Railway track condition monitoring system for detecting a partial or complete disruption of a rail of the railway track

Abstract

A railway track condition monitoring system including a plurality of sensors installed on a rail of the track and spaced by a predetermined distance from each other. The sensors having respective signal acquisition and recording means for the acquisition and recording of an acoustic signal being generated by an approaching train and propagated by the rail or by a ground carrying the rail. A signal transfer means for the transfer of an acquired and recorded signal to the passing train by wireless transfer at the time of passing of the train in proximity to sensors, and a signal analyzing means for analysing transferred ones of the acquired and recorded signal on-board the passing train. The signal analyzing means may be adapted to compare in a processing unit the signals from adjacent sensors and on that basis indicate a rail discontinuity.

Claims

1. A railway track condition monitoring system for detecting a partial or complete disruption of a rail of the railway track, comprising: a plurality of sensors installed on a rail of the track and spaced by a predetermined distance from each other, the sensors having respective signal acquisition and recording means for the acquisition and recording of an acoustic signal being generated by an approaching train and propagated by the rail or by a ground carrying the rail; a signal transfer means for the transfer of an acquired and recorded signal to the passing train by wireless transfer at the time of passing of the train in proximity to sensors; a signal analyzing means for analysing transferred ones of the acquired and recorded signal on-board the passing train; and means adapted to cross-correlate the signals from a first sensor (s.sub.n) and an adjacent second sensor (s.sub.n+m), where the first sensor and the second sensor are adjacent to one another on the same rail, and detect the partial or complete disruption of said rail of the railway track.

2. The rail condition monitoring system of claim 1, wherein the installed sensors are adapted to be stimulated by an acoustic stimulation passively generated by the approaching train collecting the data.

3. The rail condition monitoring system of claim 1, wherein the installed sensors comprise means for recording the signal for a given programmed time period upon the signal exceeding a predefined noise level.

4. The rail condition monitoring system of claim 1, wherein the signal acquisition and recording means is adapted to identify a start of the signal and to record the start of the signal, and the signal analyzing means is adapted to analyse a recorded start of the signal to derive therefrom rail breakage information indicating a rail discontinuity.

5. The rail condition monitoring system according to claim 1, further comprising: means adapted to distinguish cross-correlation coefficients obtained into a) “exact” correlation for Corr. value=1, b) “strong” correlation for Corr. value>0.8, c) “medium” correlation for Corr. value=0.5-0.8, d) “weak” correlation for Corr. value<0.5, and e) “no” correlation for Corr. value=0; and means responsive to said distinguished cross-correlation coefficients and adapted so to not send an alert message in response to “exact” or “strong” correlation, to perform double-check with neighboring sensors before an alert message is sent in response to “medium” correlation, and to send an alert message in response to “weak” or “no” correlation.

6. The rail condition monitoring system of claim 2, wherein the installed sensors comprise means for recording the signal for a given programmed time period upon the signal exceeding predefined noise level.

7. The rail condition monitoring system of claim 2, wherein signal acquisition and recording means is adapted to identify a start of the signal and to record the start of the signal, and the signal analyzing means is adapted to analyse a recorded start of the signal to derive therefrom rail breakage information indicating a rail discontinuity.

8. The rail condition monitoring system according to claim 2, comprising: means adapted to distinguish cross-correlation coefficients obtained into a) “exact” correlation for Corr. value=1, b) “strong” correlation for Corr. value>0.8, c) “medium” correlation for Corr. value=0.5-0.8, d) “weak” correlation for Corr. value<0.5, and e) “no” correlation for Corr. value=0; and means responsive to said distinguished cross-correlation coefficients and adapted so to not send an alert message in response to “exact” or “strong” correlation, to perform double-check with neighboring sensors before an alert message is sent in response to “medium” correlation, and to send an alert message in response to “weak” or “no” correlation.

9. The rail condition monitoring system according to claim 3, comprising: means adapted to distinguish cross-correlation coefficients obtained into a) “exact” correlation for Corr. value=1, b) “strong” correlation for Corr. value>0.8, c) “medium” correlation for Corr. value=0.5-0.8, d) “weak” correlation for Corr. value<0.5, and e) “no” correlation for Corr. value=0; and means responsive to said distinguished cross-correlation coefficients and adapted so to not send an alert message in response to “exact” or “strong” correlation, to perform double-check with neighboring sensors before an alert message is sent in response to “medium” correlation, and to send an alert message in response to “weak” or “no” correlation.

