Research Programme. Engineering. ERTMS adhesion management: An assessment of the available adhesion and slip risk for ERTMS

Transcription

1 Research Programme Engineering ERTMS adhesion management: An assessment of the available adhesion and slip risk for ERTMS

2 R&D Programme: report commentary An assessment of the available adhesion and slip risk for ERTMS Research aims The objectives of this research were to ensure that the level of train protection is not reduced as a result of poor adhesion conditions, and that European Rail Traffic Management System (ERTMS) can remain fully operational throughout periods when a train is experiencing wheel slip or spin. The project systematically examined the potential changes that ERTMS brings to the management of adhesion risk and determined the actions required to ensure that ERTMS integrity is maintained. This report was commissioned under the Rail Safety and Standards Board (RSSB) Research Programme, and prepared by AEAT. It is one of five reports delivered for the project titled ERTMS and adhesion management. The five reports are listed below and are available from the RSSB web site. 1. Review of automatic train protection (ATP), automatic train operation (ATO) and ERTMS. 2. An assessment of the available adhesion and slip risk for ERTMS. 3. The feasibility of adhesion control and monitoring for ERTMS. 4. Review of adhesion dependent odometry. 5. An overview of low adhesion factors for ERTMS. Research findings This report quantifies the adhesion requirement for ERTMS in Britain and the risk of the levels required not being available. It also reviews equipment settings and determines the implications for levels 2 and 3 ERTMS on adhesion demand. In commissioning this workstream on behalf of the UK rail industry, it is clear that it is essential to obtain the buy in of all affected industry stakeholders and RSSB acknowledges and supports the work undertaken by the Strategic Rail Authority, and in particular the NEP team, in leading the adoption of ERTMS in the UK. The National ERTMS Programme team (NEP) assisted in scoping and delivering this research. The views of the NEP team and RSSB on the output of this research are given below. NEP response NEP accepts the report and recommends RSSB use the findings of this report as an input into their work to determine the braking models needed for the train braking data in the ERTMS onboard assembly. RSSB response It is critical for the UK to identify a strategy to ensure that the safety benefits of ERTMS are achieved, given that the increases in line capacity brought about by ERTMS will require the more consistent use of higher braking rates in locations not currently used for braking. Clearly, in developing such a strategy it is essential to obtain the buy in of all affected industry stakeholders. To achieve such buy in, it is necessary to involve all affected stakeholders and RSSB, where it can, will assist in the facilitation process. Existing industry stakeholder groups where they exist, for example the Adhesion Working Group, are ideally placed for taking the outputs from this research workstream forward. RSSB welcomes the NEP endorsement of the work carried out to date and the support for the publication of this report. RSSB supports the outputs from this phase of this research project and in particular, the conclusions drawn. The recommendations from the report provide a Page 1 of 2

3 R&D Programme: report commentary framework for specific elements of the UK implementation of ERTMS. RSSB recommends to those industry stakeholders charged with delivery of the next phases of the ERTMS implementation to adopt the recommendations and evaluate the effectiveness on risk mitigation and performance implications. In addition, RSSB will take forward the recommendations at a European level to raise awareness of the recommended UK approach. Contact Jim Lupton Head of Engineering Research Research and Development Programme Rail Safety and Standards Board jim.lupton@rssb.co.uk Page 2 of 2

4 An Assessment of the Available Adhesion and Slip Risk for ERTMS A report produced for Rail Safety and Standards Board. Authors: B. Rowell and W.S. Richardson, with contributions from S. Brown, S. King, W. Poole and J. Tunley June 2003 AEATR-VTI , Issue 1 iii June 2003

5 Title An Assessment of the Available Adhesion and Slip Risk for ERTMS. Customer Customer ref. Confidentiality, copyright and reproduction The Rail Safety and Standards Board RS Ref. 02-T080_AEAT-002-TRT This document has been prepared by AEA Technology Rail, part of AEA Technology plc, in connection with a contract to supply goods and/or services. File reference Report number LD43179 Rp02 AEATR-VTI Report status Issue 1 AEA Technology Rail 4, St Christopher s Way, Pride Park Derby DE24 8LY Telephone +44 (0) /1351 Facsimile +44 (0) AEA Technology is a business name of AEA Technology plc AEA Technology is certificated to BS EN ISO9001: 2000 Name Signature Date Authors Reviewed by B Rowell W S Richardson Bill Poole Approved by John Tunley AEATR-VTI , Issue 1 iv June 2003

6 Revision Record Issue Revisions Draft Issue for Comments 1 Final report Copyright 2004 Rail Safety and Standards Board This publication may be reproduced free of charge for research, private study or for internal circulation within an organisation. This is subject to it being reproduced and referenced accurately and not being used in a misleading context. The material must be acknowledged as the copyright of Rail Safety and Standards Board and the title of the publication specified accordingly. For any other use of the material please apply to RSSB's Head of Research and Development for permission. Any additional queries can be directed to research@rssb.co.uk. This publication can be accessed via the RSSB website AEATR-VTI , Issue 1 v June 2003

7 DISTRIBUTION LIST 1. David Meinck, Project Manager, The Rail Safety and Standards Board, Evergreen House, 160 Euston Road, London NW1 2DX - (5 bound copies + 1 Electronic copy) 2. B Rowell, AEA Technology Rail (bound copy) 3. W S Richardson, AEA Technology Rail (bound copy) 4. J Tunley, AEA Technology Rail (bound copy) 5. Vehicle Track Interaction Group File Copy Rachel Barker, 2nd Floor South Wing (unbound master) 6. AEA Technology Rail Reports Database Andrea Anderson, Ground Floor South Wing (1 unbound copy) 7. AEA Technology Rail Reports Database to Andrea Anderson (Executive Summary) 8. President of AEA Technology Cliff Perry, 3rd Floor North Wing (Executive Summary and Distribution List) 9. Senior Vice President of Engineering Bob Mole, 3rd Floor North Wing (Executive Summary and Distribution List) 10. Vice President of Railway Infrastructure & Systems Robin Smith, 1st Floor East Wing (Cover Page, Executive Summary, Distribution List, Approval Page, Introduction, Conclusions and Recommendations) 11. Vice President of Technology, NOBO Richard Gostling, 3rd Floor North Wing (Executive Summary only) 12. Business Development Manager, Trains Engineering & Management Paul Straw, 3rd Floor East Wing (Executive Summary only) 13. Business Manager, Railway Infrastructure & Systems Ben Blackwall, 2nd Floor South Wing (Executive Summary Only) AEATR-VTI , Issue 1 vi June 2003

8 EXECUTIVE SUMMARY European law has stipulated that the European Traffic Management System, ERTMS must be implemented for upgrades of all European high-speed railway routes. Consequently, the Rail Safety and Standards Board and the Strategic Rail Authority, SRA, established a project to manage the introduction of ERTMS to the U.K. This report covers the work carried out in Stage 2 of that project, Rserv254: ERTMS & Adhesion Management to establish the risk of wheelslip occurring (slip risk), at the ERTMS adhesion demand level. AEA Technology Rail has conducted a review of historic adhesion and SPAD data. From that review it is suggested that a combination of braking at the proposed ERTMS low adhesion brake rate of 6%g and an additional 200m margin in advance of the 6%g braking point (plus an allowance for gradient) is adopted for ERTMS low adhesion operation to reduce the risk to ALARP. The adoption of the above margin should eliminate more than 80% of all low adhesion SPAD s. The margin would also reduce the severity of the remaining SPAD s which generally are caused by severe low adhesion conditions. However, it is recommended that a wider range of adhesion control measures must be adopted to minimise the risk from such severe conditions. The action of the ERTMS low adhesion intervention algorithm will also minimise the effect of driver misjudgement in such conditions, further reducing the risk of a low adhesion SPAD occurring. It is also suggested that some of the above benefits could be achieved on a non-ertms railway by the adoption of low adhesion braking point signs placed at the 6%g braking point (plus 200m and allowance for gradient) at critical low adhesion signals. This would reduce the reliance on individual judgement of distances and braking performance required of every driver, which is bound to introduce some increased SPAD risk. Tests have shown that train borne sanders constructed to the Railway Group Standard offer around a 3%g improvement in deceleration in low adhesion conditions. A review of data from OTMR and LAWS fitted trains indicates that such levels of improvement are generally achieved in practice. Sander fitted trains would be at a reduced SPAD risk when operating under ERTMS AEATR-VTI , Issue 1 vii June 2003

9 low adhesion mode. Alternatively, the 200m margin could be reduced to 100m for sander fitted trains for the same level of SPAD risk. However, in that case, sander reliability would have to be reviewed since the sanding equipment may then be deemed safety critical. The ability of older analogue WSP design to make the best use of the available adhesion is limited compared with modern microprocessor designs which have been optimised on the AEA Technology Rail WSPER facility. Because of this and the odometric accuracy problems identified in previous work, the use of older WSP systems for ERTMS purpose is not recommended. This report has considered the ERTMS low adhesion brake rate and the extra margin required to minimise the SPAD risk. However, the actual intervention brake rate to be adopted when a train crosses the ERTMS low adhesion braking curve has briefly been identified as an issue which requires further consideration. A 9%g intervention rate offers some advantages at line speeds above 60 mile/hour. AEATR-VTI , Issue 1 viii June 2003

10 CONTENTS Description Page 1 INTRODUCTION PROJECT METHODOLOGY 1 2 ADHESION DEMAND IN BRAKING SIGNAL DISTANCES 4 3 AVAILABLE ADHESION AND SLIP RISK BACKGROUND NATURALLY OCCURRING LOW ADHESION PROFILE DATASETS THE EFFECT OF SPEED 13 4 THE EFFECT OF WSP OVERVIEW OF THE WSP SYSTEMS TESTED WSPER TEST RESULTS WSP SUMMARY 20 5 EFFECT OF SANDING SANDER TRACK TEST RESULTS IN-SERVICE PERFORMANCE SANDER SUMMARY 28 6 LOW ADHESION RISK 29 7 SPAD FREQUENCY FROM HISTORIC DATA 31 8 SPAD RISK SPAD RISK SUMMARY 35 9 THE INTERVENTION BRAKE RATE INTERVENTION BRAKE RATE AT 6%G INTERVENTION BRAKE RATE AT 9%G INTERVENTION BRAKE RATE AT 12%G LIMITATIONS OF THE ANALYSIS CONCLUSIONS RECOMMENDATIONS REFERENCES ACKNOWLEDGEMENTS 46 APPENDICES 44 AEATR-VTI , Issue 1 ix June 2003

11 1 Introduction European law has stipulated that the European Traffic Management System, ERTMS must be implemented for upgrades of all European high-speed railway routes (Ref. 1). Consequently, the Rail Safety and Standards Board (RSSB) and the Strategic Rail Authority, SRA, established a project to manage the introduction of ERTMS to the U.K. This report covers the work carried out in Stage 2 of that project, Rserv254: ERTMS & Adhesion Management to establish the risk of wheelslip occurring (slip risk), at the ERTMS adhesion demand level. Slip risk depends on the probability of the required adhesion being available and this has been established from a number of sources of information. The assessment of low adhesion slip risk provides the means to establish the measures that need to be taken on an ERTMS railway to reduce the low adhesion SPAD (Signals Passed at Danger) risk to ALARP. Due consideration is given to the aims of ERTMS in reducing SPAD's and increasing line capacity. 1.1 Project Methodology In assessing adhesion related slip risk, AEA Technology Rail has brought together information from a wide range of sources: reference to published material, discussions with signalling experts within AEA Technology Rail, members of relevant European committees, reference to historic BR Research data on adhesion and the risk of wheelslip occurring, Low Adhesion Warning System (LAWS) data, Autumn 2002 OTMR incident logs from train operators, SMIS database (SPAD and overrun) statistics The Wheel/Rail Adhesion Team members have extensive knowledge of low adhesion issues developed over the past 25 years from fundamental research and from consultancy work in this field. AEATR-VTI , Issue 1 1 June 2003

