CA2728105C

CA2728105C - System and method for detecting rock fall

Info

Publication number: CA2728105C
Application number: CA2728105A
Authority: CA
Inventors: Bohdan Nedilko; Iain Weir-Jones
Original assignee: WEIR-JONES ENGINEERING CONSULTANTS Ltd
Current assignee: WEIR-JONES ENGINEERING CONSULTANTS Ltd
Priority date: 2008-06-17
Filing date: 2009-06-17
Publication date: 2013-06-18
Anticipated expiration: 2029-06-17
Also published as: AU2009267754B2; CN102123899A; US9031791B2; US20110313671A1; HK1159572A1; EP2300299A4; EP2300299B1; US20150285927A1; CN102123899B; AU2009267754A1; CA2728105A1; EP2300299A1; WO2010003220A1

Abstract

Aspects of the invention provide systems and methods for using ballast sensors to detect rock fall events in a vicinity of railway tracks or similar roadways or tracks. The ballast sensors are spaced apart from the tracks. Particular embodiments permit the use of signals from the ballast sensors to discriminate rock fall events from other types of events and to detect the hypocenter of a rock fall event.

Description

. . CA 02728105 2012-12-06 I

SYSTEM AND METHOD FOR DETECTING ROCK FALL

Technical Field [0002] This invention relates to detection of rock fall events.
Particular embodiments provide systems and methods for rock fall detection.

Background [0003] Rock fall events and other similar events (e.g. avalanches and washouts) which take place in a vicinity of railway tracks can damage the track, can damage passing trains and, in some cases, can derail passing trains which can in turn cause significant damage to the train and to people and/or property being transported by the train.
Similar events which take place in a vicinity of other transport-ways (e.g. roadways, bridges, subway tracks and the like) can cause similar damage.

[0004] Prior art technology for detecting rock fall in a vicinity of railway tracks involves so called "slide fences." Slide fences incorporate current carrying wires which extend between fence posts alongside the railway track. Falling rock may strike and break one or more of these wires, opening the corresponding circuits and preventing current flow therethrough. This change of current flow may be detected to generate a rock fall indicator. Slide fences are unreliable, because falling rock may not strike or break a wire, but may still represent a danger to a passing train. Slide fences also tend to generate false positive results, for example, when the wire are broken by animals or the like. Additionally, if a slide fence triggers (i.e. a wire is broken), then the slide fence must be repaired (i.e. the broken wire must be replaced) and rail traffic may be delayed until the slide fence is repaired.

[0005] There is a general desire for systems and methods of rock fall detection that overcome or ameliorate these and/or other deficiencies with the prior art.

Summary of the Invention [0006] One aspect of the invention provides a system for detection of rock fall in a vicinity of a section of railway track. The system comprises:
a plurality of ballast sensors spaced apart along the track section, each ballast sensor located in a ballast proximate to the track section but spaced apart from rails and ties associated with the track section and each ballast sensor sensitive to acoustic energy and configured to generate a corresponding ballast sensor signal in response to detecting acoustic 1 5 energy; and a signal processing unit operatively connected to receive the ballast sensor signals from the plurality of ballast sensors, the signal processing unit configured to detect rock fall events in a vicinity of the track section based, at least in part, on the ballast sensor signals.

[0007] Another aspect of the invention provides a method for detection of rock fall in a vicinity of a section of railway track. The method involves: providing a plurality of ballast sensors spaced apart along the track section and locating each ballast sensor in a ballast proximate to the track section but spaced apart from rails and ties associated with the track section, each ballast sensor sensitive to acoustic energy and configured to generate a corresponding ballast sensor signal in response to detecting acoustic energy; receiving the ballast sensor signals from the plurality of ballast sensors; and processing the ballast sensor signals to detect rock fall events in a vicinity of the track section based, at least in part, on the ballast sensor signals.

SUBSTITUTE SHEET (RULE 26) [0008] Other aspects of the invention provide computer program products comprising computer instructions which, when executed by a processor, cause the processor to carry out the methods of the invention.

[0009] Other features and aspects of specific embodiments of the invention are described below.

Brief Description of the Drawings [0010] In drawings which depict non-limiting embodiments of the invention:
1 0 Figure 1 is a schematic depiction of a rock fall detection system according to a particular embodiment configured to detect rock fall in a vicinity of a section of railway track;
Figure 2A shows a sensor array according to a particular embodiment which is suitable for use with the Figure 1 rock fall 1 5 detection system and which incorporates a ballast sensor;
Figure 2B shows a rail sensor which may be incorporated into any one or more of sensor arrays of the Figure 1 rock fall detection system;
Figure 3 is a schematic illustration of a signal processing unit 20 according to a particular embodiment which is suitable for use with the Figure 1 rock fall detection system;
Figure 4A is a plot showing digitized and time stamped sensor data typical for a small rock fall event obtained at the Figure 3 data processor for a number of sensors;
25 Figure 4B is a plot showing digitized and time-stamped sensor data typical for a passing train obtained at the Figure 3 data processor for a number of sensors;
Figure 4C is a plot showing digitized and time-stamped sensor data typical for a passing highrail vehicle obtained at the 30 Figure 3 data processor for a number of sensors;

SUBSTITUTE SHEET (RULE 26) Figure 4D is a plot showing digitized and time-stamped sensor data typical for the activation of an on-sited power generator obtained at the Figure 3 data processor for a number of sensors;
Figure 4E is a plot showing digitized and time-stamped sensor data typical for a significant rock fall event obtained at the Figure 3 data processor for a number of sensors;

Figure 5 schematically illustrates a number of processing parameters which may be determined from the sensor data;

Figures 6A and 6B respectively depict typical digitized, time-stamped sensor data for a rock fall event and the corresponding STA/LTA ratio;
Figures 6C and 6D respectively show a 0.4 second segment of time stamped, digital sensor data and its corresponding FFT
associated with a typical rock fall event;
Figures 6E and 6F respectively show a 0.4 second segment of time stamped, digital sensor data and its corresponding FFT
associated with a typical train event;

Figures 6G and 6H respectively show a 1.0 second segment of time stamped, digital sensor data and its corresponding FFT
associated with a typical highrail vehicle event;

Figures 61 and 6J respectively show an llsecond segment of time stamped, digitized sensor data and its corresponding FFT
associated with the operation of a generator in a vicinity of the Figure 1 track section;
Figure 7A schematically depicts a method for event detection method according to a particular embodiment;
Figure 7B schematically depicts a method for post event processing which may be performed as a part of the Figure 7A event detection method according to a particular embodiment;
Figure 7C schematically depicts a method for estimating a location of a rock fall event which may be performed as a part of the SUBSTITUTE SHEET (RULE 26) Figure 7A event detection method according to a particular embodiment;
Figure 7D schematically depicts a method for taking appropriate action in respect of a rock fall event which may be performed as a part of the Figure 7A event detection method according to a particular embodiment;
Figure 8 is a schematic depiction of the triggered state of a number of sensors in the Figure 1 rock fall detection system in response to a passing train;
1 0 Figure 9A shows a typical response of a number of sensors of the Figure 1 rock fall detection system to a rock fall event;
Figure 9B is a schematic depiction of the triggered state of the Figure 9A sensors;
Figure 10 schematically illustrates a method for deployment of the Figure 1 system according to an example embodiment; and Figure 11 schematically illustrates a method for discriminating a series of events that may be caused by a human or other animal according to a particular embodiment.

Detailed Description [0011] Throughout the following description, specific details are set forth in order to provide a more thorough understanding of the invention.
However, the invention may be practiced without these particulars. In other instances, well known elements have not been shown or described in detail to avoid unnecessarily obscuring the invention. Accordingly, the specification and drawings are to be regarded in an illustrative, rather than a restrictive, sense.

[0012] Aspects of the invention provide systems and methods for using ballast sensors to detect rock fall events in a vicinity of railway tracks or similar tracks. The ballast sensors are spaced apart from the tracks. Particular embodiments permit the use of signals from the ballast SUBSTITUTE SHEET (RULE 26) sensors to discriminate rock fall events from other types of events and to detect the hypocenter of a rock fall event.

[0013] Figure 1 is a schematic depiction of a rock fall detection system 10 according to a particular embodiment configured to detect rock fall in a vicinity of a section of railway track 12. Track section 12 may typically be located in a sloped region which may present a risk of rock fall from the up slope 14 toward the downslope 16. This is not necessary.
Track section 12 may be located in a valley and may have upward slopes on both sides thereof. In some embodiments, the length of track section 12 may be in a range of 100m-5km. In other embodiments, track section 12 may have other lengths. To facilitate this description, a number of direction conventions are used. As shown by the schematic axes shown in Figure 1, the z direction refers to the vertical direction (i.e. the direction of gravity), the y direction is oriented along track section 12 and the x 1 5 direction refers to the direction that crosses track section 12.

[0014] Rock fall system 10 comprises a plurality of sensor arrays 18 disposed along track section 12. As discussed in more detail below, sensor arrays 18 comprise one or more sensors for detecting acoustic and/or vibrational energy. In the illustrated embodiment, there are n sensor arrays 18 corresponding to track section 12. In general, the number n may be any suitable number that provides the functionality described below and may depend on the geotechnical characteristics of the substrate in a vicinity of track section 12.

[0015] Sensor arrays 18 are spaced apart from one another by distances 20 in y-direction. In some embodiments, distances 20 are in a range of 5-100m. In other embodiments, this range is 10-50m. In still other embodiments, this range is 10-30m. Distances 20 may be based on a number of factors, including, by way of non-limiting example:
characteristics of sensors used in sensor arrays 18 (e.g. types of sensors, signal to noise ratio, etc.), geotechnical characteristics (e.g. quality factor of geologic substrate), performance requirements (e.g. magnitude of rock SUBSTITUTE SHEET (RULE 26) fall which it is desired for system 10 to detect) and/or other factors (e.g.
local weather patterns, local natural and/or man-made sources of noise).
Distances 20 may be uniform within system 10, but this is not necessary.
In general, distances 20 may differ between each adjacent pair of sensor arrays 18.

[0016] Each of sensor arrays 18 generates one or more corresponding sensor signals 22. In the illustrated embodiment sensor signals 22 are analog signals, but this is not necessary. In some embodiments, sensor arrays 18 may output digital sensor signals. Sensor signals 22 are 1 0 transmitted along transmission lines 24 to central signal processing unit 26.
Transmission lines, 24 may run through protective conduits (not shown in Figure 1), such as pipes made of suitable metals, plastics, fiber or the like.

Transmission lines 24 may be electrically shielded to prevent electrical interference from external sources and/or to prevent cross-talk between 1 5 signals 22. The schematic illustration of Figure 1 shows a single signal and a single transmission line 24 for each sensor array 18. This is not necessary. In general, sensor arrays 18 may comprise multiple sensors that generate a corresponding plurality of signals 22 which in turn may be transmitted to signal processing unit 26 on a corresponding plurality of 20 transmission lines 24. It will be appreciated by those skilled in the art that signals 22 from sensor arrays 18 may be multiplexed on transmission lines 24 if desired.

[0017] In the illustrated embodiment, system 10 comprises one or more optional image capturing devices 34. Image capturing devices 34 25 may comprise closed circuit television cameras, for example. In some embodiments, image capturing devices 34 capture digital images and/or digital video. Image capturing devices 34 may be controlled by signal processing unit 26 using signals 38 which are delivered to image capturing devices along transmission lines 40. Image data 36 captured by image 30 capturing devices 34 may be transmitted to signal processing unit 26 along the same transmission lines 40. Transmission lines 40 may represent more SUBSTITUTE SHEET (RULE 26) than one actual line. In some embodiments, transmission lines 40 are not required and camera control signals 38 may be wirelessly transmitted from signal processor unit 26 to image capturing devices 34 and image data 36 may be wirelessly transmitted from image capturing devices 34 back to signal processing unit 26.

[0018] Signal processing unit 26 may be housed in a suitably protective enclosure (not shown) ¨ e.g. a small building or the like. At signal processing unit 26, sensor signals 22 are digitized and processed to detect rock fall events. Processing signals 22 to detect rock fall events, 1 0 which is described in more detail below, may involve discriminating rock fall events from other events. By way of non-limiting example, such other events may include passing trains, passing highrail vehicles (e.g. trucks that travel on track section 12), other natural noise sources (e.g.
waterfalls, falling trees or animals) and/or other man-made noise sources (e.g. power 1 5 generators or pedestrians).

[0019] System 10 may optionally include a network connection 28 to a remote workstation 30. Network connection 28 may be a wire network connection, a wireless network connection and/or a fiber optic network connection, for example. In some embodiments, remote workstation 30 20 may be connected to system 10 via network connection 28 to perform a number of functions, which may include (by way of non-limiting example): monitoring the status of system 10, logging or storing data captured by system 10, recalibrating or reconfiguring system 10, updating software used by system 10 or the like. In some embodiments, some or all 25 of the data captured by sensor arrays 18 may be transmitted via network connection 28 to remote workstation 30 and such data may be processed at the remote workstation 30 to detect rock fall events in a similar manner that rock fall events are detected by signal processing unit 26, as described in more detail below.
30 [0020] System 10 may be a modular part of a greater system (not shown) which incorporates other systems 32 similar to system 10. For SUBSTITUTE SHEET (RULE 26) example, signal processing unit 26 may be optionally linked (via network connection 28 or via some other network connection) to similar signal processing units for other systems 32 similar to system 10.

[0021] In the illustrated embodiment of Figure 1, sensor arrays 18 are located on uphill side 14 of track section 12. This is not necessary. In some embodiments, sensor arrays 18 may be additionally or alternatively located on downhill side 16 of track section 12. In some embodiments, a single sensor array 18 may comprise a plurality of acoustic or vibrational energy sensors, some of which may be located on uphill side 14 and some of which may be located on downhill side 16.

