WO2005095175A1

WO2005095175A1 - Train operating system

Info

Publication number: WO2005095175A1
Application number: PCT/GB2005/000596
Authority: WO
Inventors: John David Tunley; Stuart Duncan Brown; William Poole; Ian Hope Mitchell
Original assignee: Aea Technology Plc
Priority date: 2004-03-27
Filing date: 2005-02-18
Publication date: 2005-10-13
Also published as: EP1730009A1

Abstract

The effects of low adhesion on the operation of trains on a railway are minimised by obtaining data on a multiplicity of factors which affect the risk of low adhesion, and processing the data so as to generate predictions as to which sections of line have low adhesion. These predictions are supplied to a train-control signaling centre, which controls the movement of trains in accordance with the predictions. A warning message is also provided to the driver of any train that is about to enter a part of the railway predicted to have low adhesion. The sections of line predicted as having low adhesion may be displayed on a route diagram in the train-control signaling centre.

Description

Train Operating System

This invention relates to a methocd and an apparatus for managing the operation of trains so as to minimise the effects of low adhesion.

To operate a safe railway it is important to ensure there is sufficient adhesion between wheels and rails to sustain the required and expected traction and braking rates. The coefficient of friction between steel wheels and steel rails can vary over a very w±-de range, typically between 0.02 and 0.4, for example being at its highest with clean dry rails, and being low where there are damp leaves on the rails. The problems associated with low friction have been recognised for well over 100 years, and to a limited extent can be overcome by sanding, carried out by the driver of a train. Nevertheless, in low adhesion conditions, a driver has to drive more cautiously, and this causes delays which can effect other trains on the network. Delays can also be caused by wheelspin during traction, and by station overruns and signals passed at danger when the adhesion is insufficient for braking. According to the present invention there is provided a method of managing the operation of trains on a railway, the method comprising the steps of: a) obtaining data on a multiplicity of factors which affect the risk of low adhesion, at lea_st some of the data being real-time data, and incorporating geographical information; b) processing the data so as to g nerate predictions as to which geographical parts of the railway have low adhesion; c) supplying these predictions to a train-control signalling centre; d) the signalling centre rescheduling the movement of trains in accordance with the predictions so as to minimise the effect of low adhesion on operation of the railway; and e) transmitting data to each operating train such that the driver is warned when that train enters a geographical part of the railway predicted to have low adhesion.

It should be appreciated that the predictions that axe generated in respect of different parts of the railway may differ in their certainty. If such predictions are based on substantially real-time data from a train that is experiencing low adhesion, it will be appreciated that the corresponding predictions of low adhesion have a high degree of certainty.

Preferably the method comprises transmitting data to each train which includes the locations of all geographical parts of the railway predicted to have low adhesion. This warning message transmission system may also be used to provide the driver with warnings of other location-specific information, for example adverse weather conditions in a local area, temporary and permanent speed restrictions, sites identified as having degraded track quality, the expected location of track workers, and warnings of trespassers or vandals on or near the line. As regards speed restrictions, a warning is preferably displayed in the driver's cab indicating what the speed limit is; and a further notification is provided to the driver when the train has passed through trie speed restriction. Typically the signalling centre controls signals on the railway to achieve the rescheduled movement of the trains. Hence not only are the drivers warned of low adhesion areas, so they change their driving behaviour accordingly, but the trains are controlled in accordance with this changed driver behaviour. This ensures consistency between driver behaviour and the rescheduled timetable. By way of example the drivers may be trained to apply no more than 70% braking force, but to apply the brakes earlier, if the adhesion is predicted to be low.

Preferably the method also comprises displaying which geographical parts of the railway are predicted to have low adhesion on a route diagram in the train-control signalling centre. Alternatively or additionally such information may be displayed on a computer screen, and incorporate a drop-down menu of additional information about that part of the track, at least some of this additional information being the data on the factors which affect the risk of low adhesion.

Preferably the processing of the data to generate the predictions is carried out using a self-learning computer system, for example a neural network.

Preferably the processing generates predictions as to the values of adhesion, categorised into at least three categories: low, medium and high. However, it might make predictions in only two categories, or in more than three categories. As an extreme example, the predictions might predict the numerical value of the coefficient of friction.

The invention also provides an apparatus for performing such a method, that is to say an apparatus comprising: a) means for obtaining data on a multiplicity of factors which affect the risk of low adhesion, at least some of the data-providing means being sensors providing real-time data, and the data also incorporating geographical information; b) means to process the data so as to generate predictions as to which geographical parts of the railway have low adhesion; c) means to transmit these predictions to a train- control signalling centre; d) means in the signalling centre to reschedule the movement of trains in accordance with the predictions so as to minimise the effect of low adhesion on operation of the railway; and e) means to transmit data to each operating train such that the driver is warned when that train enters a geographical part of the railway predicted to have low adhesion.

