Method and apparatus to detect defects in a railway system

Abstract

There is disclosed a system for detecting conditions of a railway track. A plurality of sensors are provided each for mounting to railway rolling stock operating on the railway track, each comprising a sensing unit configured to measure vibration or acceleration data in each of a plurality of sampling intervals of the sensing unit, each sensing unit determing different properties in at least two sampling intervals baed on the measured data. A processor is provided, configured to append time and / or location data to the determined properties, corresponding to the time and / or location at which the measurements from which the properties were determined were made. A processor is configured to process the properties from each sensor, to identify railway track conditions at a location of the railway track based on an aggregate of determined properties from the plurality of sensing units at that location over time.

Description

[0001]DETECTING DEFECTS IN RAILWAY SYSTEM The invention relates to an apparatus, system and method for detecting defects in a railway system, including a railway track and a railway vehicle. A known railway monitoring system deploys sensors on train carriages. Data is measured at the sensors and processed on the train or transmitted off the train. Providing a small number of onboard sensors helps to limit the volume of data that needs to be processed on the train or to be transmitted off-train for processing. Typically a number of sensors are in communication with a central hub on the train, and data delivered to the central hub may be processed on or off-train. These hubs include expensive componentry, such as GPS units and mobile network wireless transmitters, and it is desirable to limit the number of hubs provided. Each sensor requires a certain bandwidth to transmit data to the hub, and overall the hub is limited to processing a maximum bandwidth, and the number of sensors in communication with each hub is limited. It is necessary to provide a small number of sensors for each hub. When data is to be transmitted off-train for processing, the size of the data that can be transmitted is limited. Even data from a small number of sensors can be excessive. On-board pre-processing facilities may be provided to process the data from the sensors to determine specific parameters from the sensor data, and then transmit only those specific parameters off-train. On-train pre-processing requires expensive on-board components and substantial on-board signal processing. It is an aim of the invention to address one or more of the above stated problems. There is provided an apparatus for mounting to railway rolling stock, comprising a single sensor configured to measure vibration or acceleration data in a plurality of sampling intervals to determine properties based on the measured acceleration or vibration data, wherein the single sensor is configured to determine different properties in different sampling intervals. Thus there is preferably provided a schedule, and a single sensor is configured to determine different properties in different sampling intervals according to the schedule. In practice, a single sensor provides different properties for different positions on a railway track, as the railway rolling stock to which the sensor is affixed moves to different positions on the railway track between successive sampling intervals. A plurality of single sensors act collectively across many vehicles to measure a large linear asset concurrently. This is different from a single sensor on a single vehicle that runs over the asset serially and has to be planned to do so. The collective sensor, being attached to all wheels (or most of them) measures the complete asset without individual sensor passes. Determined properties are aggregated over time, so that determined properties from multiple sensors (on a single or multiple rolling stock) build up an indication of track conditions at a given location, and a plot of track conditions over the whole railway track. The measured vibration or acceleration data or the determined properties may be used to monitor conditions of the railway rolling stock itself and / or a track system on which the railway rolling stock operates. This system is the complete system with which the rolling stock or train contacts. The apparatus may be for mounting to one or more of: a wheelset, a bogie, a bogie component (e.g. such as a bogie frame), a truck, a truck component, an electrical contact component (e.g. a pantograph or a hot shoe) and a vehicle body (e.g. to measure vehicle body movement) of railway rolling stock. Where the apparatus is the pantograph, this is preferably the pan head, which is the part contacting the overhead wire. A third rail component may be an implementation of the hot shoe. Sensors provided on the pantograph (such as the pan head, which is the part contacting the overhead wire) and the hot shoe (third rail contactor) may be provided with battery power. The track system may include the overhead line and / or the third rail, with which the pantograph and / or the hot shoe connect. The track may be considered as one type of railway asset. Railway linear assets include the track, the pantograph, and the hot shoe. A plurality of independent sensors may be connected to assets that are constrained to move around these linear assets, and many distributed and mobile sensing points are thus used to report on the condition of the complete asset. The measured acceleration or vibration data may be processed in one or more of: time domain (may additionally include temperature data) data, frequency domain data, statistical data to determine the properties. The sensor may further measure one or more of temperature data. The apparatus may include an accelerometer for measuring the acceleration data. The time domain data, frequency domain data, and statistical data may all be determined, and the single sensor may be configured to determine each in a different one of the plurality of sampling intervals. The single sensor may be configured to cycle through a set of property determinations in a set of sampling intervals, with different property types being determined in each sampling interval of the set. Thus a schedule is defined. The set of sampling intervals may be consecutive sampling intervals. In an alternative multiple consecutive sampling intervals measure a first value, and then other sampling intervals measure a second value. The set of sampling intervals may be repeated. Thus a schedule may define a sequence of intervals which are rotated. In an