[0022] Sensor arrays 18 may each comprise one or more acoustic energy sensors. By way of non-limiting example, suitable acoustic energy sensors may include: electromagnetic induction based sensors (which may be referred to as geophones), accelerometers, piezoelectric sensors, 1 5 electroactive polymer based sensors, optical sensors, capacitive sensors, micromachined sensors or the like. As is known in the art, some acoustic energy sensors may be directional ¨ e.g. some acoustic sensors may have one or more axes on which they are more sensitive to acoustic energy. In some embodiments, the output of these acoustic energy sensors may be generally correlated with (e.g. proportional to) the sensed acoustic energy.
In other embodiments, the output of these acoustic energy sensors may be generally correlated with (e.g. proportional to) other parameters, such as displacement, velocity or acceleration of a sensor component.

[0023] Figure 2A illustrates a sensor array 18 according to a particular embodiment which is suitable for use with rock fall detection system 10. In the Figure 2A embodiment, sensor array 18 comprises a single, uni-axial, electromagnetic induction-type sensor 50 which is located on uphill side 14 of track section 12. Sensor 50 is located in the ballast 52 which supports track section 12 and is spaced apart from track section 12 ¨ i.e. sensor 50 is not in direct contact with tracks 54 or ties 56 .
In this description, this type of sensor 50 (which is located at least in part SUBSTITUTE SHEET (RULE 26) in ballast 52 of track section 12 and is spaced apart from track section 12) may be referred to as a ballast sensor. Sensor 50 may be encased in a protective housing 58, which (in the illustrated embodiment) comprises a grout-filled enclosure which may be made from a suitable material such as suitable plastic, fiber, steel or the like.

[0024] Protective housing 58 (and sensor 50) may be located in a trench 60 which is excavated in ballast 52 alongside track section 12. In the illustrated embodiment, a region 62 surrounding housing 58 is filled with compacted sand, which may improve acoustic conduction and/or protect sensor 50 and transmission line 24 from sharp rocks which may be present in ballast 52, and a remaining region 64 of trench 60 is back-filled with ballast 52. In the illustrated embodiment sensor 50 is coupled to an anchoring stake 66 which may be driven into the substrate below ballast 52 ancUor below sand-filled region 62. Stake 66 may be situated, shaped and/or otherwise configured to provide good acoustic coupling to the geologic substrate in a region of track section 12.

[0025] As mentioned above, sensor 50 of the Figure 2A embodiment is a uni-axial sensor. The sensitivity axis of sensor 50 is the z axis and sensor 50 generates a single corresponding signal 22. In one particular embodiment, signal 22 is generally correlated with (e.g. proportional to) a sensed velocity of a component of sensor 50. However, as discussed above, in other embodiments, signal 22 may be generally correlated with (e.g. proportional to) other parameters, such sensed displacement, acceleration or energy of corresponding sensor components. The inventors have determined that uni-axial (z axis) sensors are sufficient for the purposes of detecting rock fall on suitably steep slopes. It will be appreciated that uni-axial sensors are less costly than multi-axial sensors.
In some environments or in some applications, however, it may be desirable to incorporate multi-axial sensors. Accordingly, in some embodiments, sensor 50 may be multi-axial or sensor array 18 may comprise a plurality of uni-axial sensors oriented in different directions. In SUBSTITUTE SHEET (RULE 26) such embodiments, the number of signals 22 generated by a multi-axial sensor may correspond to its number of axes or the number of signals 22 generated by a plurality of uni-axial sensors may correspond to the number of uni-axial sensors.

[0026] As discussed above, sensor 50 is electronically connected to transmission line 24 for transmission of a corresponding sensor signal 22 to signal processing unit 26. As shown in Figure 2A, transmission line 24 may run through a suitable protective conduit 68, which may be made from a suitable material such as suitable plastic, fiber, steel or the like. In some embodiments, conduit 68 may also house cables 70 (e.g. electrical and/or optical cables) which form part of network connection 28 between system 10 and remote workstation 30 and/or other systems 32 (see Figure 1) and/or transmission lines 40 associated with optional image capturing devices 34.
1 5 [0027] Figure 2B illustrates a sensor 80 which may be incorporated into any one or more of sensor arrays 18. In Figure 2B embodiment, sensor 80 is similar in many respects to sensor 50 (Figure 2A) in that sensor 80 is a uni-axial, electromagnetic induction-type sensor. Sensor 80 differs from sensor 50 in that sensor 80 is mounted (via suitable mounting hardware 82) to track 54 as opposed to being a ballast sensor which is spaced apart from track 54. Sensors which are mounted to track section 12 (including track(s) 54 and/or ties 56) may be referred to in this description as rail sensors. In other respects, sensor 80 may be similar to sensor 50 described above.
[0028] Experiments have determined that rail sensors may be more sensitive to direct contact between falling rocks and track section 12 (e.g.
track(s) 54 and/or ties 56) and may be more sensitive to passing trains or highrail vehicles. In some embodiments, therefore, it is desirable to include one or more rail sensors. However, in some embodiments, it is desirable to include ballast sensors rather than rail sensors or only ballast sensors, because: ballast sensors may be less prone to damage by trains passing along track section 12, ballast sensors may be more robust to maintenance SUBSTITUTE SHEET (RULE 26) of track section 12 which may involve physical manipulation of track section 12 (e.g. lifting track section 12 away from ballast 52), ballast sensors may produce more uniform signals, ballast sensors may exhibit greater differences in spatial attenuation and may therefore lead to more accurate location of the hypocenter of rock fall events and ballast sensors may be less sensitive to high frequency vibrations which may permit lower sampling rates and correspondingly higher bit resolution for the same data acquisition hardware.
[0029] Figure 3 is a schematic illustration of a signal processing unit 26 according to a particular embodiment which is suitable for use with rock fall detection system 10. In the illustrated embodiment, signal processing unit 26 comprises a plurality m of inputs 100 corresponding to transmission lines 24 and signals 22 from sensor arrays 18. Each input signal 100 is provided to corresponding signal conditioning circuitry 102.
1 5 Suitable signal conditioning circuitry 102 is well known to those skilled in the art and, by way of non-limiting example, may comprise anti-aliasing filter(s) and amplifier(s). Conditioned sensor signals 104 are then provided to analog-to-digital converters (ADCs) 106. ADCs 106 sample conditioned sensor signals 104 and provide corresponding digital sensor signals 108. In one particular embodiment, ADCs 106 provide 24 bits of digital resolution (i.e. digital sensor signals 108 comprise a sequence of 24 bit samples), but this is not necessary. In other embodiments, ADCs 106 may output digital sensor signals 108 having other suitable bit depths. The sampling rate of ADCs 106 may be selected to be sufficiently fast to accommodate the frequencies of interest, as described in more detail below.
[0030] Digital sensor signals 108 output from ADCs 106 are provided to data logger 110. In addition to receiving digital sensor signals 108, data logger 110 also receives timing synchronization signal 112 from timing synchronization source 114. In one particular embodiment, timing synchronization source 114 comprises a global positioning satellite (GPS) receiver which receives timing information from one or more satellite SUBSTITUTE SHEET (RULE 26) sources. A GPS-based timing synchronization source 114 is particularly useful in embodiments, where system 10 is a modular component system of a larger system that includes other component system(s) 32 (Figure 1), which other component systems 32 may have their own signal processing units 26 and their own timing synchronization sources 114. In such systems, GPS-based timing synchronization sources 114 could provide synchronous timing signals 112 across modular component system 10 and other component systems 32. In other embodiments, where there is only one signal processing unit 26, timing synchronization source 114 may comprise one or more other sources of timing information. By way of non-limiting example, timing synchronization source 114 may access timing information from an internal or external quartz piezo-electric oscillator, timing synchronization source 114 may comprise a real time clock or a suitable hardware timing chip or the like.
1 5 10031] Using timing synchronization signal 112 and digital sensors signals 108, data logger 110 time stamps, collects and logs the data generated by sensor arrays 118 (Figure 1). Data logger 110 may have access to memory (not expressly shown) and may use any suitable data structure(s) or database protocol(s) for logging digital sensor signals 108 and corresponding time stamp information from synchronization signal 112. Data logger 110 may store information in a manner that is indexed, or otherwise accessible, by time stamp indicia, by corresponding sensor, ancUor by occurrence of an event (as explained in more detail below). In some embodiments, data logger 110 may be operatively connected (via network interface 122 and network connection 28) to remote workstation and/or to other systems 32 (see Figure 1). In the illustrated embodiment, data logger 110 is also operatively connected (via interface 118) to embedded data processor 120.
[0032] In some embodiments, signal conditioning circuitry 102, 30 ADCs 106, and/or data logger 110 may be implemented by a data acquisition unit (DAU) 116. Various DAUs are known to those skilled in SUBSTITUTE SHEET (RULE 26) the art and are commercially available from a number of sources. In some embodiments, DAU 116 may also incorporate its own timing synchronization source 114. In some embodiments, DAU 116 may include other components which are not expressly shown in the Figure 3 illustration. By way of non-limiting example, such components may include digital processing components (e.g. digital filters) or the like.
Suitable DAUs include, by way of non-limiting example, the TMA-24 Microseismic Acquisition Unit available from Terrascience Systems Ltd.
of Vancouver, Canada and other suitable DAUs. In some embodiments, it 1 0 is desirable that the DAU sample at a rate greater than or equal to 500Hz with a bid resolution of 16 or more bits.
[0033] Commercially available DAUs 116 may have a limited number of inputs 100 or a limited data storage capacity. In some embodiments, therefore, signal processing unit 26 may comprise a plurality of DAUs 16, each of which may be configured in a manner similar to that described herein.
[0034] Data logger 110 is operatively connected (via interface 118) to data processor 120. Data processor 120 may be part of a suitably configured computer system (not shown) or may be part of an embedded system. Processor 120 shown schematically in Figure 3 may comprise more than one individual data processor which may be centrally located and/or distributed. Processor 120 may comprise internal memory (not shown) and/or have access to external memory 128. Processor 120 may be programmed with, or otherwise have access to, software 124. As explained in more detail below, processor 120 may execute software 124 which may in turn cause processor 120 to process data obtained from data logger 110 and to generate one or more outputs 126. Processor 120 may also control the operation of DAU 116, data logger 110 and/or system 10 via interface 118. In some embodiments, processor 120 may be operatively connected (via network interface 122 and network connection 28) to remote workstation 30 and/or to other systems 32 (see Figure 1). Processor 120 SUBSTITUTE SHEET (RULE 26) may output some or all of outputs 126 to remote workstation 30 and/or to other systems 32 via network interface 122 and network connection 28.
100351 In embodiments where system 10 includes optional image capturing devices 34, signal processing unit 26 may also comprise image data memory 130 for storing image data 36 captured by image capturing devices 34. Image data 36 may be delivered to image data memory 130 along transmission lines 40 as shown in the illustrated embodiment or may we wirelessly delivered to image data memory 130 using a wireless transceiver (not shown). Data processor 120 may also control image capturing devices 34 using camera control signals 38 which may be transmitted to image capturing devices 34 along transmission lines 40 and/or wirelessly. Camera control signals 38 may permit image capturing devices 34 to move (e.g. pan), zoom, focus or the like and may control when and how image capturing devices 34 capture image data 36.
[0036] Figure 4A is a plot showing digitized and time stamped sensor data obtained at processor 120 for a number of sensors (e.g. sensors 50) within sensor arrays 18. The vertical axis of the Figure 4A plot is measured in binary counts (e.g. digital values output by ADCs 106 and stored in data logger 110) and the horizontal axis of the Figure 4A plot is measured in milliseconds (ms). As discussed above, acoustic energy sensors within sensor arrays 18 may output signals 22 that are generally correlated with sensed velocity of a sensor element. In such embodiments, the binary counts on the vertical axis of the Figure 4A plot may also be correlated with this velocity. Where the acoustic energy sensors within sensor arrays 18 represent other parameters (e.g. displacement, acceleration, energy), the binary counts on the vertical axis of the Figure 4A plot may be correlated with such other parameters. It should be noted that the scales of the vertical axis for the individual sensor plots within Figure 4A are different for each sensor ¨ i.e. the plots corresponding to sensors #1, #(m-.7) and #rn have ranges of approximately (-100,100) binary counts, the plots corresponding to sensors #2 and #4 have ranges of SUBSTITUTE SHEET (RULE 26) approximately (-2500,2500) and the plot corresponding to sensor #3 has a range of approximately (-5000,5000).
[0037] Figure 4A shows typical digitized and time-stamped sensor data obtained at processor 120 for a small event that is detected in a region of sensors #2, #3 and #4 at a time around 3,000-7,000ms. It can be seen that, in the 3,000-7,000ms time period, the magnitude of the sensed signals of sensors #2, #3 and #4 (on the order of thousands of binary counts) is significantly greater than the background noise (on the order of hundreds of binary counts). This event is typical of a small scale rock fall, but may 1 0 also be typical of other small scale events, such as (by way of non-limiting example): raveling of a rock face, a surge in adjacent waterfall activity, one or more animals, vegetation or fencing shaken by wind or the like. For typical applications alongside railways tracks, the scale of the Figure 4A
event may be interpreted to be sufficiently small that it is not of significant concern.
[0038] Figure 4B shows typical digitized and time-stamped sensor data obtained at processor 120 for a passing train. Like Figure 4A, the vertical axis of the Figure 4B plot is measured in binary counts and the horizontal axis is measured in milliseconds. However in the Figure 4B
plot, the vertical axes for each sensor are on the same scale (-2x106,2x106).
It can be seen that the duration of the train event is significantly longer than the event of Figure 4A.
[0039] Figure 4C shows typical digitized and time-stamped sensor data obtained at processor 120 for a passing highrail vehicle. The vertical axis of the Figure 4C plot is measured in binary counts and the horizontal axis is measured in milliseconds. While the vertical scales vary between the individual Figure 4C plots for the individual sensors, it can be seen that the scales of the individual Figure 4C plots for the highrail vehicle have scales that are lower than those of Figure 4B for the train.
[00401 In some environments, there may be additional sources of events which may be particular of the environment in which system 10 is SUBSTITUTE SHEET (RULE 26) deployed. One example of a such an event is the activation of a power generator in vicinity of system 10. Where it is desired for system 10 to operate at a remote location, such a generator may be used to power system 10 itself. Such a generator is not required however and other sources of power (e.g. batteries, solar power or wind power) may be used to power system 10. Figure 4D shows typical digitized and time-stamped sensor data obtained at processor 120 for the activation of a power generator in the vicinity of system 10. The vertical axis of the Figure 4D
plot is measured in binary counts and the horizontal axis is measured in 1 0 milliseconds. The vertical axes for each sensor in the Figure 4D plots is on the same scale (-4000,4000).
[0041] Figure 4E shows typical digitized and time-stamped sensor data obtained at processor 120 for a rock fall event that is of sufficient size to be of concern for typical railway applications. The vertical axis of the 1 5 Figure 4E plot is measured in binary counts and the horizontal axis is measured in milliseconds. It will be noted that the vertical scales vary between the individual Figure 4E plots for the individual sensors. Figure 4E indicates that the rock fall event took place between approximately 19,000-21,000ms. Comparing the vertical scales of the various sensors and 20 the corresponding magnitudes of the sensed signals, it would appear that the rock fall event occurred relatively closer to sensor #m than to any of the other illustrated sensors. Comparing the vertical scales and the corresponding magnitudes of the sensed signals between the Figure 4E
rock fall event and the Figure 4A small scale event, it can be seen that the 25 Figure 4E rock fall event has significantly greater magnitude.
[0042] In some embodiments, rock fall detection by system 10 may be performed by signal processing unit 26 based on signals 22 received from sensor arrays 18 (see Figure 1). In particular embodiments, rock fall detection by system 10 may involve processor 120 processing data from 30 data logger 110 (or DAU 116) to detect rock fall events (see Figure 3).
In other embodiments, rock fall detection may be performed at remote SUBSTITUTE SHEET (RULE 26) workstation 30 and/or at other systems 32 having access to data from data logger 110 (or DAU 116) via network connection 28. For the remainder of this description, it is assumed, without loss of generality, that rock fall detection is performed by embedded processor 120 processing data received from data logger 110.
[0043] A part of rock fall detection performed by system 10 involves discriminating between rock fall events and other types of events which may be of less concern and/or between significant rock fall events and relatively small rock fall events which may be of less concern. An event that is determined by system 10 to be a significant rock fall event, but which in fact is a different event (e.g. a train, a highrail vehicle and/or a power generator) or is an insignificant rock fall event may be referred to in this description as a false positive detection result. In general, there is a desire to minimize false positive detection results.
1 5 [0044] To detect rock fall events while minimizing false positive detection results, processor 120 may process data received from data logger 110 to determine a plurality of processing parameters. Some or all of these processing parameters may be used in turn to discriminate rock fall events from other events. Figure 5 schematically illustrates a number of processing parameters 150 which may be determined by processor 120 using data accessed from data logger 110. Each of these processing parameters 150 is explained in more detail below. In some embodiments, processor 120 may output one or more of processing parameters 150 as outputs 126. As discussed above, some or all of outputs 126 may be available to remote workstation 30 and/or other systems 32 via network connection 28.
[0045] Processor 120 may process the data corresponding to one or more sensors to determine a ratio of a short term average (STA) to a long term average (LTA), which may be one of the processing parameters 150.
This ratio may be referred to as an STA/LTA average and may be computed according to:

