Method for automatically evaluating track measurement data

Abstract

The invention relates to a method for automatically evaluating track geometry and / or wavelet-transformed track measurement data (1) of a ballast bed by means of a computing device. In order to be able to automatically generate suggestions for eliminating track defects, it is proposed that a measurement series of track measurement data (1) to be evaluated, which is associated with a track section, is first subjected to wavelet transformations using a plurality of wavelets of different wavelengths, and a thermal image and a wavelet power density spectrum (3) are formed from the wavelet transformations, according to the invention, a thermal image is created in which wavelengths and wavelet transformations relating to positions in the track are defined as thermal information, and a signal strength map (4) for different wavelength ranges (D0, D1, D2, D3) is additionally calculated, from which local positions (B, C, D, F), ranges ([delta] x) and corresponding wavelength ranges (E, D0, D1, D2, D3) of the track defects occurring are determined from the thermal image, in particular from contours of the thermal image, and the local positions (B, C, D, F), ranges ([delta] x) and the corresponding wavelength ranges (E, D0, D1, D2, D3) of the track defects occurring are determined from the local positions (B, C, D, F) and ranges ([delta] x). Therefore, the type, position, range and size of the track defect can be obtained.

Description

Technical Field

[0001] The invention relates to a method for automatically evaluating wavelet transformed track measurement data of track geometry and / or of a macadam ballast using a computing device. Background Art

[0002] It is known from the prior art to analyze the measurement data obtained by a track measuring vehicle using a wavelet transform. For example, document CN 111979859 A discloses a system for detecting track irregularities, in which the data obtained by an acceleration sensor are analyzed using a wavelet transform. In addition, document CN 104032629B discloses a method for detecting long-wave track irregularities, in which the data obtained from an angular acceleration sensor are analyzed using a wavelet transform. Document CN 104947555 A shows a further method for identifying track irregularities, in which the track irregularities are determined by Fourier transforming the wavelength that is the basis for the irregularity.

[0003] Most railway tracks are designed as gravel ballast tracks. The sleepers lie flat in the gravel ballast. Due to the wheel forces of the trains running on it, the gravel ballast is rounded, partially broken and worn. This leads to irregular sinking in the gravel ballast and displacement of the lateral position geometry of the track. Due to the sinking of the gravel ballast bed, defects in longitudinal height, superelevation (in curves), twisting, gauge and directional position appear. These defects in turn increase the forces, which in turn have a destructive effect on the gravel ballast and the foundation.

[0004] If these geometrical parameters exceed certain limits or safety limits set by the railway authorities, maintenance work is planned and carried out. To eliminate and correct these geometrical track defects, track engineering machines are used. Before track construction, the geometrical track position of the track is acquired and correction values ​​are derived therefrom, which are then transmitted to the machine to carry out the corrective measures.

[0005] For acceptance of the calibration work, the machine is equipped with a tracking measuring system to check whether the rail position complies with the specified tolerances (AT516278A1).

[0006] The tamping machine with fully hydraulic tamping drive (AT513973a1) acquires the characteristics of the ballast during operation with the help of sensors integrated in the fully hydraulic tamping drive (AT515801A1). By analyzing the force path, feed speed, feed path and time sequence, the ballast parameters such as ballast hardness, tamping force, ballast stiffness and ballast damping (AT520117A1) can be measured and given for each tamped sleeper.

[0007] These parameters provide an indication of the degree of contamination of the crushed stone ballast. Above a contamination level of more than 30% of fine components in the ballast, the track position can no longer be permanently corrected by tamping. The ballast must then be replaced or cleaned. For this purpose, a crushed stone cleaning machine is used.

[0008] At significant differences in stiffness in the track (e.g. worn-in rail joints, transitions in bridges or tunnels), the high wheel-rail forces cause the sleepers to hit the gravel bed and break (and round) the gravel ballast. White spots (escaping rock dust) are often visible on the surface of these locations. At these locations, the driving dynamics cause the gravel ballast to be crushed and these locations are indicated by escaping mineral dust. These individual defects are usually only a few meters in range, but tend to spread further and their defect amplitude increases rapidly. This often leads to safety-critical defects that need to be eliminated immediately or, better still, preventively. These defects can be eliminated by a tamping machine operating in single defect elimination mode.