example implementation, a schedule comprises five successive sampling intervals, in which the following property type is successively determined: i) low pass filtered real time acceleration, ii) RMS and peak (statistical) data, iii) long sample length high resolution FFT, iv) low resolution short sample length FFT, and v) lateral and vertical statistical distribution of acceleration readings. The sensor may be configured to be enabled for a fraction of the sampling interval. The sensor is a single sensor of the apparatus. The apparatus may further comprise a transmitter operable for transmitting the determined properties in each sampling interval. The apparatus may further comprise an energy constrained power source for the sensor. The apparatus may further comprise a processor for controlling the sensor and constructing a packet comprising the determined property in each sampling interval for transmission. The determined property data in one sampling interval may be a short sample length fast Fourier transform with coarse frequency resolution. This permits corrugation to be monitored by a system, based on multiple measurements provided by multiple individual sensors associated with a particular location on the railway track. Each apparatus is preferably autonomous. There is provided a method for controlling a single sensor mounted to a railway asset, the method comprising measuring acceleration or vibration data in a plurality of sampling intervals, wherein the single sensor is configured to determine different properties in at least two sampling intervals. There is provided a system for detecting conditions of a railway track, comprising: a plurality of sensors each for mounting to railway rolling stock operating on the railway track, each comprising a sensing unit configured to measure vibration or acceleration data in each of a plurality of sampling intervals of the sensing unit, each sensing unit determining different properties in at least two sampling intervals; a processor configured to append time and / or location data to the determined properties, corresponding to the time and / or location at which the vibration or accleration data of the set was measured; and a processor configured to process the determined properties from each sensor, to identify railway track conditions at a location of the railway track based on an aggregate of determined properties from the plurality of sensing units at that location over time. Thus there is provided a large number of single sensors each providing a small amount of determined properties over a sampling period. From a system perspective an individual point on the railway track is assessed, by building up information over time based on the properties determined by lots of different sensors (multiple sensors on one train, and multiple trains) sending data packets at that location. The multiple sensors allow a system picture at a location to be built up, owing to data aggregation. The system may further comprise: a plurality of data concentrators, each configured to receive the determined properties in local packets from a plurality of sensors, and each including a processor configured to append time and / or location data to the determined properties in the packets; and a data analyser, including the processor configured to process the determined properties from each sensor, and configured to receive global packets each comprising a plurality of measured data and appended time and / or location data. The system may further comprise at least one temperature sensor mounted in proximity to at least one of the sensing units, wherein the processor is further configured to append temperature data to at least one set of determined properties, corresponding to the temperature at which the measurements from which the properties of the set were determined. Each sensor may be configured to cycle through a set of property types in a set of sampling intervals, with different property types being made in each sampling interval, wherein the processor configured to process the appended property data from each sensor is configured to aggregate property data of the same type at each of a plurality of locations over time. The system may further comprise at least one temperature sensor mounted in proximity to at least one of the sensing units, wherein the processor is further configured to append temperature data to at least one set of property data, corresponding to the temperature (of the rolling stock) at which the property data of the set was measured. The measured temperature may be used to monitor a condition of the rolling stock itself. The system preferably provides a time-evolving map of a railway system infrastructure, by using a ubiquitous sensing system, fitted to a fleet of rolling stocks for rolling stock maintenance purposes. Measured data – such as vibration data – may be compressed by limited sampling within a system sampling period. Widespread deployment of smart acceleration sensors – deployed at axle box, bogie and vehicle body mounted – delivers a flexible dataset for off-train processing, requiring only a low bandwidth (e.g. 20kbaud) per vehicle or railcar. Useful data from a ubiquitous sensor platform is measured and transmitted for processing, in such a way that a map of track condition can be generated without overwhelming either on- board data processing and transmission systems, or reducing the value of the condition information to specific indicative parameters that are insufficient to direct track maintenance actions. Track defects – and normal track – have been recognised to have a set of properties that can be described by a subset of time domain data, frequency domain data and statistical data. These properties can be determined and transmitted as pure or raw data with no specific track defect being considered. It is only after all the properties have been aggregated from all sensors that track property specific processing is done. There is provided a method for detecting conditions of a railway track, comprising: measuring a set of vibration or acceleration data at a plurality of sensors in a plurality of sampling intervals, the plurality of sensors mounted to railway rolling stock operating on the railway track, wherein different properties are determined by each sensor in at least two sampling intervals of each sensor based on the measured data; determining time and / or location data for each determined property, corresponding to the time and / or location at which the vibration or acceleration data was measured; and processing the properties from each sensor, to identify railway track conditions at a