EP2269887B1 - Method for determining short wave rail position geometry and rail deflection under load - Google Patents

Abstract

The method involves completely compensating phase distortion and reverse signal sequence by filtering with seven filters. Characteristic filter critical frequency and filter slew rate are determined from measured local equidistant vertical rail position signals. The long-wave and/or short-wave unit of the signals is determined depending on selection of the filters such that another associated short-wave and/or long-wave unit is determined by dual subtraction of doubly filtered rail position signal by an original rail position.

Description

The invention relates to a method for determining short-wave track geometry and the rail subsidence under load, wherein the short-wave geometric track position of a rail vehicle to detect and separate from the influence of the wheelset load.

Such a method is from the DE 26 17 192 A1 known.

The maintenance of the tracks is based on geometrical sizes. Special inspection vehicles are used to inspect the deviation of the track from its ideal position. Because of the measurement techniques used, the current in-service inspection vehicles can only detect long-wave fault shapes under load with the required accuracy.

When inspecting points, rail extensions and rail weld joints, however, short-wave geometric deviations are decisive. Therefore, these track components are checked manually without load. Usually rulers, gauges, etc. are used here.

For the detection of short-wave geometric deviations from the vehicle optical measuring methods are known. Furthermore, measuring systems are also known which detect the short-wave track geometry indirectly via measured wheel-set bearing accelerations and evaluate them directly or convert the wheel-set bearing accelerations into a vertical track-position deviation.

A device is known for measuring wavy deformations of the rail rolling surface of a railroad track, selecting a wavelength from a selected range of wavelengths, measuring by means of a trolley and sensors, processing the signals, determining the depth of the deformation and setting a specific ratio ( DE 30 08 440 C2 ).

From the DE 39 13 159 A1 a method and a device for measuring wavy deformations on at least one rail top (rail running surface) of a rail path is known, wherein when passing through a measuring section with the help of two sensors and an electronic evaluation unit at all occurring wavelengths and wave configurations reaches a very accurate determination of the waviness of the rail surfaces becomes. The generation of the height difference representing output signals takes place on light optical and / or electron optical and / or electronic ways.

The DE 34 44 723 C2 describes a device for detecting the track geometry with a laser, which is arranged on a rail vehicle.

Disadvantages of the prior art:

The manual check is associated with a considerable expenditure of time and personnel. In addition, in the case of manual inspection, safety precautions for occupational safety and protection of the railway operation must be taken such as, for example, Assurance items or line closures.

Optical measuring systems in rail vehicles require a high level of maintenance of the measuring technology (contamination) and are therefore not suitable for automated use.

Indirect assessment of short-wave track geometry based on measured accelerations depends on several boundary conditions, such as: Vehicle speed, error amplitude and error length, so that no clear conclusion on the short-wave track geometry is possible.

If measured accelerations are converted into a dimensionally stable track geometry, then, as with all measuring systems on rail vehicles, this includes the influence of the axle load.

To overcome the disadvantages of the prior art, the invention has for its object to detect the short-wave geometric track position of a rail vehicle and to separate the influence of the wheelset load. The proportion without the influence of the wheelset load is comparable to the manual inspection. The remaining proportion corresponds to the long-wave track deviations, which result from long-wave track position disturbances and the rail depression under the influence of the wheelset load.

This object is achieved by a method having the features of claim 1.

For this purpose, two evaluation methods are described, which are applied to locally equidistant, true to form vertical track position signals in order to divide these into short-wave and long-wave components.

For the division according to method 1, filters can be used which filter the locally equidistant true vertical track position signal in the local area. The independent, separate design of the two filters which provide the long-wave component and the short-wave component are not suitable, since these filters in addition to different amplitude responses also have different phase responses and thus lost in the separation of the signal components essential information.

In the method 1 according to the invention therefore two matched filters are used, in which the high-pass filter is developed with known filter design techniques. The filter cutoff frequencies in the local area as well as the filter steepness are determined from measured track position signals which contain long-wave and short-wave track position deviations. This is done for example by means of spectral analyzes in the local frequency range.