12 AEA Technology Rail are members of the industry sponsored Adhesion Working Group (AWG) and this source provides up-to-date information of adhesion issues within the industry as well as useful contacts with low adhesion operational experience. Nominated RSSB staffs have provided information for the project and representatives from the RSSB and the AWG have reviewed the draft report and their comments have been incorporated in the final issue. The work presented in this report forms part of a larger project targeted at addressing the adhesion issues relevant to the implementation of ERTMS. It is intended that a summary report (Ref. 2) will be produced that will bring together the related elements of these various work streams so that overall recommendations can be made to allow safe ERTMS operation in low adhesion operating conditions. AEATR-VTI , Issue 1 2 June 2003

13 2 Adhesion Demand in Braking The adhesion demand in braking is the level of adhesion between the wheel and rail that must exist to allow the required brake rate to be achieved. The normal full service brake rate in the UK is 9%g, which equates to an adhesion demand of approximately The ERTMS specification proposes that the braking rate in low adhesion conditions should be 70% of the normal braking rate (Ref. 3). Hence, for UK operation, the low adhesion ERTMS mode would require a brake rate of 6.3%g. A method of switching between normal and low adhesion ERTMS operational modes has been proposed in a related report for the Rail Safety and Standards Board (Ref. 4). Under current UK signalling practices, when braking in low adhesion conditions, 'defensive' or 'professional' driving techniques are adopted. A large proportion of passenger rolling stock is fitted with a 3-step brake where the nominal brake rates are approximately 3%g, 6%g and 9%g for brake steps 1, 2 and 3 respectively. Low adhesion driving practice encourages the use of brake steps 1 and 2 to reduce the risk of wheel slip. Hence, a maximum braking rate of 6%g (step 2) is normally utilised. Applying the above to the proposed ERTMS low adhesion brake rate, it is reasonable to select a nominal 6%g brake demand level for ERTMS operation in the UK. This is then consistent with existing 'best practice' for low adhesion operation and coincides with a brake rate that is nominally available on a wide range of existing UK passenger stock. NB Many vehicles are over-braked to some degree. It is quite possible that a 'nominal' 6%g brake demand may well be 7%g in reality. The brake rates of such vehicles are often very difficult to adjust because a reduction of one braking rate will apply an equivalent reduction throughout the range. It is possible that the brake step 1 rate will then be marginal. It follows that any detailed adjustment of the ERTMS adhesion demand level could not be replicated in service on a range of vehicle types. Thus the above 'rounding down' of the ERTMS brake demand from 6.3%g to 6%g is not a critical factor given the wide spread of achievable service brake rates and the range of low adhesion conditions encountered. It is also important to note that there may be a difference between the ERTMS on-train intervention curve at 6%g deceleration and the actual AEATR-VTI , Issue 1 3 June 2003

14 demanded deceleration, should an intervention occur. This report assumes that the intervention curve and brake demand rates are equal. However, the implication of a higher brake demand rate is briefly discussed in Section Signal Distances Braking Distance The ERTMS low adhesion brake demand rate of 6%g has a significant effect on the braking distance, assuming standard lineside signal distances and that the driver approaches the signal in ERTMS low adhesion mode at full line speed. As the distance for braking in low adhesion mode is longer than that of the normal signal length, the point at which the driver applies his brakes will have to be earlier, or alternatively, the maximum speed will have to be reduced. This is illustrated below for 100 mile/hour line speed. Distance of Brake stop with 100 miles per hour Line speed given 6%g deceleration miles/hour standard signal spacing Initial Speed (miles per hour) miles/hour 90 miles/hour Distance (m) Thus the approach speed must be limited to just below 90 mile/hour or alternatively it is necessary to apply the vehicle brake some 340m earlier than is required for normal full service braking. AEATR-VTI , Issue 1 4 June 2003

15 2.1.2 Signal spacing and its effects on ERTMS The signal distance for stopping with a 9%g brake rate includes an allowance for low adhesion and speedometer error. This allowance is currently set at a minimum of 20% of the 9%g stopping distance for level track (Ref. 5). The effect of introducing a 6%g brake demand rate for the ERTMS low adhesion mode would increase the stopping distance by 50% over normal full service braking. For the ERTMS low adhesion mode, the braking point would have to be earlier than the first cautionary aspect: The figure above shows the difference in stopping distance at 6%g brake rate as opposed to the usual 9%g full service brake rate. These are all in relation to the minimum stopping distance available from fixed line-side signalling for different line speeds. The extra braking distance required is far greater at higher line speeds than at lower line speeds as shown in the table below. AEATR-VTI , Issue 1 5 June 2003

16 Entry Speed (Mph) 9%g distance (m) 6%g distance (m) Extra distance (m) If the ERTMS low adhesion mode is adopted, the option for braking at 6%g would be incorporated into the train-borne ERTMS controller by including an intervention curve for low adhesion conditions. This will automatically provide for application of the brakes at the earlier low adhesion braking point, taking into account the extra distance required Comparison with existing lineside signal practice Currently drivers make an allowance for the extra distance required to stop based on low adhesion training, route experience and a subjective assessment of the prevailing low adhesion conditions. In addition, actual signal distances can vary from one to one and a half times the minimum distance, and signal sighting distances will also vary. The wide range of variables that the driver has to consider from day to day may contribute to the risk of a SPAD or overrun. Although outside of the scope of this work, the above review suggests that, for a non ERTMS railway, an advisory low adhesion braking point board, positioned for the defensive driving 6%g braking rate and including a small margin, could reduce SPADs at critical low adhesion signals. AEATR-VTI , Issue 1 6 June 2003

17 3 Available Adhesion and Slip Risk 3.1 Background Adhesion between the wheel and rail is mainly dependent upon the amount and type of contaminant on the railhead and on the amount of moisture present. Various contaminants are known to cause problems, in particular; light rust, grease and oil, salt, industrial fall-out and, of course, leaves which are drawn into the wheel/rail nip by the aerodynamic effects of the train passage. Any of these contaminants will lower adhesion but the addition of small amounts of moisture (dew, mist or light rain) acting as a lubricant will reduce the adhesion much further. For example, the best adhesion between a clean wheel and rail may be as high as 0.4, dry leaf film may reduce the adhesion to 0.1 and damp leaf film, in the worst case conditions, may further reduce the available adhesion to Large amounts of water, i.e. heavy rain, tend to clean the railhead and generally increase the adhesion towards the dry rail level. From the above it follows that adhesion can vary on different sections of track according to the contaminants present and can also vary with time according to the prevailing weather conditions. This variation is best described in statistical terms. An adhesion survey carried out in the early 1970 s with the BR Research Tribometer Train (Ref. 6) showed that the average adhesion on BR was 0.23 with a standard deviation of 0.05 (NB a correction factor has been applied to these figures, based on more recent unpublished studies). A normal distribution of the data was assumed, although detailed analysis was limited at that time by the computing power available. From these figures it can be calculated from standard statistical tables that the average adhesion drops below the normal full service adhesion demand of 0.09 for around 0.3% of the time. Expressed in another way, 3 in 1000 braking stops, averaged over 12 months, may result in some wheelslip. However, many of these low adhesion sections will be very short such that the extension in total braking distance will be small. In that case the extension will be accommodated by the 20% allowance incorporated in the signalling distance for such eventualities. Unfortunately, from the numbers of signals passed at danger (SPAD s) and station overruns that occur during the autumn low adhesion period, it is apparent that occasional longer lengths of low adhesion do occur which are sufficient to cause a significant risk to railway operation. The analysis of above AEATR-VTI , Issue 1 7 June 2003

18 adhesion survey results did not consider this aspect of the data. Each measurement was treated as an individual event that was part of a total adhesion population. The relationship between adjacent measurements, which would describe the profile of adhesion along the track, was not considered. However, low adhesion profiles obtained in the early 1990 s addressed these issues. 3.2 Naturally occurring Low Adhesion Profile Datasets Data collected by the BR Research Tribometer Train in the autumns of 1992 and 1993 provides statistical information on the probability of meeting different lengths of low adhesion. The Tribometer train operated as follows: the circuit consisted of Basingstoke-Eastleigh-Salisbury-Basingstoke, data was recorded over a total of 8 weeks, 5 nights a week through mid autumn 1992 and 1993, this circuit was completed twice on most nights and was 143km in length, the train measured the adhesion at 20mile/hour at all times, the measuring axle of the train was disc braked so that it did not modify the base track adhesion, all data collected was under low adhesion conditions. A large amount of data was discarded, as it did not show adhesion at a low enough level to be deemed useful for this study. This left 57.7 km of low adhesion data. For the purpose of this analysis, it is assumed that the Tribometer Train recorded for 25% of the total route distance. The outcome of these tests were adhesion profiles that show the peak adhesion available over a length of low adhesion track. This data is used in WSPER testing (see section 4) to assess the performance of WSP systems in real low adhesion conditions. The graph below shows an example of one of these profiles. AEATR-VTI , Issue 1 8 June 2003

19 AEATR-VTI , Issue 1 9 June 2003

20 The adhesion profiles were analysed to obtain the frequency of lengths of low adhesion below the 0.02, 0.03, 0.04, 0.05, 0.06 and 0.09 adhesion levels. The results of this analysis are tabulated below. Adhesion Level Length of low adhesion (m) Frequency of Occurrence. < < < < < < NB: Adhesion level frequencies are cumulative, e.g. all events for 0.09 adhesion level incorporate all events at lower adhesion levels. AEATR-VTI , Issue 1 10 June 2003

21 The data from the above table is shown below as a frequency histogram. 225 Histogram showing the frequency of occurrence of different lengths of low adhesion from the WSPER Data Sets Frequency of occurance <9%g <6%g <5%g <4%g <3%g <2%g Distance of low adhesion (m) This graph indicates that: The majority of low adhesion sections measured were less than 200m long, The frequency of low adhesion sections below 0.06 is around one third of those at 0.09, There are no low adhesion lengths longer than 400m where the adhesion is 0.04 or below. AEATR-VTI , Issue 1 11 June 2003

22 3.2.1 ERTMS and Slip Risk Probability To determine the ERTMS slip risk probability according to both the level of adhesion and the length of that low adhesion, the following procedure was employed: 1. The length of track measured over the 8 weeks of testing was calculated using the known route and frequency of testing given above. This equates to approximately 11,000km since there were some occasions when the train did not run. 2. The cumulative number of events in each adhesion band was divided by the total track distance. The resultant probability is that of meeting more than a defined minimum length of low adhesion following a brake application on any metre of track. The results are tabulated below: Adhesion Level Length of low adhesion greater than:(m) Probability of Occurrence. x AEATR-VTI , Issue 1 12 June 2003