location of the railway track based on an aggregate of measured properties from the plurality of sensing units over time. The step of processing the properties may further comprise identifying railway rolling stock conditions. The determined properties in one sampling interval may be a short sample length fast Fourier transform with coarse frequency resolution, and processing that determined property indicates corrugation or rolling contact fatigue of the track. The method may further comprise configuring each sensor to cycle through a different type of property in a set of sampling intervals, wherein the processing step comprises aggregating the determined properties for each type at each of a plurality of locations over time. There is further provided a train having a plurality of sensors for detecting conditions of a railway track, each comprising a sensing unit configured to measure vibration or acceleration data in each of a plurality of sampling intervals of the sensing unit, each sensing unit determining different properties in at least two sampling intervals. The invention will now be described with reference to the attached figures, in which: FIG. 1 illustrates a train with a plurality of deployed sensors; FIG. 2 illustrates a schematic of an exemplary sensor; FIG. 3 illustrates exemplary operation of an exemplary sensor; FIG. 4 illustrates an exemplary data concentrator; FIG. 5 illustrates an exemplary remote data processor; FIGS. 6(a) to 6(c) illustrate exemplary operation of an exemplary system; FIG. 7 illustrates a portion of an exemplary map generated by the exemplary system; and FIG. 8 illustrates an exemplary method process for the exemplary system. With reference to FIG. 1, there is illustrated an exemplary system in the context of which examples are described. The illustrated exemplary system comprises multiple carriages 12 of a railway rolling stock, deployed on a railway track of a railway system. A train carriage may be a train engine, a passenger carriage or a freight carriage for example. Each carriage 12 is associated with two bogies 14. Each bogie 14 is associated with two axles, the wheels 16 of one side of each axle being shown. The carriages are connected. The exemplary rolling stock is provided with a plurality of sensors 20 deployed on each carriage. The sensors 20 may be located at a variety of locations, and in the example of FIG.1 it is shown that the sensors may be located on each wheel and on each bogie of each train carriage. In general discrete sensors may be attached or connected to one or more locations on railway rolling stock, including: a wheelset, axle boxes, bogies, bogie components, bogie frames, gearboxes, motors, bearings (e.g. where axles and wheels are connected to a bogie), trucks, truck components and vehicle bodies. Note, the terminology trucks and truck components is consistent in some territories with the terminology bogies and bogie components. This is not an exclusive list, and other points of connection / attachment are possible. The exemplary system includes a data concentrator, and one of carriages 12 is shown equipped with a data concentrator 18. As will be described further below, in examples the data concentrator 18 is able to wirelessly connect with a number of sensors 20. A sensor may communicate with more than one data concentrator, or be associated with a single data concentrator. The sensors 20 are configured to wirelessly transmit to the data concentrator 18 via short range radio in an example implementation. The number of data concentrators may vary, and one may be provided for several carriages, one may be provided per carriage, or more than one may be provided per carriage. Short-range radio is preferably used to communicate between the sensors 20 and the data concentrator 18, and the data concentrator 18 can communicate with any sensor 20 within a particular distance. The data concentrator may be limited to communicating only with a certain number of sensors – so as more sensors are provided, more data concentrators may be needed. In general sub-sets of the plurality of sensors are each in communication with a data concentrator. Each data concentrator 18 communicates with a central data processing system 60. The central data processing system is preferably a remote central data processing system provided off- train, such as a cloud-based or office-based data processing system. Communication between the data concentrators 18 and the remote data processing system 60 is preferably via a mobile communication network, with each data concentrator 18 being provided with a SIM for establishing a dedicated wireless data connection to the remote data processing system 60. Alternatively the off-board communication link may be via an ethernet connection, via WiFi, or via other train systems. In an alternative data processing may be provided on-train. The dataset accumulated from all sensors can be interrogated together at the remote data processing system 60 to derive a range of different track properties with sufficient specificity and accuracy to trigger maintenance actions, or to be used with other measurement systems to provide an accurate basis for maintenance actions. The dataset may be used to derive rolling stock properties, and to identify maintenance actions for rolling stock. This is discussed further below. In some implementation the data concentrator is not required, and the sensors may communicate directly with a processing system provided on or off the train. With reference to FIG. 2 there is illustrated an exemplary sensor apparatus 20, and there is described the operation of the exemplary sensor apparatus 20. The exemplary sensor apparatus 20 comprises a sensing unit 22, a short range wireless transmitter 24, an optional memory 26, and a sensor processor 28 each interconnected by a control bus 32 and a data bus 34. An energy constrained power source 32 provides power to each element of the sensor 20. The system of which the sensor apparatus 20 is a part is configured to have a system sampling period, with the sensor configured to have a sensor sampling interval which is a fraction of the system sampling period. For example, the system sampling period may be 60 seconds, and the sensor sampling interval may be 4 seconds. These are exemplary values. The sensor sampling interval includes the time required to capture sensor measurements and generate and transmit local packets. By limiting the sensor sampling interval to be a fraction of the system sampling period the total communication bandwidth of the sensor can be designed to be low, a typical baud value being less than 100baud (although higher baud values may be needed for certain applications). A practical example is 200 bytes per minute, ~2000bits / minute, over 60 seconds – which is close to 30baud (bits per second). 