In the method 1 according to the invention, an FIR high-pass filter (HP filter) with the determined cutoff frequency and filter slope with a linear phase response is designed in the first step. The filter design provides the filter coefficients b _{k, HP} .

The application of the FIR high-pass filter to the locally equidistant true vertical track position signal z [n] yields the short-wave component z _KW [n].

FIR filters: $$z_{\mathit{KW}}\left[n\right]={\displaystyle \underset{k=-N}{\overset{N}{\Sigma }}{b}_{k.\mathit{HP}}}\cdot z\left[n-k\right]$$

For linear phase-locked FIR filters, the filter coefficients bk _{, TP of} the corresponding low-pass filter can be determined directly. $$b_{k.\mathit{TP}}=b_{k.\mathit{AP}}-b_{k.\mathit{HP}}$$

Wherein the filter coefficients b _{k, AP of} the all-pass filter are defined as follows: $$b_{k.\mathit{AP}}=\{\begin{array}{c}1;k=0\\ 0;k\ne 0\end{array}$$

The application of the low-pass filter to the locally equidistant true vertical track position signal z [n] yields the long-wave component z _LW [n]. $$z_{\mathit{LW}}\left[n\right]={\displaystyle \underset{k=-N}{\overset{N}{\Sigma }}{b}_{k.\mathit{TP}}}\cdot z\left[n-k\right]$$

Since in this case the order of the high-pass filter is equal to the order of the low-pass filter and both filters have linear phase response, the same group delay results for both filters. The stationary equidistant vertical track position signal z [n] was taken without loss of information into the two components z _KW and z _LW . $$z={\mathit{z}}_{\mathit{KW}}+{\mathit{z}}_{\mathit{LW}}$$

In the event that a FIR high-pass nonlinear-phase filter or an IIR high-pass filter was used in the first step, method 2 involves the high-pass-filtering of the locally equidistant true-to-life vertical track signal z [n].

The application of the high-pass filter to the locally equidistant true vertical track position signal z [n] yields the short-wave component z _KW [n].

For FIR filters: $$z_{\mathit{KW}}\left[n\right]={\displaystyle \underset{k=-N}{\overset{N}{\Sigma }}{b}_{k.\mathit{HP}}}\cdot z\left[n-k\right]$$

For IIR filters: $$z_{\mathit{KW}}\left[n\right]={\displaystyle \underset{k=0}{\overset{N}{\Sigma }}{b}_{k.\mathit{HP}}}\cdot z\left[n-k\right]-{\displaystyle \underset{k=0}{\overset{M}{\Sigma }}{a}_{k.\mathit{HP}}}\cdot z\left[n-k\right]$$

The long-wave component z _LW [n] of z [n] is obtained by subtracting z _{KW [} n] from z [n]. $${\mathit{z}}_{\mathit{LW}}\left[n\right]=z\left[n\right]-{\mathit{z}}_{\mathit{KW}}\left[n\right]$$

In order for the distribution of the locally equidistant vertical track position signal z [n] to take place without loss of information, the short-wave component z _KW [n] must not exhibit any phase distortion compared to the signal z [n]. This is achieved by applying the high pass filter twice to the signal z [n], with the second order reversing the signal order.

Due to the subtraction, the locally equidistant true vertical track position signal z [n] was taken into the two components z _KW and z _LW without any loss of information. $$z={\mathit{z}}_{\mathit{KW}}+{\mathit{z}}_{\mathit{LW}}$$

Embodiment:

Reference to an embodiment, the application of the described method will be explained in more detail below.