[0009] Due to the different settlements of the gravel ballast bed, there are suspended sleeper positions. This usually occurs at a distance of 2-5 m (D0 zone). These defects can also be eliminated by track tamping machines.

[0010] Defect wavelengths in the range between 3 and 25 m (D1 band) generally represent track defects that arise due to interaction with the vehicle (bogie wheelbase, bogie spacing or single axle spacing, suspension and damping characteristics of the running gear). Settling is caused by redistribution, wear and breakage of crushed stone ballast particles. Track position defects in this wavelength range can be eliminated by track tamping machines. It is known that highly contaminated gravel ballast beds require high compaction forces. There is no room for movement between the gravel ballast particles because the space is filled with fine material. This increases the compaction force that must be applied to move and compact the ballast. At the same time, the durability of the corrected track geometry of such a contaminated roadbed is reduced because the friction and meshing between the gravel ballast particles are smaller.

[0011] Long-wave defects (D2 zone) between 25m and 70m are defects in the roadbed adjacent to the crushed stone ballast. Usually, the cause is poor drainage or insufficient bearing capacity (soaked clay or pottery clay, etc.), which results in long-wave settlement. Drainage obstruction may occur, for example, due to the construction of sound insulation walls, which hinder the drainage of the roadbed. The longer the wavelength of the defect, the more likely its cause is to be found further down the crushed stone ballast layer. Although these defects can be compensated by track tamping machines, the cause of the defect is not permanently eliminated. Permanent elimination can only be achieved by cleaning the track bed or by roadbed repair. For this purpose, a roadbed protection layer is introduced with the help of a roadbed improvement machine (or other non-mechanized methods). If the track defects are caused by poor drainage, the drainage needs to be improved. This can be achieved by digging roadbed ditches or by cleaning and flushing drainage channels.

[0012] Defects with wavelengths greater than 70 m (D3 band) are caused by insufficient bearing capacity of the subgrade. Here, the installation of a load-dispersing subgrade protection layer or soil replacement using a subgrade improvement machine can help. Experience has shown that such defects usually manifest themselves as long-wave distortions.

[0013] Railways divide track lines into different line classes. Line classes are distinguished according to the speed range in which they are operated. Each line class is assigned its own limit values ​​for the standard deviation of track defects. This is done for track geometry parameters such as direction, height, lateral inclination, twist and gauge. There are limits for planning track operations (which should be carried out within a specific time period) and there are critical limits that require immediate correction or restriction of operation (up to blockade).

[0014] On the other hand, measuring the properties of the ballast bed by means of a fully hydraulic tamping drive allows very detailed and specific acquisition of the ballast properties for each sleeper (AT515801B1) and thus permits an objective assessment.

[0015] Currently, track maintenance is planned based on track geometry measurements. Track measuring vehicles travel on the track regularly and acquire the geometric position of the track. In this case, the track position is usually divided into sections of about 200 m in length and the standard deviations of the height position, direction, superelevation and twist are detected. In addition to these statistical values, individual single defects are also measured. If the statistical values ​​exceed certain comfort tolerances, maintenance work is planned and carried out.

[0016] The assessment of track defects is based on the standard deviation or sliding mean of the measurement signal. The determination of the exact location, extent, type and cause of the track defects remains uncertain or ambiguous. The planning and implementation of track work is often based on pre-set rules and regulations or the experience of the responsible person. Since objective measurement data are often not available, the characteristics of the gravel bed and subgrade are usually rarely included in the assessment.

[0017] Typically, a 200 m long segment is evaluated and a track quality index (TQI) is given. This index is usually calculated from a weighted combination of the standard deviations of various track geometry parameters or only from the standard deviation of the vertical height.