100 baud provides approximately 2 messages per minute, and is a satisfactory performance level. Another example implementation may provide 4 messages per minute, with 8 sensors per vehicle, and an average bit rate for the vehicle getting up to 1000baud. The sensor preferably samples at a random time during the system sampling period. Data acquisition and transmission may be synchronous or asynchronous. They may at some point get a real time clock from the data concentrator, but will still only transmit when they have enough energy. They might be forced to transmit on a particular regular fraction of a minute (15, 20, 30 seconds etc.) which will naturally synchronise readings. This will deliver a more systematic 100% coverage of the track, even if the data types are different. Under control of the sensor processor 28, each sensor 20 measures acceleration or vibration data with the sensing unit 22 in its sampling interval. Preferably the sensor does all the preprocessing, before sending packets to the data concentrator, the sensor comprising the power supply, processor, accelerometer and radio circuitry. The sensor is thus a smart transmitter, being a measurement device with processing. The measured acceleration or vibration data may be processed in one or more of the time domain data, the frequency domain data, or as statistical data in the same or different sampling intervals. Processing of the measured acceleration or vibration data is optimised, as discussed below, for rolling stock applications. In the time domain, measured acceleration or vibration data may be low pass filtered real time samples, on one or more simultaneous axes, or any combination of axes at a sample rate likely to detect long wave defects such as cyclic top or shorter discrete efforts. In the frequency domain (or frequency spectrum), acceleration or vibration data may be measured for a short sample interval at a low resolution FFT (fast Fourier transform), or for a longer sample length measured by a high resolution FFT. Low or high resolution may be 40Hz or 1Hz. Statistical data may determine the RMS and peak values of measured acceleration or vibration data. This may be statistical vibration information. The RMS may be measured for any specified bandwidth and for any combinations of axes. The peak values may be measured in any or all axes. The peak value may be the maximum value in any sample length, with or without additional location information. The peak values may be measured at specific bandwidths, e.g. 200Hz or 2048Hz. In embodiments, the measured acceleration or vibration data in each sampling interval is processed to determine properties or property data representing one or more of: i) low pass filtered real time acceleration, ii) RMS and peak (statistical) data, iii) long sample length high resolution FFT, iv) low resolution short sample length FFT, and v) lateral and vertical statistical distribution of acceleration readings. In general, the processing of the measured acceleration or vibration data may be chosen in order to detect particular railway track conditions, as discussed below with respect to the system of which the sensor forms a part, to determine performance of railway assets. The different properties determined from the data that is measured are chosen to be generally sensitive to changes in track condition – any change in track condition may cause change in multiple data types. The above data properties can be used to report track properties such as rail breaks, standard deviation (alignment), corrugation and other rail condition defects (e.g. RCF, squats, cracks or voiding). Most track defects change one or of the above measurements. In some cases, the change is in a defect specific fashion. Whilst advantageously measured data is processed for use to determine track conditions, many of the above outputs can also be used to advantageously describe a rolling stock condition, such as bearings or wheel conditions. The sensing unit 22 may be controlled by the processor 28 in order to cycle through a schedule for processing measured acceleration or vibration data, such that for a given cyclical sample interval of a sampling period, one or more particular properties are measured. Thus the sensing unit, being a smart transmitter, cycles through a processing schedule for determining properties. In an embodiment, the sensing unit 22 is controlled to determine the five types of properties identified as i) to v) above, with one of the five being determined in successive sampling intervals. The processing is repeated every five sample intervals or sample periods to define a schedule. With reference to FIG. 3 an exemplary operation of the sensor configured according to such a schedule is described. A single item of rolling stock 12, for example a railway carriage, having a sensor 20 attached to one wheel, is shown for ease of illustration. In practice rolling stock will comprises multiple carriages, and each carriage will have multiple sensors attached. The single item of rolling stock moves in distance over time along a railway track, such that the sensor 20 is at locations L1to L8at different instances. The locations may be locations associated with a section of track, corresponding to a length or distance, rather than a discrete geographical coordinate. FIG. 3 illustrates three cycles of a schedule, denoted Sch1, Sch2, and Sch3. Each cycle of the schedule is comprised of five sampling periods. The final two sampling periods Sp4and Sp5of schedule cycle Sch1, the five sampling periods of schedule cycle Sch2, and the first sampling period Sp1of schedule cycle Sch3are all shown. The five sampling periods Sp1to Sp5of schedule cycle Sch2each include a respective sampling interval Si1to Si5of the sensor 20. Also shown are the sampling intervals Si4and Si5of sampling periods Sp4and Sp5, and the sampling interval Si1of sampling period Sp1. Referring to schedule cycle Sch2, it can be seen that the sampling intervals Si1to Si5for the sensor 20, the sensor measures acceleration or vibration data to determine properties, and preferably transmits the determined properties. As described, the schedule cycles are arranged such that different properties are determined in each of sampling intervals Si1to Si5, such that over the sampling periods Sp1to Sp5of the schedule, five different properties are determined. As shown in FIG. 3, the sampling interval within each sampling period has a variable start time. The