SUBSTITUTE SHEET (RULE 26) - 19 ¨

1=n 11X al STA) 1=na where b> a > 0 and n?_a,b (1) LTA) n 7=(n¨(b--1)) where: xi represents the value of the ill' sample, n is the index of the current sample xn, a is the STA duration (number of samples) and b is the LTA

duration (number of samples). Examining equation (1), it will be appreciated that the STA and LTA durations a and b may be expressed as numbers of samples or equivalently as temporal durations.

[0046] Figures 6A and 6B respectively depict typical digitized, time-stamped sensor data obtained at processor 120 for a rock fall event and the corresponding STA/LTA ratio. For the Figure 6B plots, the STA duration a is 20ms and the LTA duration b is 1000ms. It can be seen from the Figure 6A and 6B plots that the rock fall event occurs around the 1250-1750ms time period.

[0047] The STA/LTA ratio is useful for detecting when a signal changes to stand out from background noise and may therefore be 1 5 compared against a suitable threshold to trigger the start and end of an event. For example, when the LTA/STA ratio is greater than an event start threshold (thresh start), then processor 120 may determine that an event has started and the associated time tõ,õ./. Similarly, when an event has started and the LTA/STA ratio falls below an event end threshold (thresh end), then processor 120 may determine that an event has ended and the associated time tend. The STA/LTA threshold parameters thresh start, thresh end may be experimentally determined as a part of the calibration of system 10 and may depend, by way of non-limiting example, on the STA averaging duration a, the LTA averaging duration b, the spectral characteristics (e.g. amplitude and dominant frequencies) of SUBSTITUTE SHEET (RULE 26) the background noise in a vicinity of track section 12 and/or the expected spectral characteristics (e.g. amplitude and dominant frequencies) of an event that system 10 is designed to detect. These threshold parameters may additionally or alternatively be user adjustable.

[0048] The start and end times tsrarr, tend of an event can also be used to determine the event duration tdur as one of the Figure 5 processing parameters 150 according to:

t dur = t end ¨ t start (2) 100491 An issue which may arise with the STA/LTA ratio is the so called "memory" associated with the LTA. The LTA value computed by processor 120 carries with it information about the last b samples (where b is the LTA duration used in equation (1)). In some cases, the last b samples will be influenced by an event. For example, when a train passes track section 12, it typically takes a period of time for the train to pass. In such cases, the last b samples used to compute the LTA may be influenced by the signal associated with the passing train ¨ e.g. the LTA may be relatively large during, or even after, a passing train. In such circumstances, the relatively high LTA may cause the STAJLTA ratio to lose sensitivity, even if the STA is relatively high.
[0050] In some embodiments, therefore, processor 120 processes the data from data logger 110 to determine a modified STA/LTA ratio as one of processing parameters 150. This modified STA/LTA ratio may involve replacing the actual LTA with a constant c according to:
z=n xi 7=( n-(a-1)) STA a where n > a > 0 (3) The constant c may be representative of the LTA during event free times (e.g. times without rock fall or passing trains or the like) and, in some embodiments, may be determined during calibration of system 10. For SUBSTITUTE SHEET (RULE 26) example, the constant c may be determined in a relatively noise free period in the environment where system 10 is deployed prior to the actual deployment of system 10. In one particular embodiment, the constant c may be determined to be an actual LTA during such a noise free period (e.g. determined according to the denominator of equation (1)). The constant c may be user adjustable.
[0051] The modified STA/LTA ratio (equation (3)) may be used in substantially the same manner as the actual STA/LTA ratio (equation (1)) to determine the start and end of an event and the associated times tstart, tend and to determine the associated event duration tdõ. In some embodiments, the modified STA/LTA ratio may be used in addition to or as an alternative to the actual STA/LTA ratio. In some embodiments, the thresholding decision associated with the start and/or end of an event may involve a compound decision wherein both the modified and actual 1 5 STA/LTA ratios are subject to threshold conditions. In some embodiments, the decision as to whether to use the actual STA/LTA ratio, the modified STA/LTA ratio or both (i.e. to determine the start and end of an event and the associated times cart, tend and to determine the associated event duration td,,) may be a user-selectable parameter.
[0052] Another processing parameter 150 that may be determined by processor 120 based on data from data logger 110 may be referred to as a peak particle velocity (PPV). The PPV may represent the magnitude of the sample with the largest absolute value during an event and may be determined according to:

PPV = Ilc t start- -tend / (4) where MAX{.} is an operator that returns the maximum value of the operand and x, represents the value of the ith sample.
[0053] As discussed above, in the illustrated embodiment, sensor arrays 18 comprise one or more acoustic energy sensors (e.g. sensors 50) which output signals 22 correlated with velocity of a sensor component. As SUBSTITUTE SHEET (RULE 26) such, the PPV corresponds to the maximum or peak velocity measured by such sensors ¨ hence the term peak particle velocity. In general, however, sensor arrays 18 may comprise acoustic energy sensors which output signals 22 correlated with other parameters (e.g. energy, displacement and/or acceleration). In such embodiments, PPV should be understood to represent the magnitude of the sample with the largest absolute value during an event in accordance with equation (4) and need not represent velocity in strict sense. In some embodiments, processor 120 may also determine the time tppr associated with the PPV. In some embodiments, processor 120 may also determine a global PPV value PPVglobal which represents the magnitude of the sample with the largest absolute value over all of the recorded samples ¨ i.e. a PPV which is not limited to the times between ts,õ, and tend during an event.
[0054] System 10 may use PPV to help discriminate between 1 5 significant rock falls and other types of events. In one particular embodiment, PPV is subjected to a thresholding process which may filter out small rock fall events, other low magnitude events (e.g. animals) and/or background noise events (e.g. the operation of a power generator).
For example, if the PPV of an event is less than a PPV threshold (thresh PPV), then processor 120 may determine that the event has insufficient magnitude to be a significant rock fall. The PPV threshold parameter thresh PPV may be experimentally determined as a part of the calibration of system 10 and may depend, by way of non-limiting example, on particular minimum magnitude rock fall detection required of system 10, the expected magnitude of low magnitude events (e.g. animals or humans), the expected magnitude of background events (e.g. power stations, waterfalls, wind) and/or the like. The PPV threshold parameter thresh PPV may additionally or alternatively be user adjustable.
[0055] Another processing parameter 150 that may be determined by processor 120 based on data from data logger 110 may be referred to as a the signal energy E. In some embodiments, the signal energy E used by SUBSTITUTE SHEET (RULE 26) system 10 may represent a windowed average of the sample amplitude squared and may be determined according to:

n = i=n¨(d-1) E d (5) where: x,= represents the value of the sample, n is the index of the current sample xn and d is the window duration (number of samples). Examining equation (5), it will be appreciated that the duration d may be expressed as numbers of samples or equivalently as temporal durations.
[0056] Like the STA/LTA ratio discussed above, the signal energy E
is useful for detecting when a signal changes to stand out from background 1 0 noise and may therefore be compared against suitable thresholds to trigger the start and end of an event. For example, when the signal energy E is greater than an event start threshold (E thresh start), then processor 120 may determine that an event has started and the associated time tstarr Similarly, when an event has started and the signal energy E falls below an 1 5 event end threshold (E thresh end), then processor 120 may determine that an event has ended and the associated time tõd. The threshold parameters E_thresh_start, E thresh end may be experimentally determined as a part of the calibration of system 10 and may depend, by way of non-limiting example, on the duration d of the energy window, the 20 spectral characteristics (e.g. amplitude and dominant frequencies) of the background noise in a vicinity of track section 12 and/or the expected spectral characteristics (e.g. amplitude and dominant frequencies) of an event that system 10 is designed to detect. These threshold parameters may additionally or alternatively be user adjustable. The start and end times 25 cart, tend of an event can also be used to determine the event duration tdin as described above (equation (2)).

SUBSTITUTE SHEET (RULE 26) [0057] The signal energy E may also be used in addition to or in the alternative to the STA/LTA ratio in other circumstances where it might be appropriate to use the STA/LTA ratio. The maximum signal energy Eniax=MAX{E,fict start = = = t end! also exhibits a correlation with the PPV
discussed above. In some embodiments, the maximum signal energy may be used in addition to or in the alternative to the PPV value in circumstances where it might be appropriate to use the PPV value.
[0058] Another processing parameter 150 that may be determined by processor 120 based on data from data logger 110 is the spectral power distribution (e.g. frequency content) of a signal corresponding to an event.
In one particular embodiment, processor employs a Fast Fourier Transform (FFT) technique to the sampled data during an event (i.e. between ts,õ, and tend). The spectral power may therefore be referred to in this description as the FFT. As is known in the art, however, there are a number of FFT
1 5 techniques and other techniques for determining the time-frequency content of a digitally sampled signal and any such techniques may be used to determine the time-frequency content of a signal.
[0059] Figures 6C and 6D respectively show a 0.4 second segment of a time stamped, digital sensor signal received at processor 120 and a corresponding FFT associated with a typical rock fall event. Figure 6D
show that most of the spectral power of the digital sensor signals associated with a typical rock fall event is concentrated in the frequency band less than 125Hz. Figures 6E and 6F respectively show a 0.4 second segment of a time stamped, digital sensor signal received at processor 120 and a corresponding FFT associated with a typical train event. Figure 6F
shows that the spectral power of the digital sensor signals associated with a typical train event is spread over 0-4001-Iz and has significant power at frequencies over 200Hz. Figures 6G and 6H respectively show a 1.0 second segment of a time stamped, digital sensor signal received at processor 120 and a corresponding FFT associated with a typical highrail vehicle event. Figure 6H shows that the spectral power of the digital sensor signals associated with a typical highrail vehicle event (like a train event) is spread over 0-400Hz and has significant power at frequencies over SUBSTITUTE SHEET (RULE 26) 200Hz. The sensor data shown in Figures 6C-6H is obtained from representative rail sensors, but the inventors have concluded that similar results are achievable with suitably configured ballast sensors.

[0060] Figures 61 and 6J respectively show an 11 second segment of a time stamped, digital sensor signal received at processor 120 and a corresponding FFT associated with the operation of a generator in a vicinity of track section 12. Figure 6J shows that the spectral power of the digital sensor signals associated with a the generator has a unique frequency signature with harmonics at 30.66Hz, 45.95Hz and 60Hz.
[0061] Figure 7A schematically depicts a method 200 for event detection according to a particular embodiment. Method 200 may be performed in whole or in part by embedded processor 120. Method 200 may make use of data obtained from data logger 110 and/or DAU 116 and may also make use of processing parameters 150. As discussed above, in 1 5 other embodiments, rock fall detection (including method 200 in whole or in part) may be performed by other processors, such as by processors associated with remote workstation 30 and/or other systems 32.