[0018] The disadvantage of these methods is that they are not based on an analysis of the causes of track defects, and therefore often inappropriate methods are used for correction. This leads to increased maintenance costs. For example, the wrong method may lead to a rapid increase in the number of tamping operations. The correct method should be to clean the gravel track bed. This not only leads to increased maintenance costs, but also has an adverse effect on the service life of track components (ballast, rails, sleepers, etc.) and increases the life cycle cost LCC. Gravel ballast that has been used for a long time may be seriously damaged. A large proportion of fine components and organic materials or soil squeezed upward from the roadbed may have filled the gaps in the gravel ballast. It is known from practice that the track position in such a gravel ballast structure cannot be permanently corrected with a track tamping machine.

[0019] It is also known from practice that individual defects occur randomly distributed on the track. About 40% of these local defect locations can be permanently eliminated. 60% of these defects develop again within a short time. Tracks with good crushed stone ballast are compacted on average about every four years. Individual defects that indicate that the crushed stone ballast is damaged require repair measures about every 1-3 months. During each tamping process, the tamping tool damages parts of the crushed stone ballast due to the high compaction force. Therefore, long overhaul work cycles are of great economic significance.

[0020] Track locations with high ballast bed hardness (high stiffness) form track high points. The more different the stiffness fluctuations in the subgrade (or track bed), the greater the force interaction between wheels and rails, the higher the track loads and the faster the track geometry deteriorates. Individual short defect locations in the track have a tendency to extend in the longitudinal direction under the action of high dynamic forces in the track, increase the height of the track defect and generate subsequent defects caused by excited rail vehicles.

[0021] The application of artificial intelligence methods is state of the art. The AI ​​models used can be divided into different categories. A distinction is made between Artificial Neural Networks (ANN), Adaptive Neuro-Fuzzy Inference Systems (ANFIS), Decision Support Systems (DSS) and Machine Learning models. AI models are able to map complex track position degradation behaviors or track position defects in terms of type, location, extent and wavelength with high accuracy.

[0022] Artificial intelligence models must be trained using a training data set and then tested using a test data set.

[0023] By carrying out a comparative LCC analysis of different maintenance methods (tamping with increasingly shorter maintenance intervals instead of actually necessary track cleaning), their costs can be compared with one another. In order to compare different maintenance strategies, for example, so-called standard elements or standard kilometres are determined. For this purpose, the expertise of railway engineers or actual derived figures and costs are used. Standard kilometres are divided into categories such as subgrade quality, radius, traffic load, track form and number of tracks. Summary of the invention

[0024] The technical problem to be solved by the present invention is to provide a method for automatically evaluating and analyzing a series of measurements of track data, in particular a series of measurements of track geometry and / or track measurement data of a ballast track. The analysis should automatically determine the type, location, extent and size of track defects. The method should also be able to automatically generate recommendations for eliminating track defects and provide an overall assessment of the track condition.

[0025] The invention achieves this object by means of the features of independent claim 1. Advantageous developments of the invention are described in the dependent claims.

[0026] The present invention is characterized in that a measurement series of track measurement data to be evaluated that is assigned to a track section is first wavelet transformed using wavelets of multiple different wavelengths, and a thermal image and a wavelet power density spectrum are formed from these wavelet transformations, in which the wavelength and the wavelet transformation regarding the position in the track are specified as thermal information, and signal intensity maps for different wavelength ranges are additionally calculated, based on which the local position, range and corresponding wavelength range of the occurring track defects are determined from the thermal image, in particular from the contour lines of the thermal image, in order to derive the type, position, range and size of the track defects.

[0027] For this purpose, the measured series of track measurement data (e.g. longitudinal height, orientation, twist, transverse height, ballast bed hardness, compaction force, ballast stiffness, ballast damping) are analyzed with respect to local position, range and wavelength content by means of wavelet methods and, if necessary, fractal methods. It is important that, unlike the Fourier transformation, the wavelet transformation retains the position information of the defects, so that a clear position in the track can then be assigned to each individual defect. In addition, the wavelet power density spectrum is calculated and a thermal image is generated in which the position, range and size of the track defects can be seen by color or by lines of equal intensity (height).

[0028] Furthermore, a double logarithmic fractal diagram can be generated, from which the track defects can be calculated depending on the wavelength (from the slope and the position in the wavelength range). The type, position and range of the track defects are automatically calculated from the analyzed range of the wavelengths contained in the measurement series. The intensity and size of the track defects are inferred from the integral of the wavelet power density spectrum over the corresponding wavelength band. All this is used to automatically evaluate the measured measurement series of track geometry parameters and to automatically determine the type, position, range and size of the track defects. Recommendations for eliminating track defects are automatically generated from the analysis. The analysis also enables an overall assessment of the state of the observed track.

