600 mm away. The same impact positions and sound collection positions are used for the other edge. These are measured at predetermined intervals along the longitudinal direction of the track slab to create a learning data set. In the method for evaluating the support state of a track slab according to this embodiment, the total number of sound collection positions is measured based on the installation interval of the rail fastening devices, and the total number of data is measured so that each slab track is 90 points.
[0028] Next, the time-history response waveform of the sound pressure measured in step S101 is subjected to a time-frequency analysis to create a spectrogram as shown in Fig. 4 (step S102). The spectrogram is three-dimensional image data consisting of three axes of time, frequency, and spectral intensity.
[0029] The spectrogram obtained in step S102 is stored as a learning data set (step S103).
[0030] This is repeated multiple times for each sound collection position, and the spectrograms are stored in the learning data set until a predetermined number of times is reached, thereby creating a learning data set (step S104).
[0031] Next, the learning data set compiled in step S104 is divided into data 11 for mapping generation and data 12 for threshold calculation. The data 11 for mapping generation is compressed and then restored to create a mapping 13 via principal component analysis (PCA), autoencoder (AE), or convolutional autoencoder (CAE), which are algorithms for quality learning. Compression involves reducing explanatory variables of the data that constitutes the spectrogram, and restoration involves returning the reduced explanatory variables to their original number. The data for mapping generation and the data for threshold calculation are divided randomly using pseudo-random numbers according to an arbitrarily set division ratio (for example, 50%).
[0032] Furthermore, the threshold calculation data 12 references the generated mapping 13 to generate a projection 14 of the threshold calculation data onto the mapping 13. Then, the training mean absolute error is calculated using the projection 14 and the original threshold calculation data 12. The training mean absolute error is calculated by averaging the absolute values of pixel differences in the spectrogram over the total number of pixels.
[0033] Next, threshold 15 is calculated from the sample set of training mean absolute errors. If the average of training mean absolute errors is μ and the standard deviation of training mean absolute errors is σ, threshold 15 can be calculated as threshold 15 = μ + 1.96 σ. When calculation of threshold 15 is completed, the flow of creating a training dataset is completed.
[0034] Next, a method for generating judgment data for the method for evaluating a support condition of a track slab according to this embodiment and a method for evaluating a support condition of a track slab using the judgment data will be described in detail with reference to FIG.
[0035] First, in the same manner as in step S101 described above, the time history response waveform of sound pressure is measured for the slab track to be evaluated, the support state of which is unknown (step S201). Note that the impact position and sound collection position are the same as those used to create the learning data set described above.
[0036] Furthermore, similarly to step S102, a time-frequency analysis is performed on the time history response waveform of the sound pressure measured in step S201 to obtain a spectrogram for determination (step S202).
[0037] Next, the decision spectrogram generated in step S202 is compressed and restored via the mapping 13 generated from the learning data set described above to obtain a projection 21 (step S203).
[0038] The mean absolute error is calculated using the projection 21 obtained in step S203 and the spectrogram for judgment obtained in step S202 (step S204). The mean absolute error can be calculated in the same manner as the mean absolute error for training described above.
[0039] The degree of anomaly is calculated from the ratio of the mean absolute error obtained in step S204 to the threshold value 15 calculated from the judgment dataset (step S205). The degree of anomaly (%) can be obtained by dividing the mean absolute error calculated in step S204 by the threshold value 15 and multiplying the result by 100.
[0040] Next, it is determined whether the abnormality degree is greater than 100% (step S206). At this time, if the abnormality degree is greater than 100%, it is determined that there is a gap at the sound collection position, and if it is 100% or less, it is determined that there is no gap.
[0041] The above-mentioned judgment data is collected for each sound collection position, and the void range is estimated from the relationship between the degree of abnormality and the void depth. At this time, the void range is estimated by creating a contour as shown in Figure 7 from the relationship between the degree of abnormality and the void depth (the distance from the side of the filling layer to the boundary between the void range and the support range) as shown in Figure 6.