[0062] Method 200 starts at block 201. Method 200 may involve a number of procedures which are similar for the data associated with each sensor¨ e.g. to each particular digital sensor signal 108 (Figure 3). In the illustrated embodiment, these similar procedures are shown by the representative procedures of block 202 (associated with sensor #1) and block 204 (associated with sensor #m). It will be appreciated that, depending on the number of sensors and the corresponding number of digital sensor signals 108, method 200 may generally comprise any suitable number of procedures similar to those of blocks 202, 204. The procedure of block 202 is now described in more detail, it being understood that the procedure associated with block 204 and other simlar blocks may be substantially similar to that of block 202.

[0063] The block 202 procedure starts in block 210 which involves initializing a number of parameters. For example, block 210 may involve obtaining sufficient number of data samples (a) to calculate the STA (the numerator of equation (1) and/or equation (3)) and/or a sufficient number SUBSTITUTE SHEET (RULE 26) of data samples (b) to calculate the LTA (the denominator of equation (1)).
Such data samples may be taken from the digital signal sensor signal 108 associated with block 202. Block 210 may involve resetting a number of the processing parameters 150 which may have been used during previous post event processing (described in more detail below). Block 210 may also involve initializing one or more calibration parameters and/or user-configurable parameters. The procedure of block 202 then proceeds to block 215, which involves obtaining the next data sample ¨ e.g. the next data sample from the digital sensor signal 108 associated with block 202.
[0064] In block 220, block 202 may involve updating one or more processing parameters 150 based on the newly acquired block 215 data and, in some instances, the historical data obtained prior to the current iteration of block 215. In particular embodiments, the particular processing parameters 150 which are updated in block 215 include those associated 1 5 with event-start triggering criteria. As explained above, processing parameters 150 associated with triggering the start of an event may include: the STA/LTA ratio (equation (1)), the modified STA/LTA ratio (equation (3)) and/or the energy (equation (5)).

[0065] Block 225 involves evaluating event-start criteria. The block 225 event-start criteria may involve an evaluation of whether one or more processing parameters (e.g. the STA/LTA ratio (equation (1)), the modified STA/LTA ratio (equation (3)) and/or the energy (equation (5)) are greater than one or more corresponding event-start thresholds (e.g.

thresh start(STA/LTA), thresh start(sTA/LTA),,,d, thresh start(E)). If the block 225 evaluation of the event-start criteria is negative (block 225 NO
output), then the procedure of block 202 loops back to block 215 to obtain another data sample. If on the other hand the block 225 evaluation of the event-start criteria is positive (block 225 YES output), then the procedure of block 202 proceeds to block 230.

[0066] Block 230 involves setting a value for cart' In particular embodiments, the block 230 t811 value may be based on the time associated with the current block 215 data sample. The procedure of block 202 then proceeds to blocks 235 and 240 which involve obtaining the next SUBSTITUTE SHEET (RULE 26) data sample and updating one or more processing parameters in a manner similar to that of blocks 215 and 220 described above.
[0067] Block 245 then involves evaluating event-end criteria. The block 245 event-end criteria may involve an evaluation of whether one or more processing parameters (e.g. the STA/LTA ratio (equation (1)), the modified STA/LTA ratio (equation (3)) and/or the energy (equation (5)) are less than one or more corresponding event-end thresholds (e.g.
thresh end (STA/LTA), thresh en d (STA/LTA)mod, thresh end(E) ). If the block 245¨
evaluation of the event-end criteria is negative, then the block 202 1 0 procedure loops back to block 235 to obtain another data sample. If on the other hand the block 245 evaluation of the event-end criteria is positive, then block 202 procedure determines that the event has ended and proceeds to block 250, which involves setting a value for the event end time tend. In particular embodiments, the block 250 tend value may be based on the time 1 5 associated with the current block 235 data sample.
[0068] In the illustrated embodiment, the block 202 procedure then proceeds to block 255, which involves post event processing. The post event processing of block 255 may involve discriminating between types of events or otherwise determining whether a particular event is a 20 significant rock fall event. In the illustrated embodiment of method 200, the block 255 post event processing is shown within the block 202 procedure ¨ i.e. the block 255 post event processing may be performed for each sensor whose digital signal 108 triggers the detection of an event.
This is not necessary. In some embodiments, the block 255 post event 25 processing may be performed outside of the block 202 procedure ¨ e.g. the block 255 post event processing may be performed on a global basis and/or for a subset of the sensors whose digital signals 108 trigger the detection of an event.
[0069] Figure 7B schematically depicts a method 300 for post event 30 processing which may be performed in block 255 according to a particular embodiment. In the illustrated embodiment, the post event processing of method 300 involves discriminating between a number (e.g. six) of different types of events. In other embodiments, the post event processing SUBSTITUTE SHEET (RULE 26) may involve discriminating between two types of events ¨ i.e. significant rock fall events and any other kind of event. In some embodiments, method 300 may be performed for each sensor whose digital sensor signal 108 triggers the detection of an event. In other embodiments, method 300 may be performed for a subset of the sensors whose digital sensor signals 108 trigger the detection of an event.
[0070] In the illustrated embodiment, method 300 starts in block 305 which involves determining one or more event specific processing parameters 150. Once ts,õ, and tend are determined (eg. in blocks 230, 250) 1 0 for a particular sensor in system 10, processor 120 may obtain a subset of the associated digital sensor signal 108 which occurs between t start and tend.
This data subset may in turn be processed to obtain the block 305 event specific processing parameters 150. Examples of event specific processing parameters that may be determined in block 305 include: the duration tdõõ
of the event which may be determined according to equation (2); the PPV
which may be determined according to equation (4); the time (tpp) associated with the PPV; the spectral power (FFT) of the discrete signal between t start and tend. To the extent that the STA/LTA ratio, the modified STA/LTA ratio or the energy are not determined in the block 202 procedure, then any one or more of these quantities (and/or their associated maxima, STA/LTAT,a,õ STA/LTA.õ,Lina,õ Em ax and the times of their associated maxima) may also be determined in block 305.
100711 Method 300 then proceeds to block 310 which involves evaluating event duration criteria. The block 310 evaluation may involve comparing the event duration tdõ,, to a threshold (thresh_dur) to determine whether the event durationt dur is less than the threshold (thresh dur). In some embodiments, the event duration threshold (thresh dur) may be in a range of 1-3 seconds. In other embodiments, this range may be 2-10 seconds. The magnitude of the event duration threshold (thresh dur) may depend on the typical length of the trains that pass through track section 12.
[0072] If t dur is greater than the event duration threshold (thresh dur), then method 300 may proceed along the block 310 NO
SUBSTITUTE SHEET (RULE 26) output to block 315. In the illustrated embodiment, block 315 involves concluding that the event is either a passing train or a passing highrail vehicle. From block 315, method 300 proceeds to block 320 which involves an evaluation of a magnitude criteria to determine whether the event was triggered by a passing train (block 320 YES output and the conclusion of block 325) or the event was triggered by a highrail vehicle (block 320 NO output and the conclusion of block 330). The block 320 magnitude evaluation may involve comparing the PPV of the associated digital sensor signal 108 to a suitable threshold. If the PPV is greater than 1 0 the block 320 threshold, then the event is determined to be a train (block 320 YES output and block 325 conclusion), whereas if the PPV is less than the block 320 threshold, then the event is determined to be a highrail vehicle (block 320 NO output and block 330 conclusion). Depending on the geological site conditions, in some embodiments block 320 may 1 5 additionally or alternatively involve an evaluation of spectral criteria (e.g.
comparing the FFT of an event to one or more thresholds). Such spectral criteria may be used as an alternative to or in addition to the block 320 magnitude criteria to discriminate a train event from a highrail event.
[0073] Returning to the block 310 evaluation, if tdi, is less than the 20 event duration threshold (thresh dur), then method 300 may proceed along the block 310 YES output to optional block 335. Block 335 involves the optional evaluation of spectral criteria to determine whether an event was triggered by a passing train or highrail vehicle. As discussed above, depending on the geological conditions in a vicinity of track section 12, 25 suitably configured sensors (e.g. ballast sensor 50 of Figure 2A and/or rail sensor 80 of Figure 2B) may generate distinctive frequency characteristics in response to trains and/or highrail vehicles traveling on track section 12.
These distinctive frequency characteristics may be used to discriminate trains or highrail vehicles from other types of events. In one particular 30 embodiment, the block 335 spectral criteria involves determining whether the FFT associated with a digital sensor signal 108 has a significant amount (e.g. x% or more) of its power at frequencies greater than a frequency threshold (thresh Jreq). In one particular embodiment, this SUBSTITUTE SHEET (RULE 26) threshold may be in a range of 100Hz-300Hz. In another embodiment, this threshold may be in a range of 12511z-200Hz. In one particular embodiment, the significant amount (e.g. x% or more) may be in a range of 0%-25%. In other embodiments, the significant amount (e.g. x% or more) may be in a range of 5%-15%.

10074] If the FFT of the rail sensor has a significant amount (e.g. x%
or more) of its power at frequencies greater than a frequency threshold (threshireq), then the block 335 evaluation is positive (YES output) and method 300 proceeds to block 340 which involves concluding that the 1 0 event is either a passing train or a passing highrail vehicle. Blocks 340, 345, 350 and 355 may be substantially similar to blocks 315, 320, 325 and 330 discussed above and may involve discriminating between a train (block 350 conclusion) and a highrail vehicle (block 355 conclusion).

[0075] If, on the other hand, the FFT of the rail sensor does not have 1 5 a significant amount (e.g. x% or more) of its power at frequencies greater than a frequency threshold (threshireq), then the block 335 evaluation is negative (NO output) and method 300 proceeds to block 360 which involves evaluation of magnitude criteria to determine whether the event in question is a significant rock fall event ¨ i.e. a rock fall event worthy of 20 concern. The block 360 magnitude evaluation may involve comparing the PPV of the associated digital sensor signal to a suitable threshold (thresh PPV). In some embodiments, this magnitude threshold (thresh PPV) may be in a range of 500-5000 bits. In other embodiments, this range may be 750-2,500 bits. If the PPV is less than the block 360 25 threshold (thresh PPV), then the event is determined to be an insignificant event (block 360 NO output and block 365 conclusion).
[0076] In the illustrated embodiment, however, method 300 goes beyond the block 365 conclusion of classifying the event as an insignificant event. As discussed above, it may be desirable to discriminate 30 other types of natural or human-made noise that may trigger events in a vicinity of track section 12. In one particular embodiment, a generator (not shown) is located in a vicinity of track section 12. When the generator turns on, it may trigger an event on one or more sensors of system 10. In SUBSTITUTE SHEET (RULE 26) the illustrated embodiment, method 300 proceeds from block 365 to block 370 which involves evaluation of spectral criteria. As explained above, the spectral power associated with the start up and operation of the generator has a particular spectral pattern. Accordingly, spectral criteria can be designed for the block 370 inquiry to determine whether the event was triggered by the generator. Such spectral criteria may involve evaluation of whether the FFT of the associated digital sensor signal 108 has a significant amount (e.g. y% or more) of its power at frequencies within particular frequency bands associated with the start up and operation of the 1 0 generator. If the block 370 evaluation is positive (block 370 YES output), then method 300 concludes that the event was triggered by the generator in block 375.
[0077] It should be noted that the block 370 spectral evaluation and the block 375 conclusion that the event was triggered by a generator 1 5 represent one non-limiting example of the type of criteria which may be used to discriminate other types of natural or human-made surface noise that may trigger events in a vicinity of track section 12. In other embodiments, it might be desirable to use additional or alternative criteria (e.g. in block 370 or in other similar inquiries) to discriminate additional or 20 alternative surface noise events. Such surface noise events may include (by way of non-limiting example): noise created by moving water (e.g.
waterfalls, rivers or the like); noise created by animals; noise created by nearby traffic; noise created by falling trees or the like. The types of criteria used to discriminate these events may include (by way of non-25 limiting example): magnitude criteria, spectral criteria, duration criteria or the like. It is not necessary that the evaluation of these additional or alternative criteria occur in any particular order relative to the other method 300 criteria evaluations. In general, the method 300 criteria evaluations can occur in any desirable order. For example, if it is known 30 that the generator is likely to start every 10 minutes and run for 2 minutes, then it may be desirable to locate the block 370 spectral criteria evaluation at an earlier point within method 300 to quickly conclude generator events and to thereby conserve processing power.
SUBSTITUTE SHEET (RULE 26) [0078] If the block 370 evaluation is negative (block 370 NO

output), then method 300 proceeds to block 380 which involves an evaluation of accumulation criteria to determine whether there has been a sufficient amount of low magnitude rock fall within a sufficiently short period to time to conclude that there has been rock fall accumulation that may be of concern. In one particular embodiment, the block 380 accumulation criteria involves consideration of whether method 300 has reached block 380 (i.e. small rock fall event) more than a threshold number of times (thresh #) within a recent time period AT. By way of non-limiting example, block 380 may involve evaluating whether method 300 has reached block 380 more than 5 times within the last hour. In some embodiments, the block 380 threshold number of times (thresh #) is in a range of 3-50. In other embodiments, this range is 10-20. In some embodiments, the block 380 time period AT is in a range of 30-900 1 5 minutes. In other embodiments, this range is 60-480 minutes.

100791 If the block 380 accumulation criteria evaluation is positive (block 380 YES output), then method 300 proceeds to block 385 which concludes that there has been sufficient rock fall accumulation to be of concern. If the block 390 accumulation criteria evaluation is negative (block 380 NO output), then method 300 proceeds to block 390 which concludes that the event was an insignificant rock fall event.

[0080] Returning to the block 360 magnitude evaluation, if the PPV

is greater than the block 360 threshold (thresh_PPV), then method 300 proceeds (via block 360 YES output) to block 392 which involves an inquiry into whether the current event is followed by a train event. The inventors have determined that when a train passes over track section 12, the passing train can trigger a number of events (e.g. the train can satisfy the block 225 event start criteria) prior to triggering the principal train event. These events which are triggered prior to the principal train event may be referred to as train precursor events.