[0029] According to the invention, an expert system is provided which automatically assigns the type, location and extent of track defects using objective figures. In addition to the track position measurements, however, information about the ballast can also be used for the evaluation. Any mutual influences between the different measured variables may not be captured by the expert system.

[0030] Therefore, an artificial intelligence should be trained with this expert system, which integrates these hidden relationships into the evaluation and takes them into account, thereby improving accuracy.

[0031] Wavelets originate from the idea of ​​dividing the line into shorter sections and using Fourier transform to find where track defects occur in the short waves (short-time Fourier transform).

[0032] Unlike the sine and cosine functions of Fourier transform, wavelets are localized in wavelength and position spectra. In short, wavelet transform is like filtering the signal segment by segment with a bandpass filter of a specific bandwidth. Therefore, a locally restricted and specific defect wavelength range is found. A two-dimensional representation of wavelength variation with position is generated from a simple track defect signal. Different wavelet functions can be used. Typical ones are Morelett wavelet and Mexican hat wavelet.

[0033] The Mexican Hat wavelet is mathematically expressed as follows:

[0034]

[0035] The wavelet transform is calculated as:

[0036]

[0037] And is called the wavelet transform of F(x) with respect to ψ.

[0038] The wavelet is shifted through the function F(x) (e.g. the curve of the measured ballast bed hardness in the longitudinal direction of the track) by b as the displacement factor and the wavelength of the wavelet is changed by a (scaling factor) as the wavelength parameter. This results in a two-dimensional representation (the detected wavelength is plotted at position b in the track).

[0039] By applying the fractal theory, the so-called fractal number can be calculated from the track measurement data curve of the track. To this end, the length of the polygonal conductor that fits the measurement data curve is calculated for a specific section length of the railway line, for example 200 m, and the step size is continuously reduced. The length of the polygonal conductor applies:

[0040]

[0041] L(d) = polygon length; d = polygon step size, and Dr = fractal dimension.

[0042] If we take the logarithm of this equation, we get

[0043] logL(d)=(1-D r )·log(d)+log(n)

[0044] In double logarithmic representation, the regression line is calculated (piecewise). The slope is always negative (the finer the segmentation, the greater the polygon wire length) and satisfies

[0045] k=1-Dr

[0046] k... fractal number

[0047] The study showed that the different slopes of the regression lines can correspond to wavelength ranges and their causes.

[0048] According to European standards, typical wavelength ranges are divided into 4 categories. Track defect causes can be assigned to these wavelength ranges based on practical experience and measurements.

[0049] symbol Wavelength range(m) describe Attribution D0 0.5<λ≤3(5) shortwave Interaction between sleepers and crushed stone ballast D1 3<λ≤25 medium wave Crushed stone ballast pollution D2 25<λ≤70 Long Wave Mixed Zone and Roadbed Issues D3 70<λ≤150 big wave Roadbed Problems

[0050] Table 1: Correspondence between wavelength range and orbital characteristics

[0051] The table shows the classification of wavelength ranges and the corresponding causes of track irregularities.

[0052] At present, electronic detection and measurement runs usually do not capture and evaluate the wavelength range D0. This wavelength range is mainly caused by the suspended position of the sleeper and the reaction between the sleeper and the rail fastening device. The sleepers that hit the crushed stone ballast are usually formed at 1.2m and 3 to 3.6m (i.e. 2 times and 5-6 times the usual sleeper spacing of 0.6m).

[0053] The D1 range is a typical range in which quasi-periodic track defects are generated by car body and bogie movements. Dynamic loads acting on the rails lead to degradation of the crushed stone ballast. The result is: wear of the crushed stone ballast and crushing of the crushed stone ballast particles. Tamping or cleaning of the crushed stone ballast is performed as a maintenance measure and damaged ballast is replaced with new crushed stone ballast.