10. The rail condition monitoring system according to claim 4, comprising: means adapted to distinguish cross-correlation coefficients obtained into a) “exact” correlation for Corr. value=1, b) “strong” correlation for Corr. value>0.8, c) “medium” correlation for Corr. value=0.5-0.8, d) “weak” correlation for Corr. value<0.5, and e) “no” correlation for Corr. value=0; and means responsive to said distinguished cross-correlation coefficients and adapted so to not send an alert message in response to “exact” or “strong” correlation, to perform double-check with neighboring sensors before an alert message is sent in response to “medium” correlation, and to send an alert message in response to “weak” or “no” correlation.

Description

DETAILED DESCRIPTION OF THE INVENTION

(1) In the following, a method for triggering recording an acceleration or vibration signal detected by the acceleration or vibration sensor is described, which method is also referred to herein as a signal detection algorithm for controlling when shall the recording of data be triggered.

(2) Once the train-induced accelerations or vibrations in the rail have arrived at the sensor unit and are sensed by the acceleration or vibration sensor, a portion of the sensed signal that is considered characteristic of a useful signal is recorded, preferably by data representing the sensed signal stored into a separate data file in a data storage. The data representing the sensed signal stored into the separate data file later establishes the basis for the correlation with data from neighboring sensor units. The decision to initiate a recording, herein also referred to as a signal data recording trigger, is made on basis of one or more of the following: a) Using a measured signal-to-noise ratio (SNR) within a moving short time window: The average amplitude level of all data points in a moving time window are continuously monitored. At times when no train-induced waveforms arrive at the sensor, the sensed signal output from the acceleration or vibration sensor of the sensor unit is considered a noise signal. A noise signal detector determines and stores the average amplitude level of the noise signal. As the train-induced waveforms arrive at the sensor, the average amplitude level of the sensed accelerations or vibrations increases and rises above the noise level, as illustrated in FIG. 7. A signal-to-noise ratio is computed as the ratio of the average amplitude level of the sensed accelerations or vibrations to the average amplitude level of the noise signal. A certain signal-to-noise ratio is set as threshold, and the recording of the sensed signal will be initiated (triggered) as soon as the set signal-to-noise ratio threshold is exceeded by the train-induced vibrations. The time window used is not too short in order to avoid false recordings triggered by perturbations close to the respective sensor units (e.g. induced by traffic or other ambient noise). The trigger threshold using the signal-to-noise ratio can be set as a fixed threshold value based upon the noise level, i.e., trigger threshold (SNR)=3×noise level. b) Short-time average/long-time average (STA/LTA) trigger: The average amplitudes of all data points sensed in a short and a long time window, called short-time average (STA) and long-time average (LTA), respectively, are monitored and compared with each other; once the train-induced wavetrains arrive at the sensor unit, the STA will experience a sudden increase while the LTA will only be affected in minor way; the deviation between STA and LTA (sudden increase of the STA/LTA ratio) is used to trigger the recording of the sensed signal. The trigger threshold using the STA/LTA ratio can be set as: trigger threshold (STA/LTA)>1.2×(STA/LTA).sup.noise; i.e., 20% increase of the previous (STA/LTA) for noise. c) Kurtosis of the transient signal: The shape (tailedness) of the envelope of the steadily increasing sensed transient signal upon arrival of a train-induced wavetrain and the thereby increasing amplitudes of the sensed transient signal, is identified and applied to trigger the recording of the sensed signal. d) Autocorrelation of the transient/stationary signal: The single trace data sensed at each sensor is constantly auto-correlated (i.e., cross-correlated with itself at different points in time). In case of random noise and no train-induced wavetrains arriving at the sensor unit, the maximum auto-correlation coefficient (autocorr=1) only occurs at zero-lag, else it is zero. Once the train-induced wavetrains arrive at the sensor unit, during the transient part of the signal which is characterized by steadily increasing amplitudes (see FIG. 7A), an increasing number of auto-correlation spikes will occur away from the zero-lag. When the signal becomes stationary, auto-correlating spikes of (more or less) constant amplitudes will emerge periodically (depending on the dominant period of the stationary sensed signal). e) Amplitude variations of the stationary signal lower than a fixed threshold: The average maximum amplitudes of successive data points within a certain time window are monitored and compared with each other; once the train-induced wavetrains arrive at the sensor unit, the variation of the maximum amplitudes (peaks) of the stationary part of the sensed signal (compare with FIG. 7) becomes minimal, which will trigger the recording of the sensed signal. The trigger using the amplitude variations of the stationary part of the sensed signal can in an embodiment of the invention be set as: trigger threshold (amplitude variations of stationary part)<0.05×average amplitude of sensed signal; i.e., variation smaller than 5% of the average amplitudes of the sensed signal.