23 3.3 The Effect of Speed The above adhesion data was measured generally at 20mph (32km/hr). However, the available adhesion falls with increasing speed. This effect has been reported by Office for Research and Experiments of the International Union of Railways (ORE) (Ref. 7). They suggest that the level of adhesion reduction with increasing speed is a function of the vehicle suspension and the dynamic interactions between the wheel and the rail. Subsequent unpublished analysis of the ORE test data produced a family of 'best-fit' curves according to the contaminant applied to the railhead: Mean Adhesion on UK Railways at 32 km/hr Dry Rail (UIC) Damp Rail (UIC) Damp Leaf Film (UIC) Adhesion Wet Rail (Japan) Speed (km/hr) The Variation of Adhesion with Speed. It must be emphasised that the above analysis is very approximate, being based on very limited amounts of data. However, also superimposed on the above graph is data taken from a published Japanese paper (Ref. 8) for 'wet rail'. This data is surprisingly close to the 'damp rail' curve from the ORE data. This independent corroboration provides a useful validation of the ORE results. From the above ORE data and from the distribution of adhesion on BR track measured at low speed, it is possible to predict the slip risk at different speeds, shown here graphically with a logarithmic vertical scale: AEATR-VTI , Issue 1 13 June 2003

24 %g %g 12%g % Slip Risk Speed (km/hour) Increase in Slip Risk with Speed From the above: At 100km/hour there is a ten-fold increase in slip risk when braking at 9%g compared with 6%g. Similarly, there is a ten-fold increase in slip risk when braking at 6% at 200 km/hour compared with 100 km/hour. The enhanced emergency brake rate of 12%g has around a 50 times greater slip risk than that incurred by demanding just 6%g. This increased slip risk also applies to dynamically braked stock where half the axles are motored such that a 12%g adhesion demand is required to achieve a 6%g braking rate. These above results are likely to be pessimistic because, in the majority of instances, the wheel slip will occur over a very short distance in practice, which may not substantially affect the overall braking performance. However, the results do indicate the very significant relative increase in slip risk that occurs due to increased operating speeds and brake demands. The above does not define the effect of speed on the length of low adhesion encountered, although it is likely that the effective lengths will increase at higher speeds. However, because there is no direct known correlation AEATR-VTI , Issue 1 14 June 2003

25 between speed and length of low adhesion, the effect of speed has not been included in this analysis of ERTMS SPAD risk. 4 The Effect of WSP Modern WSP, which in the UK is usually optimised on the AEA Technology Rail WSPER facility (Ref 9, Appendix A), is designed to minimise the braking distance in low adhesion conditions. By reference to WSPER data from various tests, the likely effect of variations in WSP performance on braking distance has been quantified, and the effect on the achieved deceleration compared with the ERTMS demanded levels. The WSP performance of both a modern WSPER optimised microprocessor based WSP and an older analogue WSP design have been compared on the WSPER. The results from these tests have enabled AEA Technology Rail to quantify the braking performance improvements gained through installing a modern WSP system. 4.1 Overview of the WSP Systems Tested BR MkII Deep Slide The BR MkII WSP systems are based on analogue electronics and the Deep Slide variant is the most common, being fitted to around 2,500 vehicles in the UK. The control philosophy used is relatively rudimentary and it works by: measuring the rate of deceleration of individual axles comparing relative axle rotational speeds The system attempts to control wheel slip at around the 60% slip level, but it does not maintain an accurate reference speed under very low adhesion conditions. This results in a high probability of all four wheelsets locking up at high speed and causing significant damage to wheelsets. Poor reliability compounds this poor inherent performance resulting in further damage to wheelsets. In addition, the system is difficult to set up and maintain as (amongst other considerations) the gap between the speed probe and associated toothed wheel must be set accurately and checked on a regular basis. The system performance is also limited by the use of single stage dump AEATR-VTI , Issue 1 15 June 2003

26 valves that cannot hold a constant brake pressure at a lower level than the nominal brake demand pressure. AEATR-VTI , Issue 1 16 June 2003

27 4.1.2 Modern Microprocessor Based Systems Modern WSP systems are generally based on microprocessor technology and can generally provide levels of protection superior to that available from the older systems provided they are optimised for the vehicle to which they are fitted. The control philosophies used are far more developed than the older BR systems and significant improvements have been achieved in keeping track of the true train speed under all operating conditions. In addition, modern systems have two stage dump valves which allow the pressure in the brake cylinders to be held at a reduced level. This has two advantages: air consumption is greatly reduced a greater degree of control can be exerted over the wheelset behaviour resulting in superior braking performance The systems are generally more reliable and they incorporate on-board fault diagnostics which reduces the maintenance costs and the risk of damage to wheelsets due to a malfunction of the WSP system. The problems associated with gap between the speed probe and associated toothed wheel has been reduced because modern systems are less sensitive to the gap. 4.2 WSPER Test Results The full WSPER test programme for a particular vehicle consists of 160 low adhesion braking stops for a combination of vehicle load, brake rate, wheel size, brake entry speed and adhesion profile. In the case of the 4 car EMU tested, two vehicle types were used, both Power car and Trailer car, leading in this case to a total of 320 brake stops. A typical selection of results from a full WSPER test programme is shown in the Table below. The performance of the WSP under test is shown here over 10 Naturally Occurring Variable Adhesion (NOVA) profiles that form the basis for of the basic WSPER test. AEATR-VTI , Issue 1 17 June 2003

28 Vehicle Type Vehicle Load Brake Rate Wheel Size Test No Speed (mph) NOVA Profile Distance (m) Power Car Crush Full Service Worn rz pz pr pd re pp pp pe px pe Average across the 10 test profiles: m Selection of WSPER test results The complete results from the full test sequences of the BR "Deep Slide" WSP and a modern WSP system, were collected and a value for both the average braking distance and average deceleration determined. This average included all possible test conditions (e.g. vehicle load, brake rate, wheel size and adhesion profile) for both the power car and trailer car, when braked from high and low speed. The results are shown in the table below: Vehicle Type / Line Speed BR Deep Slide WSP Modern WSP POWER, Average braking distance all tests (m): m/s Average deceleration all tests (%g): TRAILER, Average braking distance all tests (m): m/s Average deceleration all tests (%g): WSPER braking distance and deceleration results. It can be seen that the performance of the modern WSP is an improvement over the BR "Deep Slide" WSP since in all cases, for the same vehicle condition and entry speed, the overall braking distance is significantly reduced. AEATR-VTI , Issue 1 18 June 2003

29 The proposed ERTMS low adhesion mode brake demand is nominally 6%g. The WSP systems tested here, and in particular the modern WSP taken to be representative of a well optimised system, do not achieve an average rate of deceleration at or above this level. This is because in the majority of WSPER tests the average level of adhesion on the naturally occurring track profiles is not high enough to support this adhesion demand. The average rate of deceleration achieved by the modern WSP system in low adhesion conditions over the 320 WSPER tests was 5.43%g. This value is approximately 10% lower than that of the ERTMS low adhesion braking demand of 6 %g, as proposed in this report. However, it is important to note that WSPER tests are for a single vehicle and the performance of a multiple unit is likely to be better due to the wheel/rail conditioning effect along the length of the train. Additional braking distance due to sections of low adhesion - train speed = 100 mph Length of low adhesion encountered (m) Old WSP Modern WSP Additional braking distance (m) The figure above shows the resultant extension made to an ERTMS low adhesion mode brake stop when a given length of low adhesion is met. The braking rate over the section of low adhesion is reduced to the respective average deceleration rate achieved by the WSP over all of the WSPER profiles. Hence, for the BR Deep Slide WSP an average deceleration of 4.76%g is achieved, 0.65%g worse than that achieved by the modern WSP. AEATR-VTI , Issue 1 19 June 2003

30 The modern WSP is able to follow the peak of adhesion more closely than older WSP systems and can therefore make the best use of the available adhesion. This results in an improved stopping distance, since the average rate of deceleration over the entire stop is higher, even on low adhesion sections. Because these results are the average of a range of low adhesion conditions, it should be noted that, under the worst case adhesion conditions experienced on UK track, the performance benefit of the modern WSP will be proportionately greater than the results presented here. 4.3 WSP Summary Modern WSP systems are significantly better at keeping an accurate estimate of the true train speed under all operating conditions. In addition, these modern WSP systems use two stage dump valves that allow the pressure in the brake cylinders to be held at reduced levels. This has the advantage of allowing a greater accuracy of control over the wheelset behaviour. Both of these factors mean that modern WSP systems are able to make best use of the available adhesion, resulting in superior braking performance. For the purpose of this study it is assumed that a modern WSP will be able to achieve the required adhesion levels by operating close to the peak adhesion levels, making best use of the available adhesion by conditioning the wheel and rail. If stock operating under ERTMS is fitted with an older WSP system the SPAD risk will increase slightly due to an average reduction in deceleration in low adhesion conditions of 0.65%g. This factor, together with the odometry problems identified in Ref. 9, make the use of older WSP equipment less viable for ERTMS operation. AEATR-VTI , Issue 1 20 June 2003

31 5 Effect of Sanding For effective operation of the ERTMS during braking it is necessary to consider the impact of sanding on braking under low adhesion conditions. This has been achieved by consideration of track test results and from inservice sander performance data. 5.1 Sander Track Test Results The effectiveness of a train borne sander in improving the braking performance of the train that lays the sand is dependent upon a number of factors, including: The initial adhesion the base level of wheel on rail adhesion seen by the leading wheelset. The amount of sand per metre applied to the railhead The length of low adhesion track that is actually treated by sand application The length of the train the wheel and rail cleaning (conditioning) effect of each slipping wheelset means that longer trainsets benefit from a higher overall deceleration for a given initial adhesion. However, the effect of sand over and above this conditioning effect is not well understood. For a given initial adhesion condition, the adhesion improvement from sanding varies according to the amount of sand applied. Limited tests undertaken at different times with train borne sanders braking on artificially created low adhesion (dampened paper tape, which produces an adhesion condition closely replicating that of damp leaf film) have given the following results: AEATR-VTI , Issue 1 21 June 2003

32 8 Sand Flow & Performance Performance Benefit (%g) Flow Rate (kg/min) 6 In each test case the initial adhesion equated to a train deceleration of about 1.5%g. A linear best fit line is shown (dotted) which probably gives a reasonable indication of the average performance benefit for different sand flow rates at this level of initial adhesion. The highest sand flow rate (5.5 kg/min) is a nominal figure for the emergency one-shot sander used on some stock. It is important to note that these sand flow rates are fixed, measured at the end of the sand hose directed at the wheel/rail nip (i.e. taking into account any normal curvature of the hose). The Railway Group Standard for train borne sanders (Ref. 10) mandates that the rear of a train must not come to rest on sand laid at rate of 7.5 g/m or greater (the Critical Sand Density). This limit is set to protect the integrity of track circuit operation. It also mandates that a recognised method of achieving this is by a fixed laying rate approaching, but not exceeding, 2kg/minute per rail using a full service or emergency brake application. The effect of this is shown graphically below: AEATR-VTI , Issue 1 22 June 2003