• Figur 1: einen Signalflussgrafen eines Hochpassfilters mit z als Eingangssignal und z_KW als Ausgangssignal
• Figur 2: die Amplitudengänge der HP-, und TP-Filter
• Figur 3: den Amplitudengang eines Allpass-Filters
• Figur 4: einen Signalflussgrafen für die Aufteilung der Gleislagegeometrie in kurzwellige und langwellige Anteile wobei FIR-Filter mit gleicher Filterordnung und linearen Phasengängen verwendet werden
• Figur 5: einen Signalflussgraph der Subtraktion z_KW von z
• Figur 6: den Amplituden- und Phasengang des Butterworth-Hochpassfilters
• Figur 7: die vertikalen Gleislageabweichungen z[n], z_KW[n] und z_LW[n] eines Weichenherzstückes im Ortsbereich
• Figur 8: den Vergleich einer manuellen Messung an einem Weichenherzstück mit dem ermittelten kurzwelligen Anteil z_KW[n] einer Messung aus einem Regelzug

Showing:

• FIG. 1 : a signal flow graph of a high-pass filter with z as the input signal and z _KW as the output signal
• FIG. 2 : the amplitude responses of HP, and TP filters
• FIG. 3 : the amplitude response of an all-pass filter
• FIG. 4 : A Signalflussgrafen for splitting the track geometry in short-wave and long-wave proportions where FIR filters are used with the same filter order and linear phase responses
• FIG. 5 : a signal flow graph of the subtraction z _KW of z
• FIG. 6 : the amplitude and phase response of the Butterworth high pass filter
• FIG. 7 : the vertical track deviations z [n], z _KW [n] and z _LW [n] of a switch frog in the local area
• FIG. 8 : the comparison of a manual measurement on a switch frog with the calculated short-wave component z _KW [n] of a measurement from a control train

The input is a stationary equidistant vertical track position signal z [n], which was calculated from measured wheel set bearing accelerations (procedure according to the application for industrial property rights) DE 10 2008 062 143.9 ). The equidistant local increment is 0.2 m. The axle bearing accelerations were measured under a load of approx. 7 t. The signal shows the vertical geometry of a switch frog under load. This signal is divided into the short-wave component, the geometry of the core corresponding proportion, and the long-wave component, the rail depression under load (= size of the hollow layer and long-wave track deviation under the Weichenhoulder).

• b_0,HP = 0.490231200703569
• b_1,HP = -2.941387204221414
• b_2,HP = 7.353468010553535
• b_3,HP = -9.804624014071381
• b_4,HP = 7.353468010553535
• b_5,HP = -2.941387204221414
• b_6,HP = 0.490231200703569

• a_0,HP = 1.000000000000000
• a_1,HP = -4.585886346766142
• a_2,HP = 8.895035945278623
• a_3,HP = -9.319068686462629
• a_4,HP = 5.552670292723525
• a_5,HP = -1.781808943261143
• a_6,HP = 0.240326630536351

In the first step, a high-pass filter is designed. In the application example, a sixth-order Butterworth filter is used. The 3dB corner frequency is 1 / 3m, so that wavelengths larger than 3 m are filtered out of the signal z [n]. This is a nonlinear phase IIR filter (Method 2). The filter coefficients b _{k, HP} and a _{k, HP} were determined as follows.

• b _{0, HP} = 0.490231200703569
• b _{1, HP} = -2.941387204221414
• b _{2, HP} = 7.353468010553535
• b _{3, HP} = -9.804624014071381
• b _{4, HP} = 7.353468010553535
• b _{5, HP} = -2.941387204221414
• b _{6, HP} = 0.490231200703569

• a _{0, HP} = 1.000000000000000
• a _{1, HP} = -4.585886346766142
• a _{2, HP} = 8.895035945278623
• a _{3, HP} = -9.319068686462629
• a _{4, HP} = 5.552670292723525
• a _{5, HP} = -1.781808943261143
• a _{6, HP} = 0.240326630536351

FIG. 6 shows the amplitude and phase response of the Butterworth high pass filter.