(3) FIG. 7 illustrates a time signal sensed by a sensor unit which indicates the arrival of the first wavetrain induced by an approaching train at distance. Random seismic noise can be observed before and after the train-induced signal.

(4) As an example of a sensor element considered applicable for an embodiment of the present invention, reference is made to the models 4020 dual axis configuration and 4030 triaxial configuration accelerometers from Measurement Specialties Inc., a member of the TE Connectivity Ltd. Family of companies, which is of silicon MEMS type, providing DC Response and dual and triaxial output options, respectively, in respective ±2 g and ±6 g measurement ranges with a nominal 0-200 Hz bandwidth. For the model 4020, sensitivity is 1000 mV/g and residual noise over the passband is 600 microvolts RMS. For the model 4030, sensitivity is 333 mV/g and residual noise over the passband is 240 microvolts RMS. Accordingly, a sensor according to the invention employing the model 4020 dual axis configuration or the model 4030 triaxial configuration accelerometer is capable of a useful dynamic signal range of about 35 dB, which is considered suitable for implementation of the invention.

(5) In the following is described a sanity check of a section of the rail track between two consecutive sensors.

(6) Having a number N of sensors distributed along the track, the correlation coefficient matrix between the signals sensed at a first sensor s.sub.n and at a second sensor s.sub.n+m, wherein n and m are integers, n can take any value, and m can take any value different from zero, is defined as:

(7) $\mathrm{Corr}\left(s_n,s_{n+m}\right)=\frac{\mathrm{Cov}\left(s_n,s_{n+m}\right)}{{\sigma }_{s_n}{\sigma }_{s_{n+m}}}$
where σ.sub.s.sub.n and σ.sub.s.sub.n+m are the standard deviations of the signals at the respective sensors pairs and Cov(s.sub.n, s.sub.n+m) is their covariance. The correlation coefficient is defined in the interval [−1 1].

(8) FIG. 8 illustrates the recording of a signal generated by a instationary source moving in the direction of the sensors and thereby mimicking a train moving towards the sensors. The signal's wave propagation velocity is set to 6,000 m/s while the inter-sensor distance (520) is set to 1 km.

(9) FIG. 9 shows the correlation matrix between the correlation windows of all five sensors. The correlation window is located in the stationary parts of the signals (st.s.). The correlation matrix demonstrates high cross correlation coefficients even between signals recorded at sensors with inter-sensor distances of 4 km.

(10) FIG. 10 illustrates records of the synthetic signal at the eight sensors but with a broken segment inbetween sensors #3 (s.sub.3) and #4 (s.sub.3+1). Due to the break in the propagation medium, the waves generated by the introduced signal do not propagate beyond the breach point (depending on the width of the break), and the cross correlation coefficients between the sensors s.sub.3 and s.sub.3+1 located before and after the breach point, respectively, suddenly drop as it is indicated in the corresponding cross-correlation matrix, which is illustrated in FIG. 11.

(11) The sensed signals that are recorded at consecutively located sensor units s.sub.n, s.sub.n+m . . . and advantageously stored in separate data files that are picked up by the train while passing the respective sensor units, will be cross-correlated against each other, preferably using the herein described cross-correlation algorithm.

(12) The sanity check of the rail track inbetween two sensor units s.sub.n, s.sub.n+m will be based upon the cross-correlation coefficient between the sensed signals at the two sensor units s.sub.n and s.sub.n+m. In the case that the cross-correlation coefficient is lower than a certain threshold, which in an embodiment of the invention is suggested set for Corr (s.sub.n, s.sub.n+m)=0.99, an alert message is sent to the TOCC to indicate the respective section of the rail track between sensor units s.sub.n and s.sub.n+m for a potential broken rail.

(13) The sanity check of the rail track between sensor units s.sub.n, s.sub.n+m can be verified/corroborated by cross-correlating the sensed signals of sensor units that are not directly neighboring each other, i.e., sensor units s.sub.n−1 and s.sub.n+m, sensor units s.sub.n and s.sub.n+m+1, sensor units s.sub.n+m and s.sub.n+m+2, etc.

(14) The sanity check of the rail track between sensor units s.sub.n, s.sub.n+m can be verified/corroborated by cross-correlating the signals generated by consecutive trains and sensed at the vary same sensor units s.sub.n, s.sub.n+m.

(15) In the following, different implementations all covered by the invention will be exemplified.