33 20 Sand Density on Rail (Fixed Rate Sander) Sand Density (g/m) Critical Sand Density 1 kg/min 2 kg/min 5.5 kg/min (One-Shot) Speed (km/hour) This graph indicates that it is possible to lay more sand at higher speeds (above 30 km/hour in the 5.5 kg/minute case) without risking disruption to track circuits actuation whilst gaining the safety and performance benefits that higher retardation rates will give. Equally, by laying less sand at lower speeds, the risk to track circuit actuation can be further reduced. A variable rate sander provides the option for controlling the sand delivery rate according to train speed, braking demand and/or the available adhesion to further improve low adhesion braking performance whilst maintaining track circuit integrity. Ideally, such a system would be configured as part of a closed loop train braking control system such that the sand flow rate could be adjusted, within the limitations of the critical sand density criteria, to achieve the demanded train retardation. The majority of stock is currently fitted with fixed rate sanders working at the approaching 2 kg/min rate. Experience of these systems has shown that it is difficult to achieve the design sand flow rate. Variations in hopper design, hose run and hose length can all have a significant negative effect. Based on limited tests on some stock and inspection of hose routings on others, it is probable that the majority of stock achieves sand flow rates in the 1.25 to 1.75 kg/min range. From the earlier graph, this would result in a 1.5%g to 3%g improvement in deceleration when the sander is operated in low adhesion conditions. Some stock may be worse. AEATR-VTI , Issue 1 23 June 2003

34 Where the sander operates in full service and emergency braking only, its use is limited because of the professional driving techniques used during the autumn. Drivers will normally select lower brake rates during the initial phase of braking. Only when they are aware that the adhesion conditions are so low that they may not be able to stop in the required distance do they select a higher brake rate such that the sander will operate. In that case, the overall deceleration will be lower than could be achieved if the sander were to be available throughout the low adhesion stop. It is not possible to quantify this reduction in performance because it is dictated by individual driving style. In summary, the maximum adhesion improvement from train borne sanders operating at the approaching 2 kg/min rate will equate to a 3%g improvement in deceleration, although it may in some circumstances be substantially less. Hence, in absolute worst case adhesion conditions (an adhesion of, say, 0.01) a train should achieve around 4%g deceleration, 2%g below the ERTMS intervention rate. 5.2 In-Service Performance Background It is useful to relate the track testing data to that achieved by vehicles in operational service. In practice, it is very difficult to define the exact benefit that is derived from sander operation during actual service running as it is not possible to measure the adhesion performance improvement simultaneously with and without the sanding present. Realistically, this can only be achieved by operating two trains over the same section of track in quick succession and recording the data from each. However, it is possible to gather data from vehicles in operational service to give an indication of the achieved deceleration with sanding present. This at least indicates the typical lowest level of adhesion that would be experienced by an ERTMS fitted vehicle with operational sanding equipment. If sufficient data is gathered on such events then this can provide the basis of a probability of occurrence for adhesion below the ERTMS demand level. This information was potentially available from two sources: Vehicles fitted with the Low Adhesion Warning System (LAWS) and sanding equipment AEATR-VTI , Issue 1 24 June 2003

35 Vehicles fitted with train data recorders (OTMR) and sanding equipment AEATR-VTI , Issue 1 25 June 2003

36 5.2.2 LAWS Data Permission was sought from Train Operators with LAWS fitted trains to download all event data between September 2002 and January Data was filtered to identify those events where sanding would have been demanded due to WSP activity combined with a Step 3 brake demand. The overall number of events meeting this criterion was very low with only seven recorded events from the 20 LAWS fitted trains in the 14 week period. The summary of these events is attached in the report in Appendix A. The actual minimum recorded deceleration value is 3.04%g suggesting a value of adhesion below 0.01 (assuming a 3%g improvement resultant from sanding). The average of the results was 4.5%g. However, with only seven recorded events, there is insufficient data to make reliable statements about the typical minimum level of adhesion that will result from use of sanding. The low number of recorded events may in part be attributable to an improvement in driving technique resulting from defensive driving, which reduces the likelihood of slip occurring initially and which decreases the use of sand, which is often only available in full service and emergency. Where LAWS is fitted to three car or longer sets, it is usually fitted on a centre car to give equivalent data in either direction of operation. On sander fitted stock, the centre vehicle will get the full benefit from any sand application, potentially reducing the amount of wheel slip activity recorded. This factor may also partially account for the low number of events recorded OTMR Data OTMR devices typically record events continuously whenever the input signal conditions are right to do so. Therefore, assuming conditions are appropriate, there are potentially a large number of recorded events for OTMR data when step 3 brake, WSP activity and consequently sanding has occurred. Unfortunately, OTMR systems do not lend themselves to easy identification of these events. With an OTMR device it is necessary to know the date/time of an event in order to identify the occurrence in the OTMR memory. Most events have only a limited impact on train operations and will not be downloaded by the train operator. AEATR-VTI , Issue 1 26 June 2003

37 In practice, events occurring under SPAD conditions provide a good indication of very low adhesion conditions but also provide the added benefit of being recorded formally permitting easier identification from OTMR memory. This data is summarised in the Rail Safety and Standards Board SMIS database, and it was proposed to use this database to identify all SPAD and any overrun events where poor rail condition was identified as a cause. Once this had been established the Train Operators would then be contacted to obtain downloads of the particular events concerned. Unfortunately, specific low adhesion information was delayed until March 2003 in the Rail Safety and Standards Board monthly SPAD reports. This was caused in part by the significant problems arising from the storm of 27/10/03. These reports are useful because they identify the latest events and which TOC s have been affected. The Rail Safety and Standards Board was approached directly for an updated copy of the low adhesion SPAD s from the SMIS database. From the OTMR data available, interpretation of the actual achieved vehicle deceleration is an approximation. Under low adhesion conditions there is usually no clear mechanism for recording the vehicle stop point on the OTMR plot. The stop point is assumed to be where WSP activity ceases or where information marked up on the plot by the TOC indicates the approximate stop point. All of the recorded SPAD events relate to events on the 28 th and 29 th of October There was a dramatic increase in SPAD and overrun activity on parts of the network on these days following a severe storm on 27 th October which caused large quantities of leaves to be removed from trees along the line side. However, just six SPAD and station overrun events have been identified from the database where low adhesion was found to be the primary cause. There were 3 SPAD events on the SMIS database for which sanding occurred on OTMR fitted stock. Operating Standards Manager s for the respective TOC s have provided this data and the results of these are discussed in more detail in the report in Appendix B. From the SMIS database it has also been possible to identify 29 potential station overrun events where the vehicles are fitted with sanding and OTMR equipment. The TOC s have been approached for data for these events, but the data has proven more difficult to obtain. Where the station overrun is small, the data is often not downloaded. When it is downloaded, it is not necessarily retained. As an alternative, a general approach was made to Operating Standards Manager s for each TOC where it was known that vehicle fleets are sander AEATR-VTI , Issue 1 27 June 2003

38 fitted. The OSM s were asked to provide information for any SPAD or overrun event that they were aware of, particularly where this involved sand being applied. The overall deceleration obtained from the OTMR records of these SPAD and overrun events for trains fitted with sanders is 4.25%g. The minimum recorded deceleration with sand present was 2.0%g averaged over the stop. There is insufficient data to make reliable statements about the typical minimum levels of adhesion that would be experienced. An important point relating to sanding systems and train safety can be identified from the OTMR data. Under circumstances of very low adhesion where the WSP loses control of the train speed reference, the driver will usually select the Emergency brake. Assuming there is WSP activity present this will cause sand to be delivered to the rail. Unfortunately, under these low adhesion conditions, the internal speed reference of some WSP systems may not always continue to maintain an accurate speed reference and this can result in sand delivery being terminated prematurely. The impact of this is that the train braking distance can be extended. This problem could be overcome if under Emergency conditions sand delivery is not reliant upon a WSP activity signal. This is allowed under the Railway Group Standard for Sanders, but may not be utilised in all cases. 5.3 Sander Summary From the limited amount of LAWS and OTMR data available, the overall level of deceleration achieved, in worst case low adhesion conditions, with sanding is approximately 4-4.5%g. This is similar to the improvement in deceleration achieved in track testing where the initial (not sanded) adhesion was In view of this result, and the fact that there were relatively few SPAD s with sander fitted vehicles, indicates that train-borne sanders are achieving the expected in-service performance levels. However, in all but one of the SPAD examples the achieved retardation is lower than the ERTMS demand level. This means that in the most severe cases sanding will not necessarily return the rail adhesion level to within the brake demand level of the ERTMS. AEATR-VTI , Issue 1 28 June 2003

39 6 Low Adhesion Risk Based on the earlier slip risk probability data covered in Section 3.2.1, it is possible to quantify the risk of meeting a section of low adhesion whilst braking in the ERTMS low adhesion mode. It can be seen from the table in 3.2.1, that the probability of meeting any length of low adhesion below 0.06 is 9.2 x Similarly, the probability of meeting a low adhesion section of longer than 100m is 1.6 x This data is plotted as a cumulative frequency histogram below. 1.00E-05 Cumulative low adhesion risk at proposed 6%g ERTMS low-adhesion level 7.50E-06 Low Adhesion risk 5.00E-06 ERTMS low adhesion setting (6%g) WITHOUT sander ERTMS low adhesion setting (6%g) WITH sander fitted 80 % reduction of Low Adhesion risk at 6%g without additional margin 2.50E E+00 > 0 > 100 >200 >300 >400 >500 Low adhesion length (m) A reference line has been added to the cumulative low adhesion risk chart at a level which represents an 80% reduction in slip risk compared with the any length of low adhesion figure of 9.2 x Braking over a significant length of low adhesion is likely to result in a SPAD. The histogram indicates that more than an 80% reduction in the risk of any SPAD occurring could be achieved by a combination of braking at the proposed ERTMS low adhesion brake rate of 6%g and including an additional 100m margin into the braking distance. The above analysis takes no account of the relative severity of different SPAD categories. For example, in terms of a non-ertms railway where defensive driving is employed, all AEATR-VTI , Issue 1 29 June 2003

40 low adhesion SPADS of up to 100m beyond the signal would be eliminated, assuming the correct 6%g braking point was always selected. If the 6%g brake rate for the ERTMS low-adhesion mode was used in conjunction with a sander, which typically improves the adhesion level by 3%g, then the probability of meeting adhesion below 0.03 is eliminated. The risks of encountering slightly higher levels of adhesion are reduced accordingly. In that case, the above 80% reduction in risk of a SPAD occurring could be achieved without the inclusion of any additional margin. AEATR-VTI , Issue 1 30 June 2003

41 7 SPAD Frequency from Historic Data SPAD data recorded over the last 5 years was downloaded from the SMIS database and analysed according to the allocated SPAD category. In all cases the SPAD events selected were attributable to low adhesion conditions (either leaves, greasy rail, moisture, etc). The results were considered across the year as a whole: SPAD Category Description 1 Overrun 0 to 25 yards not exceeding overlap and no damage injuries or deaths 2 Overrun 26 to 200 yards not exceeding overlap and no damage injuries or deaths 3 Overrun greater than overlap and overrun greater than 201 yards no damage injuries or deaths Number of Adhesion Related Incidents 62 4 Track damage only, no casualties 1 5 Derailment with no collision and no casualties 1 6 Collision (with or without derailment) and no 0 casualties 7 Injuries to staff or passengers with no fatalities 1 8 Death of staff or passengers Total number of SPADs from June '97 to June '02 These results were then filtered to take account of only those events that occurred during the autumn period (October to December) of each year, as follows: AEATR-VTI , Issue 1 31 June 2003