[0081] Figure 8 is a schematic depiction of the triggered state of a number of sensors in response to a passing train. It can be seen from Figure 8, that a number of train precursor events 303 occur for each sensor SUBSTITUTE SHEET (RULE 26) in the time leading up to the persistent principal train events 304. The inventors have determined that the time during which train precursor events are likely to occur is within a time windowt pre-train prior to the onset of principal train events. In some embodiments, this time window tp-e-tram is in a range of 5-30 seconds. In other embodiments, this time window At pre-tram is in a range of 10-20 seconds.
[0082] In some embodiments, the block 392 inquiry as to whether the event is followed by a train event may involve an inquiry into whether the current event being processed in method 300 is followed within a time 1 0 window At pre_iõin by a persistent train event. As discussed above, a persistent train event can be discriminated on the basis of duration criteria (e.g. block 310), spectral criteria (e.g. block 335), magnitude criteria (e.g.

block 320, block 330) or any suitable combination thereof. If the block 392 inquiry is positive (e.g. the current event is followed by a train event 1 5 within the time window Ltpretrain ¨ block 392 YES output), then method 300 proceeds to block 394 which involves concluding that the current event is a train precursor event. If, on the other hand, the block 392 inquiry is negative (e.g. the current event is not followed by a train event within the time window Atpre-train block 392 NO output), then method 300 20 proceeds to block 395, where the current event is determined to be a significant rock fall event.
10083] As discussed above, method 300 represents one possible embodiment of block 255 of method 200 (Figure 7A). Returning to Figure 7A, at the conclusion of block 255 (e.g. method 300), method 200 25 proceeds to block 260 which involves an inquiry into whether the block 255 post event processing associated with any of the sensors reached a conclusion that the event was a rock fall event (e.g. either the block 390 conclusion of method 300 that the event is an insignificant rock fall event and/or the block 395 conclusion of method 300 that the event is a 30 significant rock fall event). If the event was not a rock fall event (block 260 NO output), then method 200 proceeds to block 265 which involves taking appropriate action for a non-rock fall event.

SUBSTITUTE SHEET (RULE 26) [0084] The nature of the block 265 action may depend on whether any of the non-rock fall events are considered to be important for some reason. The block 265 action may comprise logging the non-rock fall event or doing nothing. In one particular embodiment, the block 265 action may involve generating an event record associated with the non-rock fall event.
The record of the non-rock fall event may include recordal of a number of parameters associated with the event. In particular embodiments, the block 265 record may include one or more of: the event type (e.g. a block 325, 350 train event, a block 330, 355 highrail vehicle event or a block 375 surface noise event); a number of triggered sensors; start and end times of the event which may include the start time (t3.1,õ.1) for the first triggered sensor and the end time (tend) for the last sensor to remain triggered; the PPV and the associated time tppv for each triggered sensor; the maxima (and associated times) of one or more other block 305 event specific 1 5 parameters (e.g. STA/LTA., E or the like); and any other parameter of which data processor 120 may be aware. In some embodiments, the block 265 record may also include one or more of the images of track section 12 which may be captured by cameras 34.
[0085] In some embodiments, block 265 may involve storing the event record in local memory (e.g. in data logger 110, memory 128 ancUor image data memory 130) until such time as signal processing unit 26 is polled for events (e.g. by remote workstation 30 over network connection 28). Depending on the availability of local memory, in other embodiments, block 265 may involve transmitting the event record (e.g. to remote workstation 30 over network connection 28) for remote storage.
[0086] If, on the other hand, the event was a rock fall event (block 260 YES output), then method 200 proceeds to block 270 which involves estimating a location of the rock fall event. Block 270 may involve estimating a location of the rock fall event with a degree of accuracy which is finer than the minimum spacing 20 between sensor arrays 18 of system 10 (see Figure 1). Figure 7C schematically depicts a method 400 for estimating a location of a rock fall event which may be performed in block 270 according to a particular embodiment. Method 400 commences in SUBSTITUTE SHEET (RULE 26) block 410 which involves selecting a group of sensors to be considered for estimating the rock fall location. In some embodiments, block 410 may involve determining the group of sensors to include all triggered sensors whose start times ç1 are within a time window ATs,õ, of the start time t start of a first sensor to trigger on a rock fall event.
100871 This block 410 determination is shown schematically in Figures 9A and 9B. Figure 9A shows a typical response of a number of sensors to a rock fall event. Like Figure 4E, the vertical axis of Figure 9A
plot is measured in binary counts and the horizontal axis is measured in milliseconds. It will be noted that the plots for the individual sensor signals shown in Figure 9A are on different vertical scales, but that the particular scales for each sensors are omitted for clarity. Figure 9A also contrasts with Figure 4E in that Figure 9A is shown on a much smaller time scale ¨
i.e. Figure 4E spans a time period of 30s, whereas Figure 9A spans a time period of 2.2s.
[0088] For each sensor of the Figure 9A plot, Figure 9B shows when the sensor is triggered ¨ i.e. when the sensor's trigger status is ON. It can be seen from Figure 9A, that the sensor #1 is the first to trigger (at t¨istart_sensor#1)- Because acoustic waves take time to travel through the substrate in the region of track section 12, it may be assumed that sensor #1 is closest to the hypocenter of the rock fall event. The block 410 process for selecting the group of sensors to be considered for estimating the rock fall location may involve selecting all of the sensors which are triggered within a time window Tstõ, of t,õLsensor#1-[0089] This time window ZITõ is shown in Figure 9B. It can be seen from Figure 9B that sensors #2-8 and 412-13 are also triggered within this time window AT Accordingly, block 410 may involve selecting sensors 41-8 and 412-13 to be the sensors used for estimating the rock fall location. Figure 9B also shows that sensors 49-11 and #14-15 are not triggered. As such, sensors #9-11 and #14-15 are not selected for estimating the rock fall location in accordance with the illustrated embodiment. Although not explicitly shown in Figure 9B, sensors which SUBSTITUTE SHEET (RULE 26) are triggered after the time window 'A t start may be assumed to be indicative of a different event.

100901 The block 410 time window Tsiaõ may be related to a prediction of the average speed of acoustic waves in the earth near track section 12 and the length of track section 12 being considered. For example, if the length of track section 12 being monitored is lkm and the average speed of acoustic waves in the substrate near track 12 is determined to be 300m/s, then the block 410 time window AT,õ, may be set to be 3.3seconds.
[0091] Once the block 410 group of sensors is selected method 400 (Figure 7C) proceeds to block 415 which involves selecting a group ofp potential locations for the hypocenter of the rock fall event. The p potential hypocenter locations may be spaced apart from one another by a suitable interval d which depends on the location detection accuracy desired from system 10. The number p of potential hypocenter locations may depend, for example, on the processing resources associated with system 10 (e.g.

associated with embedded processor 120). As discussed above, it is logical to assume that the hypocenter of the rock fall may be most proximate to the sensor that triggers first (e.g. sensor #1 in the exemplary circumstance of Figures 9A and 9B). In particular embodiments, the group of p potential hypocenter locations may be provided in a grid around the location of the sensor that is triggered first and the grid may have a spacing of d between potential locations. In some embodiments, the spacing d may be in a range of 1-20m. In some embodiments, this spacing d is in a range of 2-5m. In some embodiments, the number p of potential hypocenter locations is in a range of 5-100. In some embodiments, this number p of potential hypocenter locations is in a range of 10-25.

[0092] Method 400 then proceeds to loop 420. Loop 420 involves carrying out a number of procedures for each of the p potential hypocenter locations determined in block 415. In the illustrated embodiment, loop 420 is indexed by the variable i, which may be referred to as the loop counter.
The loop counter i starts at i=1 on the first iteration of loop 420 and is incremented by one for each iteration of loop 420 until i=p at which point, SUBSTITUTE SHEET (RULE 26) method 400 exits from loop 420. Loop 420 commences in block 425 which involves selecting the i potential hypocenter location and determining the distances xi between the th potential hypocenter location and the locations of the block 410 sensors. It will be appreciated that where the block 410 group of sensors includes N sensors, then the quantity xi will be a IxN
vector quantity having the form [xLõ
xN if where the notation xti indicates the distance between thejth sensor and the th potential hypocenter location. Once the location of the Ih potential hypocenter location is known, the distances xi may then be determined based on pre-calibrated or otherwise known locations of the sensors of system 10.
100931 Method 300 then proceeds to block 430 which involves determining model parameters. In one particular embodiment, the spatial attenuation of the PPV of a rock fall event is modeled according to an exponential decay model:y(x) --= Ae-Bx (6) where: y(x) represents the PPV amplitude at a distance x from the hypocenter of the rock fall event, A is a model parameter representative of the PPV at the hypocenter and B is an absorption coefficient model parameter which may be representative of a quality factor of the substrate in the region of track section 12. In accordance with the above described notation, model equation (6) may be rewritten as:
yA.e-B x.

(7) _I _1 where: yj j is the expected PPV of the 1th sensor based on a rock fall at the th potential hypocenter, x1_, is the distance between the 1117 sensor and the jth potential hypocenter location, A. is a model parameter representative of the PPV at the ith potential hypocenter and Bi is an absorption coefficient model parameter for the th potential hypocenter which may be representative of a quality factor of the substrate in the region of track section 12. In other embodiments, other models may be used for the spatial attenuation of the PPV.
SUBSTITUTE SHEET (RULE 26) [0094] Block 430 then involves solving for the model parameters by minimizing a cost function. In embodiments which make use of the model of equations (6) and (7), the model parameters to be determined in block 430 are the quantities A, and Br It will be appreciated that in embodiments which use other attenuation models, the model parameters to be determined may be different. In one particular example embodiment, the cost function used in block 430 is a least squares cost function which, for the potential hypocenter location, may be given by:

=1w1(r)2 E w, ,(y, - A,e -B x )2 (8) where: F, is the cost function for the ith potential hypocenter location, N is the number of block 410 sensors, WI, is an optional weighting coefficient for theft' sensor and the Ih potential hypocenter location, y, is the actual sensor PPV for the j" sensor and the quantity r.. = y.- Aie-Ax J--1 is J

referred to as the residual for the sensor and the th potential hypocenter 1 5 location.

[0095] The cost function of equation (8) can be minimized when:

OF, - = 0 (9a) OA

and OF, - 0 (9b) Equations (9a) and (9b) can be solved for the i,h potential hypocenter location to yield the parameters A, and B,.

[0096] Solving equations (9a) and (9b) is a non-linear problem which may be simplified (e.g. linearized) by taking the natural logarithm of both sides of equation (7) to yield:

) = ln(A, ) - B,x1, = y'1_1 = A' ,- B ix (10) SUBSTITUTE SHEET (RULE 26) where y,---In(y) and Ac=ln(A). Equation (10) represents a linear regression model (as opposed to the exponential regression model of equation (7) and may be used to create a least squares cost function, which in turn may be minimized to yield the quantities A', and Br The parameter A, may then be obtained according to A. = e . Minimizing a least squares cost function for a linear regression model has a closed form solution and is well understood by those skilled in the art.

[0097] Examining the equation (6) model more closely, it can be seen that the quantity A depends on the amplitude of an event wavelet at the hypocenter and therefore varies from event to event. In contrast, the quantity B represents a quality factor which may depend on the geotechnical characteristics of the environment around track section 12. It would be expected, therefore, that the quantity B is relatively constant from event to event. The inventors have experimentally determined (using 1 5 the least squares curve fitting techniques described above) that the parameter B for a particular track section 12 typically stays within 5-10%

of some average value Bõ. Accordingly, in some embodiments, the quantity B, may be determined during calibration and thereafter the parameter B, may be taken as a constant B1=130. In such embodiments, equations (8) and (9a) may be used to solve for Aõ which is given by:

. .y e B xj-' _1 J

Ai= J=N1 (11) zw .e-2B0x, [0098] For the equation (6) attenuation model, at the conclusion of block 430, method 400 has determined the model parameters A, and B, for the ith potential hypocenter location. For other models, block 430 may yield different model parameters for the Ih potential hypocenter location.

Method 400 then proceeds to block 435 which involves determining an SUBSTITUTE SHEET (RULE 26) error metric associated with the ith potential hypocenter location. In general, the block 435 error metric may be any suitable quantity that is representative of the error associated with the model parameters determined in block 430 for the ith potential hypocenter. The block 435 error metric may involve a summation of constituent error metrics over the N block 410 sensors. Each constituent error metric may involve a difference between the PPV predicted by the loop 420 model and the PPV
measured by the sensor. In embodiments which make use of a least squares cost function (e.g. equation (8)), the block 435 error metric (E,) may 1 0 comprise a sum of the squares of the residuals for the ith potential hypocenter location over the N block 410 sensors which may be given by:
E,= J=1 (r11)2 (12) in embodiments which make use of the regression model of equation (7), equation (12) becomes:
Ei = E (y, - yõJ=1 (13) where y is the PPV value measured at the jth sensor and y11 is the PPV
value predicted by model equation (7) for the jth sensor and the ith potential hypocenter location.
[0099] Once the block 435 error metric (E,) is determined, method 400 proceeds to block 440 which involves evaluation of a loop exit condition. If there are other potential hypocenter locations in the block 415 group of p potential hypocenter locations which have yet to be examined, then the block 440 inquiry is negative (NO output) and method loops back to block 450, where the loop counter i is incremented to refer to the next potential hypocenter location and then back to block 425. If the procedures of blocks 425, 430 and 435 have been performed for all of the block 415 SUBSTITUTE SHEET (RULE 26) group ofp potential hypocenter locations, then the block 440 loop exit condition is fulfilled (YES output) and method 400 proceeds to block 445.

[0100] Block 445 involves selecting one of the block 415 group of potential hypocenters to be the estimated hypocenter of the rock fall event.

In the illustrated embodiment, the block 445 selection is based on the hypocenter having the lowest error metric (as determined in block 435). In embodiments which use the error metric of equation (12) or (13), block 445 involves selecting the estimated location of the hypocenter of the rock fall event to be the potential hypocenter having the lowest value of Ei.