[0054] D2 occurs in the ballast-subgrade mixed area and in the subgrade. This area can also be improved by tamping and track cleaning.

[0055] The D3 range can be attributed to subgrade problems, long-wave fluctuations are often characterized by twisting fluctuations. The D3 range is characterized by insufficient load-bearing capacity. Possible causes are poor drainage, poor subgrade soil (clay, clay, peat), unsuitable subgrade materials or missing or too weak subgrade protection. This track defect can be eliminated by eliminating drainage problems, improving and repairing the subgrade or replacing the soil.

[0056] According to the invention, wavelet and fractal analysis is applied to the recorded measurement series. The expert system thus created is used to train an artificial intelligence.

[0057] The AI ​​model automatically provides information on the location, extent and type of track defects. In addition, the AI ​​provides information on the overall quality state of the track and recommends the best maintenance method based on technical and economic calculations. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Embodiments of the present invention are schematically shown in the accompanying drawings. In the accompanying drawings:

[0059] Figure 1 The Mexican Hat wavelet is shown;

[0060] Figure 2 Fractal diagram showing the spectrum of track defects before and after track bed cleaning

[0061] Figure 3 Shows the evaluation of a series of rail defect measurements by wavelet analysis DETAILED DESCRIPTION

[0062] Figure 1 The shape of a wavelet, a so-called "Mexican hat", is shown by way of example. The wavelet is moved through the measurement series for evaluation. Identical wavelength components of the signal are matched and a corresponding signal is generated. Since the evaluation takes place as different wavelengths are passed through successively, a two-dimensional diagram is produced.

[0063] Figure 2The results of the fractal analysis of a track section before and after track cleaning are shown. The effects can be clearly seen in the medium-wave ranges 5 and 6 (2-15m), while in the long-wave range the track section is practically unaffected by the gravel ballast cleaning. The flat slope in the long-wave range 7 shows that the foundation has sufficient load-bearing capacity and meets the requirements. Improvements occur due to track cleaning. However, track cleaning has no effect on the long-wave range. If the change in the fractal number as a function of track load or running time is concerned, conclusions can be drawn about the remaining service life or aging rate of the gravel ballast. Likewise, a larger slope in the long-wave range indicates foundation problems. The characteristic is that the fractal analysis can be performed for any track length, for example also for the entire line or the entire track network. The values ​​provided by the fractal analysis are independent of the length of the analyzed pattern.

[0064] Contaminated crushed stone ballast remains contaminated even after tamping, just as the ground conditions do not change. The individual straight lines 5, 6, 7 represent defects in the corresponding wavelength range. The steeper the straight lines 5, 6, the greater the impact of the defect. Fractal analysis is used as a second independent method to determine the defect wavelength band. Although the fractal analysis provides the type and intensity of the defect, it does not specify the location. The fractal analysis can only determine information about the evaluated section and state that track defects with this defect wavelength component are preferentially present in this section.

[0065] Figure 3 The evaluation of the measurement signals using wavelet analysis is shown. The measurement signal 1 (upper area) can be a geometric measurement variable (longitudinal height, direction, twist, gauge, lateral height) or a physical measurement variable (track temperature, gravel bed hardness, gravel bed stiffness, gravel ballast damping or compaction force at the end of tamping). For example, in the figure you can see data for the TQI (Track Quality Index), which has been calculated, for example, from the weighted standard deviation of the track geometry and physical measurement variables. The larger the TQI, the worse the track quality. The value of the TQI can also be given for the track grade. A high-speed track on which vehicles are driven at 300 km / h has narrower tolerances than a freight car track with a maximum speed of 80 km / h.