42 SPAD Category Description Number of Adhesion Related Incidents 47 1 Overrun 0 to 25 yards overrun not exceeding overlap and no damage injuries or deaths 2 Overrun 26 to 200 yards overrun not exceeding 49 overlap and no damage injuries or deaths 3 Overrun greater than overlap and overruns 27 greater than 201 yards no damage injuries or deaths 4 Track damage only, no casualties 2 5 Derailment with no collision and no casualties 0 6 Collision (with or without derailment) and no 0 casualties 7 Injuries to staff or passengers with no fatalities 1 8 Death of staff or passengers Total number of SPADs during each autumn period from June '97 to December '02 The above data is shown as a frequency histogram per annum/autumn below: AEATR-VTI , Issue 1 32 June 2003

43 14 Number of SPAD incidents per period against SPAD category (all SPADs as a result of low adhesion conditions) Number of incidents per annum Number of incidents per autumn Number of Incidents SPAD Category The year round and autumn trends are similar. The majority (80%) of year round low adhesion SPAD events over the five-year period are categories 1 and 2, which equate to an overrun of up to 200 yards. Higher risk category SPAD s (categories 3 to 7) constituted the remaining 20% of events and 40% of these (8% of the total number of low adhesion SPAD s) also resulted in overruns of less than 200m. AEATR-VTI , Issue 1 33 June 2003

44 8 SPAD Risk The Railway Group Safety Plan (Ref. 11) objective is to reduce the risk of all SPAD s by at least 80% of the 2000/01 level. The risk of a SPAD is associated with the SPAD severity categories and includes the potential consequences of that SPAD e.g. derailment, collision, injury etc. The information derived in Sections 6 and 7 above does not provide any assessment of the risks associated with the SPAD s. In the Rail Safety and Standards Board Risk Profile Bulletin, Appendix B - Cause Precursor Risk Contributions, the frequency (events per train mile) and risk contributions are set to zero and assigned code N (no data available). It is stated that a research project has commenced to examine the methodologies available for use where little or no data exists as part of the Rail Safety and Standards Board research and development programme. Until such time as the frequency and risk contributions are established, it is not possible to predict the effect of reducing the number of low adhesion SPAD s on the total low adhesion SPAD risk e.g. removing all category 1 SPAD s may have little effect on the total risk. However a greater than 80% reduction in the risk of a low adhesion SPAD occurring can be achieved, as outlined in Sections 6 and 7 above. The ERTMS low adhesion mode could be implemented at any time throughout the year, hence the historical data relating to the total number of SPAD incidents over the 5 year period is relevant to this analysis. The historic SPAD data indicates that a combination of braking at the proposed ERTMS low adhesion brake rate of 6%g and including an additional 200m margin into the braking distance (plus an allowance for gradients) would effectively remove 88% of all low adhesion SPAD s. This 88% consists of all category 1 and 2 low adhesion SPADS (80%) plus 8% from higher risk categories. It is also important to note that the remaining 12% of SPAD s would also be reduced in severity. Under existing non-ertms operating procedures, the driver will brake earlier and at a lower brake demand (defensive driving) to minimise the SPAD risk in low adhesion. However, as was demonstrated in Section 2.1.3, the driver has no indication of the braking point required to stop before the signal at the demanded brake rate, normally 6%g (Step 2 brake). He will normally select a brake point based on route knowledge and experience. If this point is beyond the ideal braking point and low adhesion is encountered then a SPAD is likely to result. AEATR-VTI , Issue 1 34 June 2003

45 Operating under the ERTMS proposals made in this report, the undesirable situation described in the previous paragraph will not arise. The driver will have the benefit of low adhesion information about the required braking point or, if the required braking curve has been exceeded, an intervention will occur. 8.1 SPAD Risk Summary From the low adhesion risk data (see Section 6), there is a relatively high probability of encountering 100 m of low adhesion, therefore a margin of 100m would eliminate the effects of these lengths of low adhesion. With reference to the historic SPAD data above, a 200m margin is required to eliminate category 1 and 2 and some higher category SPAD s and to reduce the severity of the remaining higher category SPAD s. This difference in required margin between the results of the two analyses is likely to be because: 1. the historic SPAD data is based on the driver selected braking point which may be inaccurate, 2. the slip risk data does not include the effect of increasing operating speed, which may increase the lengths of low adhesion encountered, and hence the SPAD risk. AEA Technology Rail suggests that, in their expert opinion and in the absence of SPAD risk frequency and contribution data, a combination of braking at the proposed ERTMS low adhesion brake rate of 6%g and an additional 200m margin in advance of the 6%g braking point (plus an allowance for gradient) is adopted for ERTMS operation to reduce the operational risk to ALARP. It may be possible to reduce this margin to 100m for sander fitted stock, although sander reliability would have to be considered since it may then be deemed Safety Critical. Clearly it would also be possible to adopt a greater margin, at the expense of operational capacity, to reduce the SPAD risk further. However, the smaller number of higher risk SPAD s are likely to be caused by extreme low adhesion conditions and it is widely accepted in the industry that any one measure is unlikely to prevent such occurrences. A wide range of adhesion control measures should be employed to minimise the risk of such incidents. AEATR-VTI , Issue 1 35 June 2003

46 9 The Intervention Brake Rate The above analysis indicates that an acceptable reduction in low adhesion SPAD risk may be achieved by utilising an ERTMS low adhesion braking curve at 6%g deceleration plus an additional margin. The driver of a train operating under ERTMS will be aware from the cab display that he is driving in low adhesion conditions and that the ERTMS is operating in low adhesion mode. He must brake earlier than the intervention braking curve to prevent an ERTMS intervention. This will add a further small factor of safety. In high adhesion, ERTMS operates with a full service brake application as the intervention curve is exceeded. Should the brake fail to apply, an emergency brake application will occur. Under low adhesion conditions the intervention braking point is moved to allow the train to be brought to a stop with a lower brake rate. If the driver does not respond to the low adhesion warning, which would be provided by the adhesion management system (Ref.4) via the in-cab display, and apply the brake before the intervention point, the brake will be applied automatically. This raises a question about the actual brake rate applied as a result of the ERTMS intervention. The intervention brake rate (as distinct from the ERTMS braking curve) could be set at the defensive driving rate of 6%g, at the normal full service rate of 9%g or at the enhanced emergency braking rate of 12%g. These options are discussed below: AEATR-VTI , Issue 1 36 June 2003

47 9.1 Intervention Brake Rate at 6%g At a 6%g intervention brake rate, the train deceleration will follow the ERTMS low adhesion-braking curve. The risk of a SPAD occurring, as discussed in Section 7, will apply. Ideally, if the stock is fitted with automatic sanding equipment, it should be configured to operate at the 6%g brake rate to further reduce the SPAD risk. 9.2 Intervention Brake Rate at 9%g At a 9%g intervention brake rate, following intervention the demanded deceleration of the train would be in advance of the ERTMS low adhesion braking (6%g) curve. If, contrary to the current prediction of low adhesion, the actual wheel/rail adhesion is greater than the brake demand rate, the train will come to rest with significant margin, depending upon the line speed. The margin for different line speeds from Section (and including a 100m allowance from low adhesion risk data, Section 6) is: Line Speed (mile/hour) Margin (m) If there are sections of low adhesion during the braking stop the risk of a SPAD occurring will increase. The histogram below, introduced in Section 6, is modified here to include the 9%g brake demand SPAD risk data: AEATR-VTI , Issue 1 37 June 2003

48 3.00E-05 Cumulative SPAD risk for units with and without sanders fitted 2.50E-05 Standard stop (9%g) without sander SPAD risk 2.00E E-05 ERTMS low adhesion setting (6%g) WITHOUT sander ERTMS low adhesion setting (6%g) WITH sander fitted 80 % reduction of SPAD risk at 6%g without additional margin 1.00E E E+00 > 0 > 100 >200 >300 >400 >500 Low adhesion length (m) N.B. For the purpose of this analysis, the 9%g risk of a SPAD assumes that the train would normally stop exactly at the signal when braking in high adhesion and that any length of low adhesion encountered during that stop would then result in a SPAD. Hence no additional margin, as used in conventional signalling, has been incorporated. From the above graph, a margin of 300m in addition to the 9%g braking distance will result in the same low adhesion SPAD risk as an allowance of 100m on top of the 6%g braking distance. This is because the risk of meeting longer lengths of low adhesion reduces very rapidly the best-fit to the data in the above graph approximates to a square law. Any 9%g intervention brake event where the approach speed is such that the margin is greater than 300m will have the same or less risk of a SPAD than is proposed at 6%g + 100m allowance. From the table above, this applies to all line speeds above 60 mile/hour. For line speeds of less than 60 mile/hour the risk of a SPAD following a 9%g brake intervention will increase because the available margins are small. Where such line speeds exist on an ERTMS route, it may be necessary to consider the risk on a site by site basis. For example, it is likely that the risk of low adhesion at a particular site may be lower than predicted here if there is no vegetation in the vicinity. Alternatively, the intervention brake rate can be AEATR-VTI , Issue 1 38 June 2003

49 restricted to 6%g to maintain the required overall level of risk of a SPAD occurring. It may appear from this analysis that, with a 9%g intervention brake rate, the proposed margins are too generous at high speeds. For example the margin at 100 mile/hour is 658m, where, by calculation, just 300m is required to achieve an 80% reduction in the risk of a SPAD occurring. However the analysis does not include the effect of the variation of adhesion with speed. This is because, as discussed earlier, it is not known how increasing operating speed may vary the lengths of low adhesion encountered. However, as the effective adhesion does decrease with increasing operating speed, it is possible that the effective length of low adhesion sections will increase. It is suggested that the additional allowance should be maintained to provide an extra safety factor to compensate for any such effect. In contrast to the above slightly negative aspect, there is an advantage in selecting a 9%g intervention brake rate. Because of the natural variation in adhesion, there will be instances when braking in low adhesion where the adhesion rises above that required to sustain the nominal 6%g brake demand level. In these cases the 9%g rate, used in conjunction with a WSP which is optimised to make the best use of the available adhesion, will reduce the overall stopping distance compared with that which would be achieved by a 6%g intervention brake rate. 9.3 Intervention Brake Rate at 12%g The risk of meeting significant lengths of low adhesion below a 12%g enhanced emergency intervention brake rate has not been analysed as part of this study. However, it is likely that similar factors to those discussed above for the 9%g intervention brake rate will apply. Also the low adhesion slip risk levels will be significantly higher (50 times higher from Section 3.3) so that it may be difficult to justify such an intervention brake rate for this low adhesion application. However, it will be necessary to employ this emergency brake rate in low adhesion conditions should the required intervention brake rate not occur, i.e. in the case of a service brake failure. In that case the increased slip risk would be an unfortunate, but very necessary, consequence of the action necessary to avoid complete brake failure. AEATR-VTI , Issue 1 39 June 2003