101011 In other embodiments, method 400 may be implemented with a different attenuation model. For example, in one alternative embodiment, the attenuation model of equation (6) may be replaced by the following model:

y (x) = A e-Bx (14) 1 5 which represents a combination of absorption and geometric spreading.

For the equation (14) model, the equivalent to equation (7) is:

Ai e-BtxJ _I
v= ._1 . (15) -f \fXjJ

the least squares cost function equivalent to equation (8) is:

= E )2 = w1, (y A -Bx e '-µ) (16) J-1 j=1 VX

and the residual r1_, for the ill' sensor and the ith potential hypocenter location is given by:

SUBSTITUTE SHEET (RULE 26) A_e-Brxj-1 (17) y r. =

J J

Al X j Equation (15) may be linearized by taking the natural logarithm of both sides to obtain:

) = ln(A, ) Bix ¨ ln(x ) = = A', ¨ ln(xj_, ) (18) where y J=ln(yi_) and A' ,--ln(A ). When the assumption is made that equations (9a) and (16) may be used to solve for A, according to:

¨Box ) Ew , e J

X
_1 (19) i A = .1-=1 e-2Boxj_, EVVii X
j=1 [0102] The variables used in the model of equations (14)-(19) may have substantially the same meaning as those used in the above described model based on equations (6)-(11). Method 400 using the model set out in equations (14)-(19) may be similar to method 400 described above using the model of equations (6)-(11). More particularly, block 430 may involve determining the model parameters A, and B, for the ith potential hypocenter, block 435 may involve determining an error metric associated with the ilh 1 5 potential hypocenter (e.g. using equations (12) and (13), except that the equation (15) regression model is used in the place of the equation (7) regression model) and the remainder of method 400 may be substantially similar to that described above.

[0103] As discussed above, method 400 represents one possible embodiment of block 270 of method 200 (Figure 7A). The block 445 estimated hypocenter location may be the output of the block 270 event location estimation. In other embodiments, block 270 may be implemented SUBSTITUTE SHEET (RULE 26) by other methods. Returning to Figure 7A, at the conclusion of block 270 (e.g. method 400), method 200 may proceed to optional block 275 which involves estimating an event energy. The optional block 275 energy estimation may also be based on the spatial attenuation model used in block 270 to estimate the rock fall hypocenter. In the particular exemplary embodiment described in method 400 above, the spatial attenuation model is represented by equations (6) and (7).
[0104] If it is assumed that the trajectory associated with a falling rock is predominantly vertical, then the rock's kinetic energy may be expressed as:

KE mhg (14) where: m is the mass of the rock, h is the height from which the rock falls and g is the acceleration due to gravity. The block 275 energy estimation may also involve the assumption that the PPV of a rock fall event at the 1 5 hypocenter is proportional to the kinetic energy KE:

KE = kA (15) where: A is the value of the model parameter A. determined in block 430 and associated with the hypocenter selected in block 445 and k is a constant of proportionality which may be determined experimentally during calibration of system 10. The block determination of the energy associated with a rock fall event may be determined using equation (15).

101051 Method 200 (Figure 7A) proceeds to block 280 which involves taking the appropriate action for a rock fall event. Figure 7D
schematically illustrates a method 500 for taking appropriate action in respect of a rock fall event which may be performed in block 280 according to a particular embodiment. Method 500 commences in optional block 510 which involves obtaining one or more images of the estimated event location. The estimated event location may be the event location determined in block 270 (e.g. the block 445 hypocenter). Optional block 510 may involve controlling one or more image capturing devices 34 using camera control signals 38. Image capturing device(s) 34 may be controlled, SUBSTITUTE SHEET (RULE 26) so as to direct them toward the estimated event location and to capture corresponding image data 36. As discussed above, image data 36 may be stored in image data memory 130.
[0106] Method 500 then proceeds to block 515 which involves generating a record of the rock fall event. The record of the rock fall event may include recordal of a number of parameters associated with the rock fall event. In particular embodiments, the block 515 record may include one or more of: the event type (e.g. a block 395 or block 390 rock fall event); a number of triggered sensors; a number N of block 410 sensors;
start and end times of the event which may include the start time (t,) for the first triggered sensor and the end time (tend) for the last sensor to remain triggered; the estimated location of the hypocenter of the event (e.g. the block 445 hypocenter); the estimated PPV of the event at the hypocenter (e.g. the value of the model parameter A, determined in block 430 for the 1 5 hypocenter selected in block 445); the estimated event energy (e.g. the block 275 energy); the PPV and the associated time tppv for each triggered sensor; the maxima (and associated times) of one or more other block 305 event specific parameters (e.g. STA/LTAnia,õ Erna, or the like); and any other parameter of which data processor 120 may be aware. In some embodiments, the block 515 record may also include one or more of the block 510 images of the estimated event location.
[0107] Method 500 then proceeds to block 520 which involves evaluation of an alarm criteria. In one particular embodiment, the block 520 alarm criteria may involve comparison to determine whether the estimated PPV of the event at the hypocenter (e.g. the value of the model parameter A, determined in block 430 for the hypocenter selected in block 445) is greater than a PPV alarm threshold (thresh PPV alarm). In other embodiments, the block 520 alarm criteria may involve comparison to determine whether the estimated event energy (e.g. the block 275 energy) is greater than an energy alarm threshold (thresh KE_alarm). It will be appreciated, based on equation (15) above, that in the above-described embodiment, these two block 520 alarm criteria are equivalent and related by the experimentally determined scaling factor k. In some embodiments, SUBSTITUTE SHEET (RULE 26) the block 520 alarm criteria may be additionally or alternatively based on inquiries into one or more other parameter(s) measured or estimated by system 10.
[01081 If the block 520 inquiry is positive (e.g. the estimated PPV of the event at the hypocenter is greater than a PPV alaim threshold (thresh PPV alarm)), then method 500 proceeds along the block 520 YES output to block 530 which may involve triggering an alarm and/or transmitting the block 515 event record directly back to an offsite location (e.g. via network connection 28 to remote workstation 30). The block 530 1 0 alarm may involve triggering sensory stimulus at remote workstation 30 and/or an email at remote workstation 30 or the like. In some embodiments, when the block 530 alarm is received at remote workstation 30, the block 515 event record (including any block 510 images) may be evaluated by human personnel. If the event is determined by human 1 5 personnel to be worthy of service disruption, then vehicular traffic may be prevented from traveling on track section 12 until the event is investigated more thoroughly and/or cleared. In some embodiments, human intervention may not be desired or required and the block 530 alarm may cause a communication to be directed to rail vehicle operators to alert them 20 to the event and to cause them to stop traveling on or toward track section 12.
[0109] If the block 520 inquiry is negative (e.g. the estimated PPV of the event at the hypocenter is less than the PPV alarm threshold (thresh PPV alarm)), then method 500 proceeds along the block 520 NO
25 output to block 540. In the illustrated embodiment, block 540 involves transmitting and/or logging the event in the normal course (i.e. without triggering an alarm). Block 540 may involve storing the block 515 event record in local memory (e.g. in data logger 110, memory 128 and/or image data memory 130) until such time as signal processing unit 26 is polled for 30 events (e.g. by remote workstation 30 over network connection 28).
Depending on the availability of local memory, in other embodiments, block 540 may involve transmitting the block 515 event record (e.g. to SUBSTITUTE SHEET (RULE 26) remote workstation 30 over network connection 28) without triggering an alarm.
0110] Figure 10 schematically illustrates a method 600 for deployment of system 10 according to an example embodiment. Method 600 commences in block 610 which involves assessing geotechnical characteristics of the environment in the vicinity of track section 12. Block 610 may involve simulating rock fall events using drops of known weights from known heights at known locations (i.e. test drops). Block 610 may involve using portable sensor arrays (similar to sensor arrays 18) and 1 0 portable signal processing units 26. Block 610 may involve assessing one or more of:
= ambient noise characteristics (including, by way of non-limiting example, characterizing noise from sources such as waterfalls, running water sources, winds, nearby traffic and other sources of 1 5 surface noise);
= the surface wave velocity in the substrate in a vicinity of track section 12;
= the soil quality factor (e.g. the parameter Bo described above);
= the accuracy range of block 270 location estimation method (e.g. the 20 inventors have determined that the accuracy of method 400 may decrease with distance between the sensors and the rock fall location);
= assessing an amount of data scattering resulting from acoustic energy crossing an obstacle (e.g. the track in circumstances where 25 sensors are installed on both sides of track section 12 or if track section 12 has curvature) and, if significant energy loss takes place, then determining that data from "shadowed" sensors should not be mixed with the remaining sensors;
= verifying that gain settings associated with signal conditioning 30 circuitry 102 are suitable to capture events in a range of interest;
and/or = the like.

SUBSTITUTE SHEET (RULE 26) [0111] Method 600 then proceeds to block 620 which involves using the block 610 information to determine a sensor density for system 10 and determining the associated system layout. Block 620 may involve comparing PPVs associated with test drops at various distances against background noise. In some embodiments, the PPV of the smallest event necessary to be detected should be greater than 3 times the background noise level. In other embodiments, this ratio is 4-5 times.
[0112] Method 600 then proceeds to block 630 which involves installing system 10 in accordance with the block 620 layout. The portable sensor arrays and portable signal processing units may be replaced by permanent sensor arrays 18 and signal processing unit 26. Some of the geotechnical parameters determined in block 610 may be reassessed using the permanent system components. Optional block 640 may involve determining a background noise level (and an associated LTA constant c) which may be used, in some embodiments, to compute the modified STA/LTA ratio in accordance with equation (3) described above.
[0113] Method 600 then proceeds to block 650 which involves testing system 10 by running system 10 for a period of time sufficient to capture all types of detectable events. System 10 (and in particular software 124 used by data processor 120) may be adjusted as needed during this testing period to optimize performance. Digital sensor signals associated with particular events may be recorded so that they can be used again to evaluate changes to software 124. In block 660, system 10 is commissioned to operate, but is subject to regular routine testing and recalibration as desired.
[0114] In some embodiments, it may be desirable to attempt to discriminate events (and/or series of events) caused by human(s) and/or other animal(s) from events caused by rock fall, rail traffic and/or other source of surface noise. Figure 11 illustrated a method 700 which may be used to discriminate a series of events that may be caused by a human or other animal according to a particular embodiment. Method 700 may be performed in method 200 (Figure 7A) after the detection of an event. For SUBSTITUTE SHEET (RULE 26) example, method 700 may be performed between blocks 260 and 265 and/or between blocks 275 and 280.
[0115] Method 700 commences in block 710 which involves determining a temporal correlation between the current event and the previous M events. The parameter M may be based on empirical evidence and may depend on sensor sensitivity, the importance of detection of animals in the vicinity of track section 12 or the like. The block 710 temporal correlation may be determined using a wide variety of techniques known to those skilled in the art. One such technique involves determining a mean time of the last Mevents (e.g. the mean start time tsiõ, of the last M
events) and comparing the time that is furthest from the mean time with the mean time. In accordance with this technique, a large difference indicates a fairly weak temporal correlation and a small difference indicates a fairly strong temporal correlation. Another technique involves 1 5 computing the statistical standard deviation a of the times of the last M
events (e.g. the mean start times tstarl of the last Mevents). In accordance with this technique, a large deviation indicates a relatively weak temporal correlation, whereas a small deviation indicates a relatively strong temporal correlation.
[0116] If the block 710 inquiry indicates that the temporal correlation of the last M events is less than a threshold (temp_corr_thresh), then method 700 may proceed along the block 710 NO output to block 720 where method 700 concludes that the series of events were not produced by an animal. If, on the other hand, the block 710 inquiry indicates that the temporal correlation of the last M events is greater than a threshold (temp corr thresh), then method 700 may proceed along the block 710 YES output to block 730. Block 730 involves evaluation of a spatial correlation of the last N events. In some embodiments, the block 710 number of events M is equal to the block 730 number of events N. The block 730 spatial correlation may be determined on the basis of the event locations determined in block 270 (e.g. method 400), for example. The block 730 spatial correlations may be determined using any of a large SUBSTITUTE SHEET (RULE 26) variety of techniques known to those skilled in the art, including those described above for block 710.
101171 If the block 730 inquiry indicates that the spatial correlation of the last N events is less than a threshold (spat corr_thresh), then method 700 may proceed along the block 730 NO output to block 740 where method 700 concludes that the series of events were not produced by an animal. If, on the other hand, the block 730 inquiry indicates that the spatial correlation of the last N events is greater than a threshold (spat_corr_thresh), then method 700 may proceed along the block 730 YES output to block 750, which involves concluding that the series of events was most likely caused by human(s) or other animal(s).
101181 As will be apparent to those skilled in the art in the light of the foregoing disclosure, many alterations and modifications are possible in the practice of this invention without departing from the spirit or scope thereof. For example:
= The particular embodiments of the methods described above are exemplary in nature. In other embodiments, portions of these methods my be modified or changed. In some embodiments, aspects of these methods may be performed in suitable orders other than the orders described above. By way of non-limiting example, in some embodiments of method 300 (Figure 7B), the procedures of blocks 360-390 may be performed before the procedures of blocks 310-330 and/or blocks 335-355 or the procedures of blocks 335-355 may be performed prior to the procedures of blocks 310-330. Those skilled in the art will appreciate that there are other circumstances in which the order of particular operations may be changed in circumstances where this is desirable.
= Method 300 of the illustrated embodiment described above involves discriminating a variety of different types of events (i.e. train events, highrail vehicle events, surface noise events, insignificant rock fall events and significant rock fall events. This is not necessary. In some embodiments, it is desirable to discriminate a smaller number of events (e_g_ the two categories of significant rock fall events and SUBSTITUTE SHEET (RULE 26) other events). In such embodiments, method 300 may be suitable modified such that the block 310 NO output, optional block 335 YES output, block 360 NO output and block 392 YES output all lead to the same conclusion (i.e. other type of event) and method 300 may conclude a significant rock fall event when the block 392 inquiry is negative (i.e. block 392 NO output). In other such embodiments, some of blocks 365-390 may be maintained to discriminate small rock fall events or rock fall accumulation.
= In some embodiments, method 200 and/or method 300 may be modified to provide an inquiry into a minimum delay between events (devent). Events which occur at within a time (and/or number of samples) separation less than the minimum delay devent from one another may be deteimined to belong to the same event. In particular embodiments, such closely spaced events may be merged into a single event or one or more of such closely spaced events may be ignored.