sampling interval of a sensor is random within each sampling period. In the illustrated embodiment, in sampling periods Si1, Si2, Si3, Si4, Si5, respective different properties or property data data A, B, C, D, E is measured, corresponding to the five properties identified as i) to v) above. Under control of the sensor processor 28, each sensor apparatus 20 creates local packets including the determined properties, and transmits the local packets from the short-range wireless transmitter 24. The local packets may be 200 bytes long. A variety of different data packets containing different determined properties may then be sent at regular or irregular intervals. The determined properties data may be transmitted as it is measured, and not stored at the sensor 20, and the memory 26 may be optionally provided for buffering purposes only. The properties data is preferably measured and transmitted when the train is moving. Each sensor may transmit local packets at a random time within the sampling period, associated with the random sampling interval. Each sensor may be provided with a frequency spreader to control the transmission of local packets. The time of transmission of local packets may depend on train activity. Each packet may comprise a header identifying the type of determined properties contained therein, and may be appended with time and / or location data. The location data may be derivable, in a subsequent processing step, based on the time the data was measured. These forms of data descriptor are low complexity and easy to measure / calculate in a low power smart sensor. In general a local packet comprises a set of determined properties data, typically property data of one type. Time and / or location information is appended to this set to form the local packet, with any appropriate header added. The header may include the address of a data concentrator with which the sensor apparatus is associated. Alternatively data concentrators may be configured to receive and process any broadcast local data packets which are detected. The on-board sensor data processing required for processing the measured acceleration or vibration data creating and transmitting local packets is simple and can be implemented with low power electronic devices. The sensing unit 22 may include any element required for measuring acceleration or vibration data In the illustrative example the sensing unit is shown as including an accelerometer 36, and a gyroscope 38, and a temperature sensor 37. The sensing unit 22 can therefore measure additional data, such as temperature data, and provide additional data. The data measured by the sensor is vibration data or acceleration data. The local packets may also be referred to as vibration packets, when the measured data is determined by a smart acceleration sensor. The fact that the sensors sample only for a fraction of the system sampling period and that the sensor is only required to be active for a fraction of the sampling period means that the sensors can be readily equipped with an energy constrained power source. The implementation of low power electronics devices for the required post-sampling processing at the sensor device is consistent with this. A preferable energy constrained power supply is an energy harvester, which converts ambient energy such as vibration energy into electrical energy for use to power electronic devices. A battery may also be used as the energy constrained power supply – due to the operation during a fraction of the system sampling period, a battery lifetime may be up to 5 years. The implementation of a sensor with an energy constrained power source and a wireless data transmitter allows it to be a stand-alone device which can be mechanically connected / attached without need for any wired connection. Preferably no modification / adaptation is required to the train vehicle. In alternatives sensors may also be equipped with other forms of power sources, and may have a wired power source. FIG. 1 shows a system is deployed utilising a data concentrator to communicate with multiple sensors. With reference to FIG. 4 there is illustrated an exemplary data concentrator 18, and there is described the operation of the exemplary data concentrator 18 in such a system. The data concentrator 18 may also be referred to as a data collector or a data aggregator. The exemplary data concentrator 18 comprises a short-range wireless receiver 40, a memory 42, a processor 44, a wireless transmitter 46 and an ethernet transmitter 47 interconnected by a control bus 50 and a data bus 52. A clock 30 and a GPS unit 31 are connected to the processor 44. A power source 48 provides power to each element of the data concentrator 18. A hard-wired connection 51 may provide an ethernet connection between the The short-range wireless receiver 40 of the data concentrator 18 receives the local packets transmitted by each sensor 20 within short range wireless range of the data concentrator 18. The processor 44 of the data concentrator 18 is configured to extract the properties and apply a time stamp from the clock 30 to the information from these local packets and store it in the memory 42. The processor 44 accesses this data from the memory 42 to create global packets which contain the properties from the multiple sensors from which the data concentrator 18 receives measurements. A header of each global packet may identify the type of properties contained therein. The processor may append the time of capture of the properties to the global packets. The processor may additionally append location information to the global packets – by comparing the time information to GPS information at the data concentrator. The processor 44 preferably transmits the global packets from the wireless transmitter 46 having a radio antenna on line 51, on a data connection of a mobile communications network. The wireless transmitter may have a combined UFH / GPS / LTE antenna. Global packets may be transmitted whenever a connection is available. Alternatively, the processor 44 transmits the global packets from the ethernet transmitter 47 having a wired connection 53, which connects to other on-board circuitry or to off-board circuitry, via the Internet or otherwise. The ethernet transmitter is an example of a hardwired connection from the data concentrator, but other hard-wired communication mechanisms may be provided. The data concentrator may also receive inputs from other sensors, such as in-cab accelerometers or gyroscopes built-in to the data concentrator itself. Other sensors may provide temperature information. In embodiments the data concentrator – or multiple data