50 10 Limitations Of The Analysis 1. In conducting the low adhesion risk analysis of section 6, the effect of speed on the length of low adhesion sections has not taken into account. There is also a risk that multiple low adhesion events may occur in a single braking event. Again this was not considered due to a lack of data. However these factors potentially will be an element of the historic data considered in Section 7 and may partially explain why that data is slightly worse than the calculated results indicate. The 12% of low adhesion SPADs that would remain after utilising a 6%g deceleration rate plus 200m margin e.g. those that are beyond the current overlap distance, are likely to be due to long lengths of very low adhesion which occur very occasionally. Whilst these SPADs would be reduced in severity, the measures proposed in this report would not significantly alter them. A significant change in adhesion levels would be required. However, it is possible that some of these extreme events are a combination of very low adhesion and the driver failing to take full account of the prevailing conditions. In that case ERTMS intervention would also help minimise the severity of a SPAD. 2. This analysis assumes a level of adhesion as seen by a single leading braked wheelset, as measured by the BR Research Tribometer Train. However, the effective adhesion seen by longer trains may be higher. This is because heat is generated at the wheel/rail interface according to the level of slip and the friction coefficient. This local temperature rise, together with braking shear forces (a squeegee effect of the wheelset slipping over the rail), reduces the moisture at the interface, and potentially also removes some of the railhead contaminant. It follows that the next wheelset will meet a drier, cleaner rail and the adhesion available to support the braking forces will be greater. AEA Technology Rail has measured this rail conditioning effect for a maximum of four slipping axles, and the results are incorporated into the WSPER single vehicle model. However, the conditioning effect will continue down the train. As the initial adhesion gets higher, so the energy dissipated increases, such that, at some point, the moisture in the interface is vaporised. Under these conditions the adhesion will probably rise above the demanded adhesion level such that the full braking performance of that wheelset is achieved. The Japanese have measured the frequency of WSP activity along the length of a Shinkansen train (Ref. 8), which gives an indication of the improvement in adhesion AEATR-VTI , Issue 1 40 June 2003

51 down the train. This data indicates that, when the first vehicle experiences wheel slip, the last vehicle in a nine-car rake is 90% less likely to slip. It follows that longer train sets such those operated on inter-city services are more likely to be able to make full use of the conditioning effect. Common UK operating practice supports this analysis because short multiple units are generally coupled together during autumn leaf fall operations to reduce the slip risk. It is likely that the extra margins proposed in this report could be relaxed for longer train formations, although further work would be required to improve understanding of this phenomena. 3. The adhesion data used in this report was obtained using the Tribometer Train in autumn periods of 1992 and 1993 and the data is therefore a snapshot in time. Measurements were on a circuit between Basingstoke to Eastleigh (high speed InterCity 3 rd rail electrified route), Eastleigh to Salsibury (Non-electrified route used by local services and Cardiff- Portsmouth trains class 15x units) and Salisbury to Basingstoke (non-electrified used by the Waterloo-Exeter service etc class 15x and 170). It is assumed that data from these routes is representative of the network. These routes had significant lineside vegetation over approximately 15% to 20 % of the length of each section when surveyed during Vegetation management reduces the probability of low adhesion but new vegetation or re-growth will increase the probability. The level of vegetation at many low adhesion sites will not have changed because management is not possible due to tress not being on railway land or because of environmental reasons. Hence it is assumed that, on a network wide basis, the overall proportion of routes that are currently low adhesion sites is unlikely to have changed significantly. 4. The report considers the risk issues in relation to disc braked rolling stock as these operate at higher braking rates and operating speeds than older tread braked designs (Ref. 12). Because of these higher braking rates and operating speeds, the risk of wheel slip is also greater (see Section 3). Hence the analysis conducted in this report considers the worst case condition. Where the full service braking rate is lower, the degree of modification of the normal adhesion intervention curve for low adhesion operation will be correspondingly reduced. AEATR-VTI , Issue 1 41 June 2003

52 5. Tread braked stock may have a non-linear braking characteristic, usually resulting in increased train deceleration at low speeds. Linear braking characteristics have been assumed in this report for the train-borne ERTMS controller. However, where non-linear braking characteristics are utilised, matching non-linear train-borne ERTMS controller curves could also be adopted. These will require separate consideration of the requirement for a low adhesion operating mode, which should be based on the considerations contained in this report. 6. No allowance has been made for the brake build-up or driver response time in calculating stopping distances. The distances quoted will therefore be shorter than the actual stopping distance by an amount that will vary according to operating speed and train performance. At the time of writing the method of including the above allowances in the ERTMS train-borne equipment for normal (i.e. high adhesion) operation has not yet been determined. When this methodology has been decided, it is suggested that the results of this report are re-evaluated to ensure that they are inline with that practice. AEATR-VTI , Issue 1 42 June 2003

53 11 Conclusions ERTMS will incorporate a train-borne low adhesion intervention braking curve that decreases at an effective train deceleration of 6%g. This rate is consistent with existing UK low adhesion driving best practice. A review of adhesion data from the BR Research Tribometer Train (section 3) and historic SPAD data (section 7) has been conducted. From this review AEA Technology Rail suggests that, in their expert opinion and in the absence of SPAD risk frequency and contribution data, a combination of braking at the proposed ERTMS low adhesion brake rate of 6%g and an additional 200m margin in advance of the 6%g braking point (plus an allowance for gradient) is adopted for ERTMS operation to reduce the operational risk to ALARP. The adoption of the above margin should eliminate more than 80% of all low adhesion SPAD s. The margin would also reduce the severity of the remaining SPAD s which generally are caused by severe low adhesion conditions. However, it is recommended that a wider range of adhesion control measures must be adopted to minimise the risk from such severe conditions. The action of the ERTMS low adhesion intervention algorithm will also minimise the effect of driver misjudgement in such conditions, further reducing the risk of a SPAD occurring. It is suggested that some of the above benefits could be achieved on a non- ERTMS railway by the adoption of low adhesion braking point signs placed at the 6%g braking point (plus allowance) at critical low adhesion signals. This would reduce the reliance on individual judgement of distances and braking performance required of every driver, which is bound to introduce some increased SPAD risk. Tests have shown that train borne sanders constructed to the Railway Group Standard (Ref. 10) offer around a 3%g improvement in deceleration in low adhesion conditions (section 5). A review of data from OTMR and LAWS fitted trains indicates that such levels of improvement are generally achieved in practice. Sander fitted trains would be at a reduced SPAD risk when operating under ERTMS low adhesion mode. Alternatively, the 200m margin could be reduced to 100m for sander fitted trains for the same level of AEATR-VTI , Issue 1 43 June 2003

54 SPAD risk. However, in that case, sander reliability would have to be reviewed since it may then be deemed safety critical. Modern microprocessor based WSP equipment makes the best use of the available adhesion and is used to operate the sanding equipment when required (Section 4). Older analogue WSP designs used in the UK, which are 20+ years old, tend to allow the wheelsets speeds to drift too far from the true train speed. From WSPER test data, this means that the overall deceleration is reduced by an average of 0.65%g compared with a modern WSP system under identical low adhesion conditions. Also the axles can cease to rotate in very low adhesion conditions. In that case, to the WSP, the train will appear to have come to a stop such that no further WSP activity occurs and the sander would be inhibited. For these reason, and the odometric accuracy problems identified in Ref. 9, the use of older WSP systems for ERTMS purpose is not recommended. This report has considered the ERTMS low adhesion brake rate and the extra margin required to minimise the SPAD risk. However, the actual intervention brake rate to be adopted when a train crosses the ERTMS low adhesion braking curve has briefly been identified as an issue which requires further consideration. A 9%g intervention rate offers some advantages at line speeds above 60 mile/hour. AEATR-VTI , Issue 1 44 June 2003

55 12 Recommendations The ERTMS low adhesion braking curve should be based on a 6%g deceleration rate with an additional 200m margin in advance of the 6%g braking point plus an allowance for gradient (Section 7). It may be possible to reduce the above margin to 100m for sander fitted stock, although sander reliability would have to be considered since it may then be deemed Safety Critical (Section 7). To get the best advantage from sanding systems in reducing ERTMS low adhesion SPAD risk, they should be configured to provide sand at 6%g brake demand (see Section 8.1). There would also be a benefit in using this configuration for existing operations where defensive driving uses a 6%g brake demand in the low adhesion autumn period. Modern WSP should be used under ERTMS to make the best use of the available adhesion and ensure sander activation where available (Section 4). Under extreme low adhesion conditions, the internal speed reference of some WSP systems may not always continue to maintain an accurate speed reference and this can result in sand delivery being terminated prematurely. This problem could be overcome if under Emergency conditions sand delivery is not reliant upon a WSP activity signal. This is allowed under the Railway Group Standard for Sanders, but may not be utilised in all cases (Section 5). AEATR-VTI , Issue 1 45 June 2003

56 13 References 1. Waboso D, The ERTMS Programme Team Final Report, Railway Safety Report Issued to HSC April AEATR-VTI Project summary report to be published. 3. Corke A, Review of ATP, ATO & ERTMS, AEA Technology Rail, Report Reference AEATR-VTI , March Poole W, The Feasibility of Adhesion Control and Monitoring for ERTMS, AEA Technology Rail, Report Reference AEATR-VTI , March Lineside Signal Spacing Railway Group Standard GK/RT Pritchard C, Brakes and Wheel/Rail Adhesion, Institute of Mechanical Engineering Conference on Railway Braking, York, 26/27 September ORE Report 2, Question B164, Adhesion during Braking, and Anti-Skid Devices, Utrecht, Ohyama Dr.T, Adhesion Characteristics of Wheel/Rail System and its Control at High Speeds, 4th RTRI Symposium, Tokyo Nov 7th Rowell B, Review of Adhesion Dependent Odometry, AEA Technology Rail, Report Reference AEATR-VTI Issue 1, March Railway Group Standard GM/RT2461 Sanding Equipment Fitted to Multiple Units and On-Track Machines. 11. Railway Group Safety Plan 2003/ Schofield K, Tunley J, & Waring J A comparison of tread and disc braking for passenger rolling stock with particular reference to low adhesion, BR Research Report, RR-SAM-135, March Acknowledgements The Rail Safety and Standards Board has sponsored the work and their support during completion of the task is appreciated. The contributions from colleagues in the Wheel/Rail Adhesion team is also acknowledged and appreciated. AEATR-VTI , Issue 1 46 June 2003

57 Appendices AEATR-VTI , Issue 1 47 June 2003

58 Appendix A SPAD and Overrun Data from OTMR s 1.1 Introduction This report represents OTMR information gathered from discussion with TOC s regarding SPAD and station overrun events where sand was applied to the rail head and where low adhesion was the cause. The purpose of this data is to establish a typical minimum value of achieved retardation by a train when braking and applying sand under low adhesion conditions. This information is intended to provide a comparison with track based test data where simulated low adhesion conditions were produced and sand applied. The results obtained from each are then to be used as part of the consideration of train braking strategy under low adhesion conditions for ERTMS. 1.2 Method Information was obtained by: Discussion with TOCS s regarding low adhesion events they had records of. Review of SMIS data for SPADS and station overrun events where low adhesion had been identified as a cause. The TOC were then approached to obtain records if they existed. It was intended to monitor monthly SPAD reports from the Railway Safety website. Unfortunately, due in part to severe weather conditions on 27/10/03, this information was not published until March As an alternative, in January 2003 a general approach was made to each train operator to identify any SPAD or overrun events that could be attributed to low adhesion. The SMIS data had to be filtered to identify fleets that are sander and OTMR fitted. A summary of the events identified is listed below: SPAD events Date Time Signal Location Unit/Veh Operator WTT No Dep time Origin Dest 28-Oct-02 09:16 WJ189 Kings Langley Silverlink 1W44 07:48 Northampton Euston AEATR-VTI , Issue 1 48 June 2003