The locally equidistant dimensionally accurate vertical track position signal z [n] is filtered twice with the HP filter, with the second filters reversing the signal sequence and thus the phase distortion of the HP filter is completely compensated. This high-pass-filtered signal is the short-wave track geometry in the center section without load.

Since the Butterworth high-pass filter used in the first step has a non-linear phase response, the long-wave component z _LW [n] of z [n] is determined by subtraction. For this purpose, the short-wave component z _KW [n] is subtracted from the locally equidistant dimensionally accurate vertical track position signal z [n]. The result is the long-wave component z _LW [n]. This proportion corresponds to the rail depression under the influence of the wheelset load and corresponds to the size of the hollow layer and the long-wave track deviation under the switch frog.

• im oberen Teil: das ortsäquidistante formtreue vertikale Gleislagesignal z[n]
• im mittleren Teil: der kurzwellige Anteil z_KW[n] von z[n]
• im unteren Teil: der langwellige Anteil z_LW[n] von z[n]

In FIG. 7 are shown in the local area:

• in the upper part: the stationary equidistant vertical track position signal z [n]
• in the middle part: the short-wave component z _KW [n] of z [n]
• in the lower part: the long-wave component z _LW [n] of z [n]

In FIG. 8 the comparison of a manual measurement on a switch frog with the determined short-wave component of a measurement from a rule train is compared.

Claims (6)

1. A method for the distortion-free separation of true-to-form track position geometry into short-wave and long-wave fractions without information loss,
   characterized in that
   a locally equidistant, true-to-form, vertical track position signal is separated into a short-wave fraction corresponding to the geometry without a load and a long-wave fraction corresponding to the subsidence of the track under a load with long-wave track position deviations, wherein filters are used in the local range for the distortion-free separation and a filter is designed with known filter design techniques in the first step, wherein the characteristic filter limit frequency and the filter steepness are determined beforehand from the measured track position signals that contain long-wave and short-wave track position deviations, wherein this is preferably carried out by means of spectral analyses in the spatial frequency range and the separation of the locally equidistant, true-to-form, vertical track position signal takes place in dependence on the chosen filter type, as well as the phase response of the filter, wherein the second filter has, when using a FIR-filter with linear phase response, also a linear phase response and the same filter arrangement as the first filter and is calculated from the filter coefficients of the first filter, and wherein the locally equidistant, true-to-form, vertical track position signal is separated into a short-wave and a long-wave fraction due to the utilization of the two filters
   or,
   when using a FIR-filter with nonlinear phase response or an IIR-filter for the distortion-free separation of the locally equidistant, true-to-form, vertical track position signal into a short-wave and a long-wave fraction, the track position signal is filtered and a phase distortion is completely compensated by means of a second filtering with the same filter and the reverse signal sequence, wherein the characteristic filter limit frequency and the filter steepness are determined beforehand from the measured track position signals that contain long-wave and short-wave track position deviations, wherein this is preferably carried out by means of spectral analyses in the spatial frequency range and the long-wave or the short-wave fraction of the track position signal is respectively determined in dependence on the choice of filter, and wherein the associated second short-wave or long-wave fraction is respectively determined by subtracting the twice-filtered track position signal from the original track position.
2. The method according to Claim 1, characterized in that the short-wave fraction of the signal corresponds to the crossing geometry of switches, feathered joints or insulated joints and is evaluated.
3. The method according to Claim 1, characterized in that the long-wave fraction of the signal corresponds to the rail subsidence under a load and the long-wave track position deviation and is evaluated.
4. The method according to Claims 1 to 3, characterized in that the locally equidistant, true-to-form, vertical track position signals are determined by means of measurements carried out by track position inspection vehicles.
5. The method according to Claims 1 to 4, characterized in that the locally equidistant, true-to-form, vertical track position signals are determined by means of measurements carried out by vehicles during normal operation.
6. The method according to Claims 1 to 5, characterized in that the locally equidistant, true-to-form, vertical track position signals are determined by means of manual measurements.

Images