(16) Using statistics terminology, the cross-correlation coefficients resulting from correlation of signals from rail sensors are advantageously distinguished into: 1) “exact” correlation (Corr. value=1), 2) “strong” correlation (Corr. value>0.8 and <1), 3) “medium” correlation (Corr. value=in range 0.5-0.8), 4) “weak” correlation (Corr. value<0.5 and >0), and 5) “no” correlation (Corr. value=0),
though the values for these thresholds for action could show some variations from one reference to another reference.

(17) In a first example, now referred to as solution 1, thresholds for action are advantageously applied as follows:

(18) a) “exact” and “strong” correlation=>no alert message is sent

(19) b) “medium” correlation=>set flag (“warning”), and perform double-check with neighboring sensors (e.g. sensors s.sub.n and s.sub.n+2) before an alert message is sent

(20) c) “weak” and “no” correlation=>alert message is sent

(21) In a second example, now referred to as solution 2: an alert message trigger logic comprised in an apparatus or method according to the invention is implementing at least part of the procedure described in FIG. 12.

(22) According to a first aspect of the invention, an electronic rail monitoring sensor unit for a railroad rail condition monitoring system is proposed, the rail sensor comprising

(23) a seismic sensor means adapted to sense acoustic energy in the rail,

(24) a coupling means for keeping the sensor unit attached to the rail and the seismic sensor means acoustically coupled to the rail,

(25) a signal data recording means having a clock and in communication with the seismic sensor means, and adapted to make a record of signal data representing acoustic energy sensed by the seismic sensor means as a function of time,

(26) a short-range wireless data communication means in communication with the signal data recording means and adapted to establish a wireless data link to another short-range wireless data communication means located on-board a train for making a transfer of the record of signal data to another second wireless data communication means,
a controller means in communication with the seismic sensor means, the signal data recording means, and the first wireless data communication means, and adapted to identify from acoustic energy sensed by seismic sensor means, an acoustic energy signature of a distant train in motion on the rail, and, in response thereto, to activate the signal data recording means to make the record of signal data and to activate the first wireless data communication means to establish the wireless data link to the second wireless data communication means when within communication range and to make a transfer of the record of signal data to the second wireless data communication means.

(27) An embodiment of the electronic rail monitoring sensor unit according to the first aspect above is also proposed, wherein the clock is a real time clock, and wherein the signal data recording means is adapted to make the record of signal data representing acoustic energy sensed by the seismic sensor means as a function of real time.

(28) According to a second aspect of the invention, an electronic rail monitoring signal data receiver and processing arrangement for a railroad rail condition monitoring system is proposed, the receiver and processing arrangement being adapted to be installed and to operate on-board a train in motion and comprising

(29) a data storage means adapted to store a plurality of records of signal data representing acoustic energy in the rail as a function of time,

(30) a short-range wireless data communication means in communication with the data storage means and adapted to establish, when within wireless communication range another short-range wireless data communication means of a rail monitoring sensor unit attached and coupled to the rail, a wireless data link for receiving a transfer of a record of signal data from the another short-range wireless data communication means and to forward the received transfer of record data for storage in the data storage means with information for locating the respective rail monitoring sensor unit, and
a data processing means adapted to receive the plurality of records of signal data representing acoustic energy in the rail as a function of time at respective ones of the rail monitoring sensor unit, to analyse, immediately after each receiving of a transfer of a record of signal data, the signal data to determine signal data representing acoustic energy in the rail generated by a train in motion on the rail distant from the respective sensor unit and propagated thereto but not to an adjacently located sensor unit, and to output an indication of a propagation discrepancy as an indication of a rail discontinuity.

(31) An embodiment of the electronic rail monitoring signal data receiver and processing arrangement according to the second aspect above is also proposed, comprising a clock, and wherein the data processing means is adapted to analyse the received transfer of record of signal data using time as a parameter for analysis.

(32) According to a third aspect of the invention, an electronic railroad rail condition monitoring system for providing an indication of a rail discontinuity, such as a rail break, is proposed, comprising a plurality of the electronic rail monitoring sensor unit according to the first aspect above attached to the rail in a spaced relationship, and at least one of the electronic rail monitoring signal data receiver and processing arrangement according to the second aspect above installed to operate on-board a train assigned to operate on the railroad.

(33) The present invention is easily adapted for application at different railways from the fact that railway tracks generally consist of standardized components and standardized installation means. This especially applies to the rails which are standardized steel profiles supported by standardized sleepers/crossties, which are typically precast reinforced concrete sleepers/crossties. In respect of the present invention, the entire rail-sleeper system can be considered as a stable system that is not undergoing significant changes, e.g. in terms of stiffness or mass variations, over a period of time of reasonable length.