59 28-Oct-02 21:17 LR209 Little Bowden LC Mid ML 1F57 19:30 St Pancras Derby 29-Oct-02 04:30 TT439 Hathern Mid ML 5C01 04:12 Derby, EP Leicester Station overrun events First North Western 17-Oct-02 QNW/OCT/ Colwyn Bay 2002/ Oct-02 QNW/OCT/ 2002/734 Scotrail Railways Ltd 24-Oct-02 HSC/2002/O CT/3376 1D33 07:17 Manchester Piccadilly Llandudno Rail contamination leaves Llanfairfechan 1K57 07:49 Holyhead Crewe Rail contamination leaves Broughty Ferry 1A79 16: Other line defect Edinburgh Waverley Dyce Connex South Central Trains 27-Oct-02 QSR/2002/0 8/702 Wandsworth Common 2R17 22:33 Dorking Victoria Rail contamination leaves 29-Oct-02 QSR/2002/0 Reedham 2P71 18:45 Tattenham Corner Rail contamination 8/790 (Surrey) Victoria leaves Midland Main line Ltd 31-Oct-02 QMD/61044 /2002/08 South West Trains Ltd 17-Oct-02 QSR/2002/0 8/ Oct-02 QSR/2002/0 8/568 Luton 1B26 10:52 Nottingham St Pancras Rail contamination other The driver failed to tell the signaller that there were exceptional rail head conditions in the area line contaminated by leaf mulch No data available Oxshott 2G18 08:20 Guildford Waterloo Rail contamination leaves Effingham Jcn 2D96 16: Rail contamination (Leatherhead Guildford Waterloo leaves Line) 28-Oct-02 QSR/2002/0 Surbiton 2G03 06:28 Wimbledon Rail contamination 8/579 Guildford leaves Connex South Eastern Ltd 29-Oct-02 QSR/2002/0 Beckenham Hill 2B67 20: /887 Sevenoaks Blackfriars 29-Oct-02 QSR/2002/0 Beckenham Hill 2B61 18: Rail contamination Sandite found on rail 8/908 Sevenoaks Blackfriars other 01-Nov-02 QSR/2002/0 Beckenham Jcn 2D74 20: Rail contamination 8/927 (Main Line) Orpington Victoria other 11-Nov-02 QSR/2002/0 Dunton Green 2S65 17: /97 Cannon St-Sevenoaks 12-Nov-02 QSR/2002/0 Grove Park 2F52 10: Data not suitable for 9/67 (Main Line) Orpington-Charing Cross use 12-Nov-02 QSR/2002/0 St Mary Cray 2B14 08: Data not suitable for 9/72 Blackfriars-svenoaks use Thames Trains Ltd 28-Oct-02 QSR/2002/0 Blackwater 1O52 04:35? No data available 8/549 Reading Gatwick Airport 30-Oct-02 QGW/82655 Shipton 1D45 16:18 Paddington Worcester Foregate Data not suitable for use Street Thameslink Rail Ltd 28-Oct-02 QMD/60937 /2002/08 Harlington 2L50 16:25 Moorgate Bedford Rail contamination leaves 28-Oct-02 QMD/60940 /2002/08 Harlington 2L54 17:10 Moorgate Bedford Rail contamination leaves 14-Nov-02 QMD/61413 /2002/09 Radlett 2O19 10:01 Luton Luton Rail contamination other AEATR-VTI , Issue 1 49 June 2003

60 Connex C2C 24-Oct-02 QER/2002/0 8/ Oct-02 QER/2002/0 8/ Oct-02 QER/2002/0 8/ Oct-02 QER/2002/0 8/ Oct-02 QER/2002/0 8/ Nov-02 QER/2002/0 8/9578 East Tilbury 2D48 20:20 Fenchurch Street Shoeburyness Upminster 2B08 07:10 Fenchurch Street Shoeburyness East Tilbury 2D48 20:20 Fenchurch Street Shoeburyness Grays 2R21 08:16 Pitsea Fenchurch Street Stanford le 2D58 23:55 Hope Upminster Pitsea Grays 2D55 20:20 Southend Central Fenchurch Street 12-Nov-02 QER/2002/0 9/12500 West Horndon 2B27 10:35 Shoeburyness Fenchurch Street 12-Nov-02 QER/2002/0 Laindon 2B27 10:35 9/9968 Shoeburyness Fenchurch Street 14-Nov-02 QER/2002/0 Laindon 1B11 06:52 9/10172 Shoeburyness Fenchurch Street? Rail contamination other Additional Data collected Additional data which is not listed in the SMIS database was supplied by Thameslink. Date Time Signal Location Unit/Veh 29-Oct-02 19:05? Unknown Oct-02 19:17? Unknown Dec-02 04:29? Unknown Findings The response has been low, producing only 6 usable events. There were 3 SPAD events on the SMIS database for which sanding and OTMR occurred. From the SMIS database it has been possible to identify 29 potential station overrun events where the vehicles are fitted with sanding and OTMR equipment. Unfortunately, for the majority of these overrun events it has not been possible to obtain the data from the TOC s. The three remaining events were supplied by TOC s in response to a general request for SPAD and station overrun data. All of the recorded SPAD events relate to events on the 28 th and 29 th of October There was a dramatic increase in SPAD and overrun activity AEATR-VTI , Issue 1 50 June 2003

61 on parts of the network on these days following a severe storm on 27 th October which caused large quantities of leaves to be removed from trees along the line side. From the data available, interpretation of the actual achieved vehicle deceleration is an approximation because there is usually no clear mechanism for recording the vehicle stop point on the OTMR plot when braking in low adhesion conditions. The stop point is assumed to be where WSP activity is seen to stop or where information marked up on the plot by the TOC indicates the approximate stop point. A summary of the results obtained is indicated below: Details Achieved retardation MML, SPAD 29/10/02, %g MML, SPAD 28/10/02, %g Silverlink, SPAD 28/10/02, %g Thameslink, O run, 29/10/02, %g Thameslink, O run, 29/10/02, %g Thameslink, O run, 14/12/02, %g The overall retardation value with sanders fitted based upon this data is 4.25%g. The minimum recorded value where sand should have been applied was 2%g, although the driving technique used is suspect. With the limited sample of events there is insufficient data to make reliable statements about the typical minimum levels of adhesion that would be experienced. From the OTMR plot data, it can be seen that in circumstances of very low adhesion where the WSP loses control of the train speed reference, the driver will have often selected Emergency brake. Assuming there is WSP activity present this will cause sand to be delivered to the rail. Unfortunately, under these low adhesion conditions, the WSP internal speed reference may not always continue to maintain an accurate speed reference and this can result in sand delivery being terminated prematurely. The impact of this is that the train braking distance can be extended. This problem could be overcome if under Emergency conditions sand delivery is not reliant upon a WSP activity signal. AEATR-VTI , Issue 1 51 June 2003

62 1.4 Conclusions it is beneficial to have sanding available when in Step 2 brake and WSP activity It is beneficial to have sanding exclusive of WSP activity when in Emergency Brake demand Based on the limited recorded data, the general level of adhesion achieved under braking with sanding present is around 4%g. 1.5 OTMR Data Midland Mainline 29/10/02 04:31:30, Unit 170, Veh: This SPAD event occurred on one of two particularly bad days, during the last leaf fall season in which an unusually severe storm affected the whole country. The storm occurred on 27/10/02 with strong winds and heavy rain resulting in significant quantities of leaves to fall from the trees along the line side in a short time period. A special investigation was requested by Network Rail following this storm (Rail Industry Special Review Stage 2). AEATR-VTI , Issue 1 52 June 2003

63 In the event concerned the driver has initially selected step 3 brake, presumably unaware that track conditions may be poor. This immediately resulted in WSP activity of which the driver will have been aware due to its severity. After approximately 20 seconds the driver selects Emergency Brake. The vehicle speed reference can be seen to quickly drift to zero. The internal speed reference of the WSP system will maintain a level to compensate for the lack of wheel speed reference and this results in continued WSP activity for a short period of time after the axle speed (corresponding to the speedometer indication) has fallen to zero. The AWS magnet indications show that the start of braking roughly coincides with the double yellow aspect but that the stop point indicated by the dotted line is a considerable time after the Red aspect. This point is assumed to be the stop point based upon it being the last point at which there is any WSP activity. This may be a little optimistic. With the assumed stop point the equivalent retardation of the train is 2.75%g. Sanding would have been delivered for most of this event whilst there was WSP activity when in Step 3 and Emergency brake selected. However, the apparent axle lock-up indficates a fault with the WSP which is likely to have inhibited sander activation. Midland Mainline 28/10/02 21:08:29, Unit 170, Veh: AEATR-VTI , Issue 1 53 June 2003

64 Once again this SPAD event occurred on one of two particularly bad days, during the last leaf fall season. In this event the driver initially selected step 2 brake. This resulted in some WSP activity indicating that the level of available adhesion was below the Step 2 brake demand level. The driver, presumably aware of the WSP activity, selects Step 3 brake, which enables sand to be applied. When the train passes a Permanent Speed Restriction (PSR) the driver responds by selecting Emergency brake demand. The driver response is most likely to be as a consequence of entering the PSR with the train speed too high. The WSP activity continues for approximately 30 seconds after this until the WSP internal speed reference has fallen to zero. A consequence of this is that the sanding equipment will cease operation. The driver is clearly aware that he cannot control the speed of the train and in accordance with the Rule Book issues a train in distress signal by pulsing the warning horn. Based on the assumed stop point the equivalent retardation of the train is 3%g. It is important to identify that loss of WSP speed reference restricts the amount of sand delivered and may as a consequence have caused a resultant extension in braking distance. This problem can be overcome by not making delivery of sand reliant upon having a WSP activity signal when in Emergency brake demand. The difficulty with this approach is that it may deliver a larger amount of sand than having sand delivery based upon WSP activity. AEATR-VTI , Issue 1 54 June 2003

65 Selecting Emergency brake demand by definition indicates that the train is in distress. Under such circumstances this is reasonable justification for delivery of additional sand to the railhead. Silverlink 28/10/02 09:08:01, Unit , Veh: As with the Midland Mainline events, this SPAD event occurred on one of two particularly bad days, during the last leaf fall season. In this event the driver is initially travelling at full line speed. The TOC have provided some indications on the plot of the circumstances as they evolved. Presumably aware of low adhesion conditions, the driver selects step 2 brake at the double yellow aspect signal. This almost immediately results in considerable WSP activity of which the driver would have been aware. The speed reference signal taken from axle 2 rapidly enters a deep slide. After approximately 12 seconds of WSP activity the driver selects step 3 brake to take advantage of sanding. After a further 12 seconds of WSP activity the driver is presumably concerned enough to select Emergency brake. The TOC have identified on the plot that the train had stopped 200m past the Red aspect signal WJ189. It is not possible to define this point exactly on the plot but an approximation has been made. This gives a resultant train retardation of 3.5-4%g. In this example the sand was demanded for most of the event as the WSP internal reference did not fall to zero until near the end of the braking event. The WSP fitted to this unit is a BR Advanced Flywheel design. This has an internal reference which will assume a 3%g deceleration if it loses its external speed references from the axles. As the resultant retardation was around 4%g this indicates that the adhesion level may have been very low and that the sand applied kept the adhesion level with the range of control of the WSP. At low speed it is possible for the internal speed reference of the WSP to be locked out and this appears to have occurred at approximately 09:10:20. The result of this is that the WSP thinks that the train has stopped and the train will continue to slide at low speed with the wheels locked. AEATR-VTI , Issue 1 55 June 2003