[0042] In this way, according to the method for evaluating the support condition of a track slab according to this embodiment, by performing evaluation using a spectrogram, it is possible to reduce the number of impact positions and sound collection positions in the impact test without generating erroneous judgments, and to improve the efficiency of their arrangement. Furthermore, by reducing the number of impact positions and sound collection positions, it is possible to reduce the size and weight of the impact sound test device for slab track used in the method for evaluating the support condition of a track slab according to this embodiment, making it easier to evaluate the support condition of a slab track.
[0043] In addition, in the above-mentioned method for evaluating the support condition of a track slab according to the present embodiment, a case where a track slab is placed on a concrete roadbed via a filling layer has been described as a slab track, but the configuration of the slab track is not limited to this, and the present invention can also be applied to a slab track, for example, a configuration in which a track slab is placed on a concrete floor plate of an elevated bridge via a filling layer. It is clear from the claims that such modified or improved forms are also included in the technical scope of the present invention. [Explanation of symbols]
[0044] 1 slab track, 2 concrete track bed, 3 filling layer, 4 track slab, 5 rail, 6 rail fastening device, 11 data for mapping generation, 12 data for threshold calculation, 13 mapping, 14 projection, 15 threshold, H1 impact position, M1~M3 sound collection positions, S gap.
Claims
1. A method for evaluating the support condition of a track slab in a slab track having a filling layer below the track slab, comprising the steps of: striking an arbitrary striking position of the track slab; measuring a time history response waveform of the sound pressure caused by the strike at a plurality of sound collection positions that are a predetermined distance away from the strike position; A step of performing a time-frequency analysis on the time history response waveform of the sound pressure to obtain a spectrogram for judgment; calculating a mean absolute error using the decision spectrogram and a mapping and a projection generated from a training data set prepared in advance; determining a threshold value from the training data set; Calculating an anomaly degree from a ratio of the mean absolute error and the threshold value; A method for evaluating the support condition of a track slab, comprising estimating the range of the gap from the relationship between the degree of abnormality and the gap depth.
2. The method for evaluating the support condition of a track slab according to claim 1, The method for evaluating the support condition of a track slab is characterized in that the learning data set is a collection of learning spectrograms obtained by conducting multiple impact tests on a slab track where the track slab is fully supported by a filling layer, and performing time-frequency analysis on the time history response waveforms of sound pressure obtained in the impact tests.
3. The method for evaluating the support condition of a track slab according to claim 1, The step of obtaining a threshold value from the training data set includes a step of dividing the training data set into mapping generation data and threshold calculation data; a step of compressing and decompressing the mapping generation data to generate a mapping; generating a projection of the threshold calculation data onto the mapping; A method for evaluating the support condition of a track slab, characterized in that the threshold is obtained by averaging the absolute value of the pixel difference in the spectrogram from the projection and the threshold calculation data over the total number of pixels.
4. The method for evaluating the support condition of a track slab according to claim 1, the impact position is an edge portion of the track slab in a width direction, A method for evaluating the support condition of a track slab, characterized in that the sound collection positions are arranged at predetermined intervals widthwise inward from the impact position.
5. The method for evaluating the support condition of a track slab according to claim 1, A method for evaluating the support condition of a track slab, characterized in that the estimation of the gap range is performed by creating a contour of the degree of abnormality in the track slab.
6. The method for evaluating the support condition of a track slab according to claim 1, A method for evaluating the support condition of a track slab, wherein the judgment spectrogram is three-dimensional image data having three axes of time, frequency, and spectral intensity.
Citation Information
Patent Citations
Abnormality detection device, abnormality detection method, and abnormality detection program
JP2020042519A
Classification device, classification system, classification method and program
JP2020107138A
Striking sound testing device for slab track
JP2022080347A
Grout filling status evaluation system and grout filling status evaluation program
JP7216238B1
Device and Method for Assessing a Quality Class of an Object to be Tested
US20070100623A1