concentrators - communicates with a processor of a remote central processing unit. With reference to FIG. 5 there is illustrated an exemplary data processor 60, and there is described the operation of the exemplary data processor 60. The exemplary data processor 60 is a remote, off-train data processor. The exemplary data processor 60 comprises a wireless receiver 62, a memory 64, a processor 66, a display 68 and an API 70 all interconnected by a control bus 70 and a data bus 72. At the data processor 60, global data packets received at the wireless receiver 62 are stored in the memory 64 by the processor 66. The memory 64 thus stores a large amount of data representing determined properties from multiple sensors over time, together with the time at which the acceleration or vibration data on which that property is determined was measured. The stored properties may be used to determined performance of railway assets, as now described. The stored properties may be accessed by the processor 66 to extract the relevant properties to generate a map of the railway track, which may then be displayed on the display 68. This map may be used, possibly in conjunction with other outputs, to drive maintenance actions. With reference to FIGs. 6(a) to 6(c) an exemplary operation of the central processing provided by the system comprising multiple sensors is described. Referring to FIG. 6(a), a plot of distance against time shows that different railway carriages 12 having a single sensor (as per FIG. 3, for simplicity of illustration) are positioned at certain locations at different points in time. As shown, at location L3sensor 20 associated with railway carriages RS1 to RS5 are detected at different times t1to t5. As the example also shows, at location L4the sensor 20 associated with railway carriages RS6 to RS 10 are detected at the different times t1to t5. The locations may be sections / lengths of track, rather than a precise geographical location. As shown in FIG. 6(b), for the location L3, five sensor measurements are provided at times t1to t5from the sensors of the carriages RS1 to RS5. In practice, many more sensor measurements will be provided for location L3. As denoted in FIG. 6(b), at location L3the sensor of RS1 determines property data of type A, the sensor of RS2 determines property data of type C, the sensor of RS3 determines property data of type A, the sensor of RS4 determines property data of type D, and the sensor of RS5 determines property data of type A. As also denoted in FIG. 6(b), the sensor of RS6 determines property data of type C, the sensor of RS7 determines property data of type A, the sensor of RS8 determines property data of type B, the sensor of R94 determines property data of type A, and the sensor of RS10 determines property data of type B. The types A to D correspond to the types of properties according to FIG. 3. In the embodiment, consistent with the schedule of FIG. 3, in FIG. 6 multiple properties of types A to E will be determined at each location, although this is not shown in FIG. 6 for ease of illustration. A shown in FIG. 6(c), for each of the locations L3and L4an aggregate may then be compiled for each of the determined properties from multiple sensors, showing the variation of the determined properties at that location over time. In practice, multiple sensors will provide determined properties of each given category at various time instants. In this way an aggregate of each of the five determined properties, denoted AG1to AG5, can be compiled at each location at a given time instant and over time. The aggregates AG1to AG5are the aggregates for determined properties A to E at the respective location. The data processor 60 thus aggregates the packets received from the sensors by location and time, to generate a map of changing track conditions. The term ‘track’ includes rail, sleeper, ballast, sleeper pads or any other component of the system associated with the railway infrastructure. The data processor 60 thus uses the properties determined from measured acceleration or vibration data to determine performance of railway assets. Short packets are thus transmitted from many discrete sensors attached to many railway vehicle assets. As each sensor only samples and transmits for a fraction of the system sampling period, the total bandwidth of signals from multiple sensors is much less than that from a few sensors sampling continuously. The size and bandwidth of the local data packets transmitted by each sensor 20 is much reduced compared to a conventional sensor transmission, where the sensor sampling interval is the system sampling period. With reference to FIG. 7, there is illustrated a portion of an exemplary network map. The map shows a section of rail track 80, on which a particular location 82 is highlighted (which may, for example, be location L3). As denoted by waveform segments 84a to 84f six sensors (either different sensors, and / or one or more sensors at different times) have provided property data indicating unexpected vibrations at this location 82, identifying a potential track defect for inspection. For example a common strong frequency peak indicates a defect. With reference to Table I below there is illustrated a list of exemplary track defect types in column 1, and an identity of what determined properties are needed to identify that defect in column 4. Column 2 indicates how those track defect types are seen with prior art mapping systems, and column 3 indicates how these defects will be seen using the described techniques. Column five lists the associated advantage of the described techniques over the prior art. Track defect Prior art Map Identifies Detection Advantage type seen as mechanism, components Corrugation Elevated Defined Short Identify an RMS and corrugation sample accurate peak frequency, length FFT, location, values. location to with coarse corrugation Poor <10m. frequency type and location resolution, severity to accuracy giving a assist in (~40m). well planning. defined narrowband frequency and amplitude. Rolling contact Elevated Definitive Broadband, Accurate fatigue RMS on identification high location corners. of accurate frequency and location of random severity of affected rail. vibration, specific measured in RCF. a short sample length FFT. Localised Unusually High peak Aggregated Localised defects (squats, high, values, spatial / tim defects can breaks) localised consistent point e domain be an peak defect indicated data indicator values in real time showing a of existing packet, The lack localised rail cracks of specific defect. or frequency domain Combination conditions features of peak and