= In some embodiments, method 200 and/or method 300 may be modified to provide an inquiry into a minimum number of triggered sensors (#_sensor_min). If the number of sensors triggered by an even is less than this minimum number of sensors (# sensor_min), then the event can be determined to be too small to be of concern.
= Figure 9B described above makes use of a parameter L Ttari to determine the sensors to be included in the block 410 group of sensors. This parameter dtõaõ or a similar (possibly larger) temporal parameter may be used to determine a maximum arrival time difference. If a first sensor is triggered at a time tstart_sensor# I and one or more sensors become triggered after this maximum arrival time difference, then the subsequently triggered sensors can be determined to belong to a separate event. The maximum arrival time difference can be determined based at least in part of the experimentally determined surface wave velocity in the substrate in a vicinity of track section 12.

SUBSTITUTE SHEET (RULE 26) = Some of the above described embodiments describe using an experimentally determined average Bo for the model parameter B, of equation (7) and/or equation (15). In some embodiments, this parameter Bo may be the same for a particular track section 12, but may differ as between each of a plurality of modular track sections 12 which may be incorporated into a overall system or this parameter Bo may vary locally within a track section (12).
= The methods described above involve the discrimination of a number of events. These events represent non-limiting examples of events that may be discriminated by system 10. In other embodiments, system 10 may be configured to discriminate other types of events, such as, by way of non-limiting example: switch points being moved, locomotive bells/horns and thermal expansion and the accompanying rail creep atop the ties caused by solar 1 5 heating of the rail.

Accordingly, the scope of the invention should be determined in accordance with the following claims.

SUBSTITUTE SHEET (RULE 26)

Claims

WHAT IS CLAIMED IS:

1. A system for detection of rock fall in a vicinity of a section of railway track, the system comprising:
a plurality of ballast sensors spaced apart along the track section, each ballast sensor located in a ballast proximate to the track section but spaced apart from rails and ties associated with the track section and each ballast sensor sensitive to acoustic energy and configured to generate a corresponding ballast sensor signal in response to detecting acoustic energy;
a signal processing unit operatively connected to receive the ballast sensor signals from the plurality of ballast sensors, the signal processing unit configured to detect rock fall events in a vicinity of the track section based on the ballast sensor signals.

2. A system according to claim 1 wherein the signal processing unit is configured to detect a plurality of different types of events comprising rock fall events, train events wherein a train travels over the track section and highrail vehicle events wherein a highrail vehicle travels over the track section and wherein the signal processing unit is configured to discriminate rock fall events from train events or highrail vehicle events.

3. A system according to claim 2 wherein the signal processing unit is configured to detect an event for a particular one of the ballast sensors based on its corresponding ballast sensor signal.

4. A system according to claim 3 wherein the signal processing unit is configured to detect a start of the event, for the particular one of the ballast sensors, when a short term average/long term average STA/LTA parameter associated with the corresponding ballast sensor signal is greater than a start trigger threshold thresh_start.

5. A system according to claim 4 wherein the signal processing unit is configured to detect a time t start associated with the start of the event.

6. A system according to claim 5 wherein the signal processing unit is configured to detect an end of the event, for the particular one of the ballast sensors, when the short term average/long term average STA/LTA parameter associated with the corresponding ballast sensor signal is less than an end trigger threshold thresh_end.

7. A system according to claim 6 wherein the signal processing unit is configured to detect a time tend associated with the end of the event.

8. A system according to any one of claims 4 to 7 wherein the signal processing unit is configured to determine the short term average/long term average STA/LTA parameter for the corresponding ballast sensor signal according to:

where: x i represents a value of an i th sample of the corresponding ballast sensor signal, n is an index of a current sample x n, a is a short term average STA duration constant, b is a long term average LTA duration constant and is the short term average/long term average STA/LTA parameter.

9. A system according to any one of claims 4 to 7 wherein the signal processing unit is configured to determine the short term average/long term average STA/LTA parameter for the corresponding ballast sensor signal according to:

where: x i represents a value of an i th sample of the corresponding ballast sensor signal, n is an index of a current sample x n, a is a short term average STA duration constant, c is an experimentally determined constant that is representative of a long term average LTA during event free times and is the short term average/long term average STA/LTA parameter.

10. A system according to claim 3 wherein the signal processing unit is configured to detect a start of the event, for the particular one of the ballast sensors, when an energy parameter associated with the corresponding ballast sensor signal is greater than a start trigger threshold E_thresh_start, the energy parameter comprising a windowed average of a squared amplitude of the corresponding ballast sensor signal.

11. A system according to claim 10 wherein the signal processing unit is configured to detect a time t start associated with the start of the event.

12. A system according to claim 11 wherein the signal processing unit is configured to detect an end of the event, for the particular one of the ballast sensors, when the energy parameter associated with the corresponding ballast sensor signal is less than an end trigger threshold E_thresh_end.

13. A system according to claim 12 wherein the signal processing unit is configured to detect a time t end associated with the end of the event.

14. A system according to any one of claims 10 to 13 wherein the signal processing unit is configured to determine the energy parameter for the corresponding ballast sensor signal according to:

ballast sensor signal, n is an index of a current sample x n, d is a window duration constant and E n, is the energy parameter.

15. A system according to any one of claims 7 and 13 wherein the where: x i represents a value of an i th sample of the corresponding signal processing unit is configured to determine a duration t dur

16. A system according to claim 3 wherein the signal processing unit is configured to determine a duration t dur associated with the event for the particular one of the ballast sensors based on its corresponding ballast sensor signal.

17. A system according to any one of claims 15 and 16 wherein the signal processing unit is configured to compare the event duration t dur with one or more duration criteria and to determine on the basis of this comparison that the event is not a rock fall event.

18. A system according to claim 17 wherein the one or more event duration criteria comprise a duration threshold thresh_dur and the signal processing unit is configured to determine that the event is not a rock fall event when the event duration t dur is greater than the duration threshold thresh dur.

19. A system according to 18 wherein the signal processing unit is configured to determine that the event is a train event or a highrail vehicle event when the event duration t dur is greater than the duration threshold thresh_dur.

20. A system according to claim 19 wherein the signal processing unit is configured to determine, for the particular one of the ballast sensors, a peak particle velocity PPV parameter which represents a magnitude of a sample of the corresponding ballast sensor signal with the largest absolute value during the event and wherein the signal processing unit is configured to compare the peak particle velocity PPV parameter with one or more magnitude criteria and to determine on the basis of this comparison whether the event is a train event or a highrail vehicle event.

21. A system according to any one of claims 3 to 18 wherein the signal processing unit is configured to determine a spectral power distribution associated with the event for the particular one of the ballast sensors based on its corresponding ballast sensor signal.

22. A system according to claim 21 wherein the spectral power distribution comprises a fast Fourier transform FFT of the corresponding ballast sensor signal.

23. A system according to any one of claims 21 to 22 wherein the signal processing unit is configured to compare the spectral power distribution with one or more spectral criteria and to determine on the basis of this comparison that the event is not a rock fall event.

24. A system according to claim 23 wherein the one or more spectral criteria comprise a frequency threshold thresh_freq and the signal processing unit is configured to determine that the event is not a rock fall event when the spectral power distribution comprises more than a particular percentage of its power at frequencies above the frequency threshold thresh_freq.

25. A system according to 24 wherein the signal processing unit is configured to determine that the event is a train event or a highrail vehicle event when the spectral power distribution comprises more than the particular percentage of its power at frequencies above the frequency threshold thresh_freq.

26. A system according to claim 25 wherein the signal processing unit is configured to determine, for the particular one of the ballast sensors, a peak particle velocity PPV parameter which represents a magnitude of a sample of the corresponding ballast sensor signal with the largest absolute value during the event and the signal processing unit is configured to compare the peak particle velocity PPV parameter with one or more magnitude criteria and to determine on the basis of this comparison whether the event is a train event or a highrail vehicle event.

27. A system according to any one of claims 3 to 26 comprising one or more rail sensors, each rail sensor operatively contacting the rails or the ties associated with the track section and each rail sensor sensitive to acoustic energy and configured to generate a corresponding rail sensor signal in response to detecting acoustic energy and wherein the signal processing unit is operatively connected to receive the rail sensor signal.

28. A system according to claim 27 wherein the signal processing unit is configured to determine a rail sensor spectral power distribution associated with the event for a particular one of the rail sensors based on its corresponding rail sensor signal.

29. A system according to claim 28 wherein the signal processing unit is configured to compare the rail sensor spectral power distribution with one or more spectral criteria and to determine on the basis of this comparison that the event is not a rock fall event.

30. A system according to any one of claims 3 to 29 wherein the signal processing unit is configured to determine a peak particle velocity PPV parameter associated with the event for the particular one of the ballast sensors and its corresponding ballast sensor signal, the peak particle velocity PPV parameter representing a magnitude of a sample of the corresponding ballast sensor signal with the largest absolute value during the event.

31. A system according to claim 30 wherein the signal processing unit is configured to compare the peak particle velocity PPV with one or more magnitude criteria and to determine on the basis of this comparison that the event is not a significant rock fall event.

32. A system according to claim 31 wherein the one or more magnitude criteria comprise a magnitude threshold thresh_PPV
and the signal processing unit is configured to determine that the event is not a significant rock fall event when the peak particle velocity PPV is less than the magnitude threshold thresh_PPV.

33. A system according to any one of claims 3 to 32 wherein the signal processing unit is configured to ascertain whether the event occurs within a time window .DELTA.t pre-train of a subsequent train event and if the event does occur within the time window .DELTA.t pre-train to determine that the event is a train precursor event.

34. A system according to any one of claims 3 to 33 wherein the signal processing unit is configured to ascertain whether the event is one of a threshold number thresh_# of events that have occurred within a recent time period .DELTA.T and if the event is one of the threshold number thresh_# of events that have occurred within the recent time period .DELTA.T to determine that the event is part of a rock fall accumulation.

35. A system according to claim 3 wherein the signal processing unit is configured to:
determine a duration t dur associated with the event for the particular one of the ballast sensors based on its corresponding ballast sensor signal and to determine that the event may be a rock fall event if the event duration t dur is less than a duration threshold thresh_dur; and determine a spectral power distribution associated with the event for the particular one of the ballast sensors based on its corresponding ballast sensor signal and to determine that the event may be a rock fall event if the spectral power distribution comprises less than a particular percentage of its power at frequencies above a frequency threshold thresh_freq; and wherein the signal processing unit is further configured, after determining that the event duration t dur and the spectral power distribution indicate that the event may be a rock fall event, to determine a peak particle velocity PPV parameter associated with the event for the particular one of the ballast sensors and its corresponding ballast sensor signal, the peak particle velocity PPV parameter representing a magnitude of a sample of the corresponding ballast sensor signal with the largest absolute value during the event and to determine that the event is a significant rock fall event when the peak particle velocity PPV is greater than a magnitude threshold thresh_PPV or a small rock fall event when the peak particle velocity PPV is less than the magnitude threshold thresh_PPV.

36. A system according to claim 3 wherein the signal processing unit is configured to:
determine a duration t dur associated with the event for the particular one of the ballast sensors based on its corresponding ballast sensor signal and to determine that the event may be a rock fall event if the event duration t dur is less than a duration threshold thresh_dur; and wherein the signal processing unit is further configured, after determining that the event duration t dur indicates that the event may be a rock fall event, to determine a peak particle velocity PPV parameter associated with the event for the particular one of the ballast sensors and its corresponding ballast sensor signal, the peak particle velocity PPV parameter representing a magnitude of a sample of the corresponding ballast sensor signal with the largest absolute value during the event and to determine that the event is a significant rock fall event when the peak particle velocity PPV is greater than a magnitude threshold thresh_PPV
or a small rock fall event when the peak particle velocity PPV is less than the magnitude threshold thresh_PPV.

37. A system according to any one of claims 35 and 36 wherein, prior to determining that the event is a significant rock fall event, the signal processing unit is configured to ascertain whether the event occurs within a time window .DELTA.t pre-train of a subsequent train event and if the event does occur within the time window .DELTA.t pre-train to determine that the event is a train precursor event rather than a significant rock fall event.

38. A system according to any one of claims 35 to 37 wherein if the signal processing unit determines that the event is a small rock fall event, the signal processing unit is configured to ascertain whether the small rock fall event is one of a threshold number thresh_# of small rock fall events that have occurred within a recent time period .DELTA.T and if the event is one of the threshold number thresh_# of small rock fall events that have occurred within the recent time period .DELTA.T to determine that the event is part of a rock fall accumulation.

39. A system according to any one of claims 1 to 38 wherein the signal processing unit detects a plurality of rock fall events associated with a corresponding plurality of ballast sensors based on their corresponding plurality of ballast sensor signals and wherein the signal processing unit is further configured to estimate a location of a rock fall hypocenter based on fitting one or more parameters determined from the corresponding plurality of ballast sensor signals to a model representative of the spatial attenuation of an acoustic signal associated with the rock fall using a curve-fitting optimization technique.

40. A system according to claim 39 wherein the signal processing unit is configured to implement the curve-fitting optimization technique by:
looping through a plurality of potential hypocenter locations and for each of the plurality of potential hypocenter locations:
using the one or more parameters determined from the corresponding plurality of ballast sensor signals to determine model parameters that minimize a cost function associated with the model; and determining an error metric between one or more parameters predicted by the model and the one or more parameters determined from the corresponding plurality of ballast sensor signals; and after completing the loop, selecting a potential hypocenter location with a lowest error metric to be the estimated location of the rock fall hypocenter.