The invention will now be further and more particularly described, by way of example only, and with reference to the accompanying drawings, in which:

Figure 1 shows a diagrammatic view of part of a railway network incorporating an apparatus of the present invention; and

Figure 2 shows a diagrammatic view of part of the apparatus of the present invention. Referring to Figure 1, part of a railway network is shown diagrammatically, incorporating railway lines 10 and railway stations 12. At some places the railway line passes over a bridge 14; at other places it passes through a cutting 16; and at other places 18 trees grow alongside the line. The number of tracks on each line are not indicated, and the junctions are not shown in any detail. A number of trains 20 are shown on the lines 10, and the movement of the trains 20 is controlled by signals (not shown) , operated from a signalling centre 22.

Adhesion between the wheels of the trains 20 and the rails is essential to safe operation of the railway, and the value of this adhesion at any particular position along a line depends on a number of parameters . For example it depends upon the presence of any moisture on the railhead; it is affected by rain and snow; and it depends upon the presence of any contaminants on the railhead, such as industrial contaminants or dead leaves. Furthermore, if a train 20 is found to slide or slip, this is an indication that that particular part of the line has low adhesion.

Referring also to figure 2, the signalling centre 22 incorporates a neural network computer 24 which is provided with data on these parameters. The computer 24 can receive data input from a keyboard 26, from remote sensors via one or more aerials 28, and from direct-wired sensors 30 (via electrical or optical cables) . In particular several moisture sensors 30 may be installed at different locations around the network, and data may be received' from each train 20 (via the aerial 28) if low adhesion is found as a result of the train 20 sliding or slipping. Preferably sensors are also provided to detect rainfall and other relevant weather parameters. Geographical data can be input into the computer by an operator, for example via the keyboard 26 (and/or a mouse) , firstly as to the routes followed by the lines 10, and also as to the location of geographical features which may affect adhesion such as cuttings 16 (where dead leaves may collect) , and places where trees 18 grow alongside the track. Preferably data is also input into the computer 24 about the density of vegetation near the track, the type of vegetation, and information on wind speeds and directions, and leaf fall. For example this may be input by an operator, or information may be provided by automatic sensors detecting parameters relating to leaves such as leaf colour.

These types of data can be considered as falling into three different categories: dynamic factors, which must be measured in real-time, such as weather conditions, observations of trains slipping, moisture on the railhead, leaf mobility etc; quasi-static factors which must be regularly reviewed and updated, such as tree density, type and size, the presence of overhanging branches, the distance of the vegetation from the track, etc, and also the presence of other sources of contamination such as dust from furnaces or quarries; and static factors such as altitude, track gradient, line speed, braking or traction site (e.g. approaching or departing from a station 12), and the nature of the track cross-section such as cuttings 16 or embankments. Data on such static factors may be provided from existing databases of geographical information; data on some of the dynamic factors may be obtained from external organisations, for example data on weather conditions from the Meteorological Office, and data 'on leaf fall from agricultural advisory bodies .

The computer 24 is arranged to assess all the data relevant to a particular section of track (for example a section of length 100 m) , and to determine whether that section is likely to have low adhesion, medium adhesion, or high adhesion. The computer 24 acts as a neural network, and is self-learning. Initially it is provided with such data on a historical basis, and learns from this historical data how best to predict the likely value of adhesion. The weights given to the different data by the neural network may be different for different sections of track. Subsequently, during use of the system, the computer 24 carries out this assessment for each different section of the network; these assessments may be carried out sequentially, or possibly a number may be carried out in parallel. Consequently the computer 24 produces output data 34 providing an indication of the likely adhesion level for each section of each line 10.

This output data 34 is supplied to a traffic scheduling computer 36 in the signalling centre 22. The computer 36 is associated with a display panel 38 for an operator, showing the location of trains 20 on the lines 10; the display panel 38 identifies, for example in different colours, those sections of the lines 10 which are predicted as having low or medium adhesion. The operator can therefore reschedule the traffic to take this information into account. In this example, however, the computer 36 reschedules all the movements of the trains 20 to optimise traffic flow, taking into account the predictions of low adhesion. For example this may entail letting through a train 20 on high adhesion, and restricting the speed of another train 20 on a low adhesion section of line. The computer 36 then operates the signals 40 accordingly. Alternatively the computer 36 might merely advise the operator, by proposing a new schedule; or might provide a new schedule to a separate route setting system (not shown) which operates the signals accordingly. The computer 36 preferably also displays the information about track adhesion on a computer screen for an adhesion manager, and each section of track identified as having low or medium adhesion may display further information (using a drop-down menu) about the factors which are relevant to the risk of reduced adhesion on that section.