leading to indicates a realtime a rail data. crack. localised defect. Stiffness Changes in Combination of Vibration Track the improved FFT caused by stiffness amplitude resolution and the is a key of the standard sleepers measurement frequency deviation. defining of track peak track support and caused by bending stability. sleeper under the Not periodicit wheels currently y. generates a measurable fixed by spatial conventiona frequency l mobile in the track wheels. measurement The machines. amplitude We have had of this a UK patent seems to for the depend on basic track method stiffness. only. Alignment / standa Increased Measurement of Short time This can be rd deviation RMS, no alignment / standa series of used to specific rd deviation measurement test track relationsh from time series s that can condition ip to low pass be used to directly standard filtered calculate against deviation acceleration standard track values in deviation standards. vertical and directly, lateral then direction. aggregated into a continuous value. Cyclic top Increased Direct Amplitude This can be RMS, no measurement in of cyclic used to specific the short time top can be test track relationsh series of low measured condition ip to pass from time directly standard acceleration series data against deviation data. directly, track then standards. aggregated to achieve a continuous value. Switches and Increased Identify RMS and Early crossing health. RMS and specific point realtime detection peak defects and high data should of values. vibration from differentia mechanical peak, RMS and te between problems time series a localised with data. Better fault switches clarity for caused by and vertical and excess crossing track may enable lateral defect bending (or earlier location. crack) and interventio increased n, general preventing vibration more of the serious crossing failures that might (such as indicate an broken underlying rail). problem leading to switch or crossing failure. Table I Track properties may be extracted as measured values with units or as condition indicators. Different types of stored data are used to extract different track properties either collectively or individually. Additional post-processing on historical data may be applied to extract new track condition values or indicators. Data aggregation is thus used to indicate the probability of a defined defect within a given length and location of track. Defects may include at least one of: rolling contact fatigue, squat defects, cyclic top, track stiffness or alignment, cracks or breaks twist, dip angles seasonal variation. The map may be made available to remote clients via a webpage or through the API 70. Changes in any reported track condition may be used to generate location specific alerts. With reference to FIG. 8, there are illustrated method steps according to an exemplary arrangement. In a step 90 the autonomous sensors are deployed across many railway rolling stock. In a step 92 the autonomous sensors are configured to measure one or more characteristics – specifically acceleration vibration data - in a sampling interval, which is a fraction of a system sampling period. This involves processing that measured data to determine a property or property data. This may involve further configuring a sensor according to a schedule, to determine different properties in different measuring intervals. In a step 94 the autonomous sensors transmit the determined properties in multiple short local packets. These transmissions may take place over a period of hours or days. In a step 96 the multiple short local packets may be received by the data concentrator. In a step 98 the data concentrator processes the received local packets, to apply time stamps and / or location information to the determined properties. The processed received data may be stored. In step 100 the data concentrator prepares global packets combining multiple sets of determined properties (and the associated time stamp and / or GPS data). In step 102 the data concentrator transmits the global packets. In step 104 the global packets are received from one or more data concentrators at the data processing centre. In step 106 the determined properties are processed and aggregated, based on time and / or location, at the data processing centre. In step 108 one or maps may be generated to provide information for defect diagnostics. Where a data concentrator is not provided, the local packets may be transmitted directly to a central processor. The time and / or location may be appended to the local packets by the processor of the sensor apparatus. The time and / or location may be appended to the global packets by the concentrator (which can record the time the local packet is received as the time the data was measured by the sensor). The time may be applied to the local packets and / or the global packets, and the central processing function may determine a location based on the time. A location may be added to the global packets by a processor at the data concentrator based on information provided by a GPS unit at the data concentrator. A processor at the central processing unit may process the measured data and any appended data to generate maps and / or providing diagnostic information. The sensors are described as measuring the acceleration or vibration data and processing this measured data to determine properties. In an alternative, the processing of this measured data to determine the properties may be done based on acceleration or vibration data contained in the local packets from the sensors. Any method described herein may be implemented as instructions of computer program code. A computer program product may store instructions of such computer program code. The computer program code may be stored, in whole or in part, on each of the sensor apparatus, and / or the data concentrator, and / or the central data processing unit. The computer program code may be executed, in whole or in part, on each of the sensor apparatus, and / or the data concentrator, and / or the central data processing unit. Thus a computer program code may be provided which, when executed on a processor, such as the processors illustrated in the examples above, may perform any method or process, at least in part. Various examples and embodiments have been set out to illustrate the invention. Aspects of examples and embodiments may be combined. The invention has been described by way of reference to various embodiments and implementations. The invention is not limited to the specifics of any example. The scope of protection afforded by the invention is defined by the appended claims.