41. A system according to claim 40 wherein the cost function comprises a least squares cost function which comprises a sum of terms, each term comprising a squared difference between the one or more parameters predicted by the model for a particular ballast sensor and the one or more parameters determined from the ballast sensor signal corresponding to the particular ballast sensor.

42. A system according to any one of claims 39 to 41 wherein the one or more parameters determined from the corresponding plurality of ballast sensor signals comprise, for each ballast sensor signal, a peak particle velocity PPV parameter representing a magnitude of a sample of the ballast sensor signal with the largest absolute value during its associated rock fall event.

43. A system according to claim 42 wherein the model comprises at least one of:

where: y(x) represents an amplitude of the peak particle velocity PPV parameter at a distance x from the rock fall hypocenter, A is a model parameter representative of the peak particle velocity PPV at the hypocenter and B is an absorption coefficient model parameter.

44. A system according to claim 43 wherein the signal processing unit is configured to determine whether to trigger an alarm condition based on the model parameter A associated with the lowest error metric and representative of the peak particle velocity PPV at the estimated location of the rock fall hypocenter.

45. A system according to any one of claims 1 to 44 wherein the signal processing unit detects a plurality of events associated with a particular ballast sensor based on its corresponding ballast sensor signal and wherein the signal processing unit is further configured to determine whether the plurality of events is likely to be a human or other animal in a vicinity of the track section, wherein determining whether the plurality of events is likely to be a human or other animal in the vicinity of the track section comprises determining that a temporal correlation of the plurality of events is greater than a temporal correlation threshold temp_corr_thresh and determining that a spatial correlation of the plurality of events is greater than a spatial correlation threshold spat_corr thresh.

46. A method for detection of rock fall in a vicinity of a section of railway track, the method comprising:
providing a plurality of ballast sensors spaced apart along the track section and locating each ballast sensor in a ballast proximate to the track section but spaced apart from rails and ties associated with the track section, each ballast sensor sensitive to acoustic energy and configured to generate a corresponding ballast sensor signal in response to detecting acoustic energy;
receiving the ballast sensor signals from the plurality of ballast sensors; and processing the ballast sensor signals to detect rock fall events in a vicinity of the track section based on the ballast sensor signals.

47. A method according to claim 46 comprising detecting a plurality of different types of events comprising rock fall events, train events wherein a train travels over the track section and highrail vehicle events wherein a highrail vehicle travels over the track section and discriminating rock fall events from train events or highrail vehicle events.

48. A method according to claim 47 comprising detecting an event for a particular one of the ballast sensors based on its corresponding ballast sensor signal.

49. A method according to claim 48 comprising detecting a start of the event, for the particular one of the ballast sensors, when a short term average/long term average STA/LTA parameter associated with the corresponding ballast sensor signal is greater than a start trigger threshold thresh_start.

50.

associated with the start of the event.
A method according to claim 49 comprising detecting a time t start

51. A method according to claim 50 comprising detecting an end of the event, for the particular one of the ballast sensors, when the short term average/long term average STA/LTA parameter associated with the corresponding ballast sensor signal is less than an end trigger threshold thresh_end.

52. A method according to claim 51 comprising detecting a time t end associated with the end of the event.

53. A method according to any one of claims 49 to 52 comprising determine the short term average/long term average STA/LTA

parameter for the corresponding ballast sensor signal according to:

where: x i represents a value of an i th sample of the corresponding ballast sensor signal, n is an index of a current sample x n, a is a short term average STA duration constant, b is a long term average LTA duration constant and is the short term average/long term average STA/LTA parameter.

54. A method according to any one of claims 49 to 52 comprising determining the short term average/long term average STA/LTA
parameter for the corresponding ballast sensor signal according to:

where: x i represents a value of an i th sample of the corresponding ballast sensor signal, n is an index of a current sample x n, a is a short term average STA duration constant, c is an experimentally determined constant that is representative of a long term average LTA during event free times and is the short term average/long term average STA/LTA parameter.

55. A method according to claim 48 comprising detecting a start of the event, for the particular one of the ballast sensors, when an energy parameter associated with the corresponding ballast sensor signal is greater than a start trigger threshold E_thresh_start, the energy parameter comprising a windowed average of a squared amplitude of the corresponding ballast sensor signal.

56. A method according to claim 55 comprising detecting a time t start associated with the start of the event.

57. A method according to claim 56 comprising detecting an end of the event, for the particular one of the ballast sensors, when the energy parameter associated with the corresponding ballast sensor signal is less than an end trigger threshold E_thresh_end.

58. A method according to claim 57 comprising detecting a time tend associated with the end of the event.

59. A method according to any one of claims 55 to 58 comprising determining the energy parameter for the corresponding ballast sensor signal according to:

where: x i represents a value of an i th sample of the corresponding ballast sensor signal, n is an index of a current sample x n, d is a

60. A method according to any one of claims 52 and 58 comprising window duration constant and E n is the energy parameter.
determining a duration t dur associated with the event wherein t dur = t end -t start.cndot.

61. A method according to claim 48 comprising determining a duration t dur associated with the event for the particular one of the ballast sensors based on its corresponding ballast sensor signal.

62. A method according to any one of claims 60 and 61 comprising comparing the event duration t dur to one or more duration criteria and determining on the basis of this comparison that the event is not a rock fall event.

63. A method according to claim 62 wherein the one or more event duration criteria comprise a duration threshold thresh_dur and the method comprises determining that the event is not a rock fall event when the event duration t dur is greater than the duration threshold thresh_dur.

64. A method according to 63 comprising determining that the event is a train event or a highrail vehicle event when the event duration t dur is greater than the duration threshold thresh_dur.

65. A method according to claim 64 comprising determining, for the particular one of the ballast sensors, a peak particle velocity PPV
parameter which represents a magnitude of a sample of the corresponding ballast sensor signal with the largest absolute value during the event and comparing the peak particle velocity PPV
parameter to one or more magnitude criteria and determining on the basis of this comparison whether the event is a train event or a highrail vehicle event.

66. A method according to any one of claims 48 to 63 comprising determining a spectral power distribution associated with the event for the particular one of the ballast sensors based on its corresponding ballast sensor signal.

67. A method according to claim 66 wherein determining the spectral power distribution comprises computing a fast Fourier transform FFT of the corresponding ballast sensor signal.

68. A method according to any one of claims 66 to 67 comprising comparing the spectral power distribution with one or more spectral criteria and determining on the basis of this comparison that the event is not a rock fall event.

69. A method according to claim 68 wherein the one or more spectral criteria comprise a frequency threshold thresh_freq and the method comprises determining that the event is not a rock fall event when the spectral power distribution comprises more than a particular percentage of its power at frequencies above the frequency threshold thresh_freq.

70. A method according to 69 comprising determining that the event is a train event or a highrail vehicle event when the spectral power distribution comprises more than the particular percentage of its power at frequencies above the frequency threshold thresh_freq.

71. A method according to claim 70 comprising determining, for the particular one of the ballast sensors, a peak particle velocity PPV
parameter which represents a magnitude of a sample of the corresponding ballast sensor signal with the largest absolute value during the event, comparing the peak particle velocity PPV
parameter to one or more magnitude criteria and determining on the basis of this comparison whether the event is a train event or a highrail vehicle event.

72. A method according to any one of claims 48 to 71 comprising:

providing one or more rail sensors, each rail sensor operatively contacting the rails or the ties associated with the track section and each rail sensor sensitive to acoustic energy and configured to generate a corresponding rail sensor signal in response to detecting acoustic energy; and receiving the rail sensor signal.

73. A method according to claim 72 comprising determining a rail sensor spectral power distribution associated with the event for a particular one of the rail sensors based on its corresponding rail sensor signal.

74. A method according to claim 73 comprising comparing the rail sensor spectral power distribution to one or more spectral criteria and determining on the basis of this comparison that the event is not a rock fall event.

75. A method according to any one of claims 48 to 74 comprising determining a peak particle velocity PPV parameter associated with the event for the particular one of the ballast sensors and its corresponding ballast sensor signal, the peak particle velocity PPV parameter representing a magnitude of a sample of the corresponding ballast sensor signal with the largest absolute value during the event.

76. A method according to claim 75 comprising comparing the peak particle velocity PPV to one or more magnitude criteria and determining on the basis of this comparison that the event is not a significant rock fall event.

77. A method according to claim 76 wherein the one or more magnitude criteria comprise a magnitude threshold thresh_PPV

and the method comprises determining that the event is not a significant rock fall event when the peak particle velocity PPV is less than the magnitude threshold thresh_PPV.

78. A method according to any one of claims 48 to 77 comprising ascertaining whether the event occurs within a time window .delta.T pre-train of a subsequent train event and if the event does occur within the time window .delta.T pre-train determining that the event is a train precursor event.

79. A method according to any one of claims 48 to 78 comprising ascertaining whether the event is one of a threshold number thresh_# of events that have occurred within a recent time period .delta.T and if the event is one of the threshold number threh # of determining that the event is part of a rock fall accumulation.
events that have occurred within the recent time period .delta.T

80. A method according to claim 48 comprising:
determining a duration t dur associated with the event for the particular one of the ballast sensors based on its corresponding ballast sensor signal and determining that the event may be a rock fall event if the event duration t dur is less than a duration threshold thresh_dur; and determining a spectral power distribution associated with the event for the particular one of the ballast sensors based on its corresponding ballast sensor signal and determining that the event may be a rock fall event if the spectral power distribution comprises less than a particular percentage of its power at frequencies above a frequency threshold thresh_freq; and spectral power distribution indicate that the event may be a rock fall event, determining a peak particle velocity PPV parameter after determining that the event duration t dur and the associated with the event for the particular one of the ballast sensors and its corresponding ballast sensor signal, the peak particle velocity PPV parameter representing a magnitude of a sample of the corresponding ballast sensor signal with the largest absolute value during the event and determining that the event is a significant rock fall event when the peak particle velocity PPV
is greater than a magnitude threshold thresh PPV or a small rock fall event when the peak particle velocity PPV is less than the magnitude threshold thresh_PPV.

81. A method according to claim 48 comprising:
determining a duration t dur associated with the event for the particular one of the ballast sensors based on its corresponding ballast sensor signal and determining that the event may be a rock fall event if the event duration t dur is less than a duration threshold thresh_dur; and after determining that the event duration t dur indicates that the event may be a rock fall event, determining a peak particle velocity PPV parameter associated with the event for the particular one of the ballast sensors and its corresponding ballast sensor signal, the peak particle velocity PPV parameter representing a magnitude of a sample of the corresponding ballast sensor signal with the largest absolute value during the event and determining that the event is a significant rock fall event when the peak particle velocity PPV is greater than a magnitude threshold thresh_PPV or a small rock fall event when the peak particle velocity PPV is less than the magnitude threshold thresh_PPV.

82. A method according to any one of claims 80 and 81 comprising, prior to determining that the event is a significant rock fall event, ascertaining whether the event occurs within a time window .delta.t pre-train of a subsequent train event and if the event does occur within the time window .delta.t pre-train determining that the event is a train precursor event rather than a significant rock fall event.

83. A method according to any one of claims 80 to 82 comprising, if it is determined that the event is a small rock fall event, ascertaining whether the small rock fall event is one of a threshold number thresh_# of small rock fall events that have occurred within a recent time period .delta.T and if the event is one of the threshold number thresh_# of small rock fall events that have occurred within the recent time period .delta.T determining that the event is part of a rock fall accumulation.

84. A method according to any one of claims 46 to 83 comprising:
detecting a plurality of rock fall events associated with a corresponding plurality of ballast sensors based on their corresponding plurality of ballast sensor signals; and estimating a location of a rock fall hypocenter based on fitting one or more parameters determined from the corresponding plurality of ballast sensor signals to a model representative of spatial attenuation of an acoustic signal associated with the rock fall using a curve-fitting optimization technique.

85. A method according to claim 84 comprising implementing the curve-fitting optimization technique by:
looping through a plurality of potential hypocenter locations and for each of the plurality of potential hypocenter locations:
using the one or more parameters determined from the corresponding plurality of ballast sensor signals to determine model parameters that minimize a cost function associated with the model; and determining an error metric between one or more parameters predicted by the model and the one or more parameters determined from the corresponding plurality of ballast sensor signals; and after completing the loop, selecting a potential hypocenter location with a lowest error metric to be the estimated location of the rock fall hypocenter.

86. A method according to claim 85 wherein the cost function comprises a least squares cost function which comprises a sum of terms, each term comprising a squared difference between the one or more parameters predicted by the model for a particular ballast sensor and the one or more parameters determined from the ballast sensor signal corresponding to the particular ballast sensor.

87. A method according to any one of claims 84 to 86 wherein the one or more parameters determined from the corresponding plurality of ballast sensor signals comprise, for each ballast sensor signal, a peak particle velocity PPV parameter representing a magnitude of a sample of the ballast sensor signal with the largest absolute value during its associated rock fall event.

88. A method according to claim 87 wherein the model comprises at least one of:
y(x)= Ae-Bx ; and where: y(x) represents an amplitude of the peak particle velocity PPV parameter at a distance x from the rock fall hypocenter, A is a model parameter representative of the peak particle velocity PPV at the hypocenter and B is an absorption coefficient model parameter.

89. A method according to claim 88 comprising determining whether to trigger an alarm condition based on the model parameter A
associated with the lowest error metric and representative of the peak particle velocity PPV at the estimated location of the rock fall hypocenter.

90. A method according to any one of claims 46 to 89 comprising detecting a plurality of events associated with a particular ballast sensor based on its corresponding ballast sensor signal and determining whether the plurality of events is likely to be a human or other animal in a vicinity of the track section, wherein determining whether the plurality of events is likely to be a human or other animal in the vicinity of the track section comprises determining that a temporal correlation of the plurality of events is greater than a temporal correlation threshold temp_corr_thresh and determining that a spatial correlation of the plurality of events is greater than a spatial correlation threshold spat_corr_thresh.