[0066] The two-dimensional thermal image 2 calculated using wavelet (Morelet) is shown below the signal curve. If the thermal image is shown in color, the position and range of the defect intensity can be clearly seen. The wavelength is drawn in the vertical direction, and the position in the track is drawn in the horizontal direction. For example, a section of 400m is evaluated separately so that the defect wavelength up to 200m can still be analyzed. The evaluation in the image shown takes into account the defect wavelength up to 150m and therefore covers all four relevant wavelength bands from D0 to D3. The intensity of the defect is shown in the thermal image as a contour or in the form of color grading. For example, the defect J in area G of signal 1 is correspondingly shown in the thermal image 2 in the area of ​​the D1 band with a wavelength of about 20m. This is usually a defect caused by the state of crushed stone ballast, which can be eliminated by tamping. The defect is located in the area between 250m and 350m. Since defects also occur around 220m, tamping from 200 to 350m is the best choice. In the D2 and D3 bands, it can be seen that there are drainage problems and load-bearing capacity problems in this area. This can be seen as the actual cause of the track defects that occur in this area. The defect intensity is shown in three wavelength bands D0 to D2 in the figure below. These three wavelength bands make it easy to assign the position, range and defect intensity. The wavelet power density spectrum LD is shown on the right side of the thermal image. The wavelength is plotted vertically and the power density is plotted horizontally. The wavelength bands are plotted in the figure. In order to determine the average power density of the wavelength band, an integral is calculated (marked A). This is done for all four bands. It is also important to detect the maximum value, since the maximum value represents the main track defect wavelength. With the help of this method, individual defects (for example with range x in the image) can also be detected and their position and range can be described.

[0067] The defect type can be assigned based on the above table 1 and the optimal repair can be assigned on this basis. A comparative LCC analysis can explain why, for example, tamping (as in the present case) is more advantageous than track cleaning in this short area.

[0068] Marker B shows, for example, a short-wavelength track defect in the 40m region, which can be attributed to the suspended sleeper position. At the same time, the image in the PSD spectrum shows that the intensity is low and therefore tamping is not yet required in this area. For wavelengths around 35m, marker F finds a weakness in the boundary layer. It is recommended to check this section for effective drainage.

Claims

1. A method for automatically evaluating wavelet-transformed track measurement data (1) of track geometry and / or ballast bed using a computing device, It is characterized in that First, a measurement series of track measurement data (1) to be evaluated and assigned to a track section is wavelet transformed using wavelets of multiple different wavelengths, and a thermal image and a wavelet power density spectrum (3) are formed from these wavelet transformations, in which the wavelength and the wavelet transformation regarding the position in the track are specified as thermal information, and signal intensity maps (4) for different wavelength ranges (D0, D1, D2, D3) are additionally calculated, based on which the local position (B, C, D, F), range and corresponding wavelength range (E, D0, D1, D2, D3) of the occurring track defect are determined from the thermal image, in particular from the contour lines of the thermal image, in order to derive the type, position, range and size of the track defect.

2. The method according to claim 1, It is characterized in that In the wavelet power density spectrum (3), an integral (A) of the power density over the wavelength, in particular over a wavelength range (D0-D3), is formed in order to determine the intensity of the track defect.

3. The method according to claim 1 or 2, It is characterized in that The determined orbital defects are assigned to wavelength ranges (D0-D3).

4. The method according to claim 2 or 3, It is characterized in that The intensity of the track defect (signal intensity, A) is also determined based on the local extent and location of the thermal information in the thermal images (2, C, D, E).

5. The method according to any one of claims 1 to 4, It is characterized in that The location and size (D, Δx) of individual defects are determined from the thermal images.

6. The method according to any one of claims 1 to 5, It is characterized in that A measurement series of track measurement data (1) is analyzed with respect to its wavelength content using a fractal analysis method (FIG. 2), wherein track defects are classified in wavelength ranges (D0, D1, D2, D3) depending on the scope of the wavelength analysis, wherein the size and intensity of the track defects are determined based on the slope of the fractal lines (5, 6, 7) and the state of the track (TQI) is calculated.

7. The method according to any one of claims 1 to 6, It is characterized in that The expert system thus formed is used to train an artificial intelligence which automatically learns to determine track defects from the provided data (1, 2, 3, 4, TQI).

8. The method according to claims 1 to 7, It is characterized in that As a result of the evaluation, recommendations for eliminating track defects are generated.

9. The method according to claims 1 to 8, It is characterized in that The assessment results and recommendations for eliminating track defects are automatically summarized in a report (Figure 3).

10. The method according to claims 1 to 9, It is characterized in that The expected durability of the tamping operation is calculated and estimated from the evaluation results.

Images