Claims

CLAIMS 1. An apparatus for mounting to railway rolling stock, comprising a single sensor configured to measure vibration or acceleration data in a plurality of sampling intervals to determine properties based on the measured acceleration or vibration data, wherein the single sensor is configured to determine different properties in different sampling intervals.

2. The apparatus of claim 1 wherein the measured vibration or acceleration data is used to monitor conditions of the railway rolling stock itself and / or a track system on which the railway rolling stock operates.

3. The apparatus of claim 1 or claim 2 is for mounting to one or more of: a wheelset, a bogie, a bogie component, a truck, a truck component, an electrical contact component and a vehicle body of railway rolling stock.

4. The apparatus of claims 1 to 3 wherein the measured acceleration of vibration data is processed in one or more of a time domain data, a frequency domain data and / or as statistical data to determine the properties.

5. The apparatus of any one of claims 1 to 4 wherein the sensor further measures temperature data.

6. The apparatus of any one of claims 1 to 5 further comprising an accelerometer for measuring the acceleration data.

7. The apparatus of any one of claims 4 to 6 wherein time domain data, frequency domain data, and statistical data are all determined, and the single sensor is configured to determine each in a different one of the plurality of sampling intervals.

8. The apparatus of any one of claims 1 to 7 wherein the single sensor is configured to cycle through a set of property determinations in a set of sampling intervals, with different property types being determined in each sampling interval of the set.

9. The apparatus of claim 8 wherein the set of sampling intervals are consecutive sampling intervals.

10. The apparatus of claim 8 or claim 9 wherein the set of sampling intervals are repeated.

11. The apparatus of any one of claims 1 to 10 wherein the property type determined in each sampling interval is one of: i) low pass filtered real time acceleration; ii) RMS and peak (statistical) data; iii) long sample length high resolution FFT; iv) low resolution short sample length FFT; and v) lateral and vertical statistical distribution of acceleration readings.

12. The apparatus of claim 11 when dependent on any one of claims 8 to 10 wherein the single sensor is configured through five sampling intervals, with each of the defined property types being determined in one of the five intervals.

13. The apparatus of any one of claims 1 to 12 wherein the single sensor is configured to be enabled for a fraction of the sampling interval.

14. The apparatus of any one of claims 1 to 13 further comprising a transmitter operable for transmitting the determined property in each sampling interval.

15. The apparatus of any one of claims 1 to 14 further comprising an energy constrained power source for the single sensor.

16. The apparatus of any one of claims 1 to 14 further comprising a processor for controlling the single sensor and constructing a packet comprising the determined property in each sampling interval for transmission.

17. A method for controlling a single sensor mounted to a railway asset, the method comprising measuring vibration or acceleration data in a plurality of sampling intervals, determining different types of performance data in at least two sampling intervals.

18. A system for detecting conditions of a railway track, comprising: a plurality of sensors each for mounting to railway rolling stock operating on the railway track, each comprising a sensing unit configured to measure vibration or acceleration data in each of a plurality of sampling intervals of the sensing unit to determine different properties in at least two sampling intervals;a processor configured to append time and / or location data to the determined properties, corresponding to the time and / or location at which the vibration or acceleration data was measured; and a processor configured to process the determined properties from each sensor, to identify performance of a railway asset at a location of the railway track based on an aggregate of determined properties from the plurality of sensing units at that location over time.

19. The system according to claim 18 further comprising: a plurality of data concentrators, each configured to receive the determined properties in local packets from a plurality of sensors, and each including a processor configured to append time and / or location data to the determined properties in the packets; and a data analyser, including the processor, configured to process the determined properties from each sensor, and configured to receive global packets each comprising a plurality of determined properties and appended time and / or location data.

20. The system of claim 18 or 19 wherein each sensor is configured to cycle through a set of property determination types in a set of sampling intervals, with different property types being determined in each sampling interval, wherein the processor, configured to process the appended properties from each sensor, is configured to aggregate properties of the same type at each of a plurality of locations over time.

21. A method for detecting conditions of a railway track, comprising:measuring acceleration or vibration data at a plurality of sensors in a plurality of sampling intervals, the plurality of sensors mounted to railway rolling stock operating on the railway track, wherein different properties are determined by each sensor in at least two sampling intervals of each sensor; determining time and / or location data for each determined property, corresponding to the time and / or location at which the acceleration or vibration data was measured; and processing the performance data from each sensor, to identify railway track conditions at a location of the railway track based on an aggregate of the determined properties from the plurality of sensing units over time.

22. The method of claim 21 wherein the step of processing the determined properties further comprises identifying railway rolling stock conditions.

23. The method of claim 21 or claim 22 wherein the determined property in one sampling interval is a short sample length fast Fourier transform with coarse frequency resolution, and processing that determined property indicates corrugation or rolling contact fatigue of the track.

24. The method of any one of claims 21 to 23 further comprising configuring each sensor to cycle through a different type of property determination in a set of sampling intervals, wherein the processing step comprises aggregating the determined property determination for each type of property at each of a plurality of locations over time.

25. A train having a plurality of sensors for detecting conditions of a railway track, each comprising a sensing unitconfigured to determine properties based on acceleration or vibration data in each of a plurality of sampling intervals of the sensing unit, each sensing unit determining a different property in at least two sampling intervals.

Images