In addition, the output data 34 is transmitted, via an aerial 42, to each train 20, the data being transmitted in such a form that in each train 20 a warning to the driver can be provided whenever the train 20 is about to enter a section of track where the adhesion is predicted to be low or medium. The way in which this is achieved will depend on the equipment available within the train 20. In one example, each train 20 incorporates a GPS instrument which determines its location, and the GPS instrument is provided (via the aerial 42) with the coordinates of the sections of track where the adhesion is low or medium. The GPS instrument provides a warning whenever the train 20 is within a preset distance along the line of such a section, this distance taking into account driver reaction time and an allowance for the adhesion conditions and their effect on a vehicle retardation. This may be referred to as a geofence. The GPS instrument may also be arranged to provide another warning when the train actually enters the track section of low adhesion. The instrument would also be arranged to notify the driver when the train has left the low adhesion section. Alternatively, if the location of each train 20 is known to the scheduling computer 36 (for example from track circuit data) , this computer 36 can transmit a warning signal to a train 20 that is approaching a track section of lower adhesion. Preferably the warning is given when the train 20 is far enough away to be able to brake, and so should take the train's speed into account. Consequently the driver can adjust the way he operates the brakes, or accelerates, to take this predicted value of adhesion into account. In a modification of the apparatus described above, there may be several such neural network computers 24 in lineside cabinets or other fixed locations, for example one every 2 km, each such computer 24 performing the adhesion predictions described above for the adjacent sections of track. Each computer 24 would then provide its output data 34 to the scheduling computer 36 at the signalling centre 22, either by radio transmission or through a cable. Rather than performing adhesion predictions for sections of track of length 100 m, such predictions may be carried out for sections a different length, preferably between 20 m and 5 km. Indeed it may be appropriate for these sections to be of different lengths, for example longer lengths of section would be appropriate where adhesion conditions are unlikely to vary (for example a long straight section of track across flat countryside with no trees) , and shorter lengths of section would be appropriate where circumstances vary from section to section. It is desirable also to transmit the output data 34 to mitigation trains, that is to say to dedicated rail- cleaning trains, so that the use of these trains to treat those sections which are predicted as having reduced adhesion can be improved. And clearly this output data 34 should also be sent to the workstations of track adhesion managers. The system may also provide a review facility for the autumn period, to allow managers to improve long-term planning. The computer 36 may also be used to supply data to the trains 20 to provide the locations of other track- related events, not relating specifically to adhesion, such as speed restriction information, and adverse weather events such as floods and snow. In the case of such speed limits, the train 20 preferably includes a display in the cab to show the driver the speed limit.

Claims

1. A method of managing the operation of trains on a railway, the method comprising the steps of: a) obtaining data on a multiplicity of factors which affect the risk of low adhesion, at least some of the data being real-time data, and incorporating geographical information; b) processing the data so as to generate predictions as to which geographical parts of the railway have low adhesion; c) supplying these predictions to a train-control signalling centre; d) the signalling centre rescheduling the movement of trains in accordance with the predictions so as to minimise the effect of this low adhesion on operation of the railway; and e) transmitting data to each operating train such that the driver is warned when that train enters a geographical part of the railway predicted to have low adhesion.

2. A method as claimed in claim 1 also comprising transmitting data to each train which includes the locations of all geographical parts of the railway predicted to have low adhesion.

3. A method as claimed in claim 1 or claim 2 also comprising transmitting data to each train to warn the driver of other location-specific information.

. A method as claimed in any one of the preceding claims also comprising displaying which geographical parts of the railway are predicted to have low adhesion on a route diagram in the train-control signalling centre .

5. A method as claimed in any one of the preceding claims also comprising displaying on a computer screen which geographical parts of the railway are predicted to have low adhesion, the display incorporating a drop-down menu of additional information about that part of the track, at least some of this additional information being the data on the factors which affect the risk of low adhesion.

6. A method as claimed in any one of the preceding claims wherein the processing of the data to generate the predictions is carried out using a self-learning computer system.

7. An apparatus for managing the operation of trains on a railway, the apparatus comprising: a) means for obtaining data on a multiplicity of factors which affect the risk of low adhesion, at least some of the data-providing means being sensors providing real-time data, and the data also incorporating geographical information; b) means to process the data so as to generate predictions as to which geographical parts of the railway have low adhesion; c) means to transmit these predictions to a train- control signalling centre; d) means in the signalling centre to reschedule the movement of trains in accordance with the predictions so as to minimise the effect of this low adhesion on operation of the railway; and e) means to transmit data to each operating train such that the driver is warned when that train enters a geographical part of the railway predicted to have low adhesion.