66 Thameslink 29/10/02 19:05:18, Unit , Veh: As with the previous SPAD events, this event occurred on one of two particularly bad days, during the last leaf fall season. The driver initially takes step 1 brake, followed very quickly by step 2. At this point the track condition cannot support the demand and some of the wheel sets enter deep slide. The driver presumably aware of the WSP activity selects Emergency brake. The plot shows that there is a slight recovery on the speed reference for axle 2 (used by the OTMR for speed indication). This is most likely to be attributable to sand from the AEA Technology Rail SmartSander improving the railhead adhesion level. There is considerable WSP activity from this point, which would arise from the demand being in excess of the available adhesion. However, it is interesting to note that the WSP is able to keep the speed reference for axle 2 from entering lock. Class 319 units are fitted with BR MkII WSP Deep Slide variant. This profile is consistent with operation of this WSP although in low adhesion conditions it is prone to lose control entering complete lock. This has not occurred on this occasion and may this may be a direct benefit of the sand applied to the railhead. Based upon the assumed cut off point for the event where WSP activity has ceased, the overall achieved retardation is 7.25%g. This is a very acceptable braking deceleration for low adhesion conditions, and may reflect a higher initial condition as well a s the improvement achieved by the variable rate sand application. AEATR-VTI , Issue 1 56 June 2003

67 Thameslink 29/10/02 19:17:45, Unit , Veh: This event is the same unit experiencing low adhesion immediately after the previous event. The driver initially takes step 1 brake, presumably forewarned of low rail head adhesion by the previous event. The plot indicates that the initial achieved retardation was approximately 2%g indicating the severity of the conditions. Even with a step 1 brake demand the wheel sets enter a deep slide, and there is WSP activity as a consequence. In this instance the defensive driving technique employed by the driver may have made matters worse. The driver stays in step 1 brake, but unfortunately due to the low adhesion conditions present, the wheels enter lock before the driver decides to initiate an emergency brake. When the emergency brake is selected by the driver, it is already too late to provide any benefit through increased sand application. Once the WSP internal speed reference is lost by all four axles locking, the WSP activity signal disappears and no sand will be delivered to the railhead. The result of this would be that the driver would longer hear WSP activity. AEATR-VTI , Issue 1 57 June 2003

68 The driver must also have been aware that the unit was not decelerating very quickly and appears to try to give the WSP a chance to recover train speed by releasing the brakes. This results in the WSP reference rapidly recovering to train speed. The driver is then seen to make further brake attempts in step 1, again resulting in momentary slide. The overall retardation for the initial section of low adhesion is 2%g. This figure is with sanding, but at a rate of 1kg/min. If the driver had selected step 2 brake, 2kg/min of sand would have delivered to the railhead and this may have helped to maintain control for a little longer. Thameslink 14/12/02 04:29:25, Unit , Veh: This slightly unusual event is a station overrun reportedly caused by detergent on the railhead deposited from cleaning of the station canopy. The event resulted in an overrun by one coach length (20m). The driver initially takes step 1 brake, followed quickly by step 2 and then step 3 brake. The step 3 brake may be a response by the driver to WSP activity which started when step 2 brake was initiated. It is interesting to AEATR-VTI , Issue 1 58 June 2003

69 note that there is no WSP activity on BDV1 and that it only occurs on BDV2 for a approximately 5 seconds. This implies that axle 1 which is the leading axle was maintaining true train speed. This is doubtful, it is normal for the leading axle to experience worse conditions than the trailing axles. This may be an indication that the WSP dump valve was not operating due to a WSP fault or that the OTMR input from the dump valve may be defective. Even so the overall achieved deceleration is approximately 6%g. For most of the event a step 3 (9%g) was requested. Detergent in large quantities can result in a lowering of the rail head adhesion level to around 5%g. In this instance it is unlikely that the contamination was continuous or over a particularly long section of track. Consequently, it has had a limited affect on the overall retardation. Some of the improvement could be attributable to sand delivered to the rail head. AEATR-VTI , Issue 1 59 June 2003

70 Appendix B Analysis of Braking Performance using LAWS Data AEATR-VTI , Issue 1 60 June 2003

71 2 Introduction This report gives information about a study of LAWS data for Railway Safety. The aim of the study was to investigate the minimum braking performance of sander fitted trains during low adhesion conditions in order to allow appropriate specifications for the ERTMS signalling system to be drawn up. 3 LAWS Description 3.1 What is LAWS LAWS is a real time passenger vehicle based system which identifies track sections with low adhesion by monitoring the traction and braking performance of the vehicles on which it is fitted. This information is then automatically transmitted to a central computer so that remedial action can be taken. 3.2 What problem does LAWS address? The poor braking and traction performance of trains under low adhesion conditions can be a great cause for concern, particularly in relation to safe operation of a train service. Further, many Train Operators rely on high acceleration and deceleration rates to operate a satisfactory service. Such low adhesion between wheel and rail causes extensive problems for rail operators, AEATR-VTI , Issue 1 61 June 2003

72 infrastructure owners and maintainers throughout the year but is a particular concern during the autumn leaf fall season. Deterioration in adhesion is often caused by a combination of railhead contamination and moisture. The principal source of contamination is usually fallen leaves, whilst moisture on the railhead can be caused by rain, humidity and fog/mist. However the relationship between amount of water and level of adhesion is not well understood. By monitoring the adhesion performance of trains over a rail network, drivers can be more easily forewarned of the likelihood of low adhesion conditions, and additional preventative measures, such as sanditing, can be targeted more accurately. Further, more qualitative information can be collected about the types of low adhesion events that do occur and their effect on the safety and punctuality of the train service. 3.3 How does LAWS work? The AEA Technology Rail LAWS system consists of an office based database and information display system (IChex ) linked by cellular radio to a permanently mounted on-train LAWS computer ON TRAIN Equipment The on-board computer is typically mounted underneath a seat in the passenger coach. Motor vehicles are the best option as then the connection to the WSP system can receive indications of wheel slide or spin activity. Two aerials are provided (one each for the navigation and communications) mounted on the longitudinal centreline of the roof. AEATR-VTI , Issue 1 62 June 2003

73 Additional cable connections are made via suitable interfaces to a train speed signal, the traction interlock (trains with power doors) and the driver s direction selector On-Board Software The LAWS on-board computer software performs the following tasks: collect raw data from the WSP system and other trains wires. record speed and position (using Ordnance Survey co-ordinates). transmit real-time low adhesion event information. undertake a range of computer house keeping tasks to ensure reliably. The LAWS computer monitors WSP activity of the vehicle during braking and when the vehicle is under power. The data is recorded from a single vehicle and this is assumed to be typical of the whole train. This assumption is most accurate for short trains operating in very low adhesion when the adhesion variation (conditioning) down the train will be small. Any WSP activity is recorded and classified against brake step or power notch in order to identify both wheel slide and wheel spin events. The time span of the WSP activity is also recorded along with train speed location, and heading. The determination of geographical position is achieved using a combination of satellite navigation with distance derived from train speed. The satellite navigator is used to determine which section the train is occupying and the tachometer pinpoints position along the track. 3.4 INFORMATION DISPLAY SYSTEM Information from LAWS is displayed through the IChex suite of software, used for the display of all infrastructure-related data produced by AEA Technology Rail s monitoring systems. The IChex suite of software runs on a PC with the Windows operating system (Windows 96, 98 or NT) and allows the user to: Receive warnings in real-time about low adhesion events. View the development of low adhesion conditions over time. Identify and view event severity via colour coded map sections. Search the database for particular events or sites. AEATR-VTI , Issue 1 63 June 2003

74 The record of low adhesion events is based on defined sections of operating routes. As fitted units report low adhesion events on particular sections a map of low adhesion sites is built up, and the severity of sites can be assessed by the number and type of incidents that occur. 3.5 History of LAWS The first system was provided for Thames Trains in 1996 and this was followed by three more the following year. In 1999 an additional thirty systems were provided and fitted to a wide range of vehicles operating over central and southern England. 3.6 Data Generated BY LAWS As stated above LAWS monitors the operation of the vehicles on board Wheel Slide Protection (WSP) system, however it also takes inputs from other vehicle systems to enable characterisation of the low adhesion event. AEATR-VTI , Issue 1 64 June 2003

75 The basic data recorded by LAWS consists of Time at which the WSP was activated Position at which the WSP was activated in terms of Ordnance Survey co-ordinates. Vehicle Speed when the WSP was activated Maximum brake step during the event Length in seconds of WSP activity Distance vehicle travelled during WSP activity Vehicle speed at end of WSP activity From these parameters further information such as average deceleration can be calculated to determine the severity of low adhesion at any particular location. AEATR-VTI , Issue 1 65 June 2003

76 4 Method of analysis In order to investigate the braking performance of LAWS fitted vehicles a study of all the events recorded by the LAWS systems between September 2002 and January 2003 was carried out. The data collected was first filtered to identify those vehicles fitted with sanders as opposed to those vehicles not fitted with sanders. A second filter was applied to identify all events where the sander should have operated, this was done by identifying braking events where the driver selected step three during the WSP active period. From the data collected from LAWS fitted vehicles in autumn 2002, analysis showed that a small number of sander fitted vehicles had experienced sufficiently bad low adhesion for the driver to select braking in step 3. The data collected from these events was analysed in more detail to investigate typical low adhesion properties. Only seven events with step three braking and WSP activity were identified as applicable to this study suggesting that defensive driving techniques were being used to good effect by most drivers. In the course of the data analysis it became clear that a more precise definition of the low adhesion event severity was required. Depending how the start and end of events was defined could lead to different answers so it was necessary to formulate a consistent definition. The daily summary data recorded by LAWS was discovered to be insufficient in its current format. Detailed analysis of the entire braking event was required to compare the brake steps selected by the driver and the resultant achieved deceleration. In most of the braking stops the driver did not go straight to step 3 and leave the brake in that state so the overall deceleration requested by the driver was lower than a step 3 demand. To gain a fair comparison between low adhesion events a comparison between the nominal deceleration requested by the driver with that achieved by the train during braking was compared. Nominal deceleration requests were calculated using the values in the following table: AEATR-VTI , Issue 1 66 June 2003

77 Brake step Nominal Deceleration 0 0 %g 1 3 %g 2 6 %g 3 9 %g Emergency 12 %g The nominal deceleration request was calculated by summing the product of the deceleration for that step and the time in that step and dividing the resultant total by the total time of the low adhesion event. ( Brakestep Time) Re quested Deceleration = 3 Total Time Unlike with the traditional LAWS calculations, where the low adhesion event is defined as the time for which the WSP system is active, events were defined as the time for which the reference speed was lower than the estimated vehicle speed and in which brake step 3 was selected for some or all of the time. This definition was necessary because LAWS gets its speed input from the wheelset tacho, which does not reflect true vehicle speed during braking. In order to calculate actual decelerations the actual vehicle speed is required so this had to be estimated using the recorded data. The graph below shows details of a typical Low Adhesion event that a train may experience. AEATR-VTI , Issue 1 67 June 2003

78 The driver selected brake step 1 for a short period however the vehicle did not slow. After a short period of coasting, the driver then selected brake step 3 and the wheels decelerated rapidly (shown in Dark Blue). Sensing this impending lock-up the WSP was activated. The combination of step three braking and WSP activity would have caused the sander to activate. Shown by the varying yellow line the WSP was active to try and regulate the wheel speeds. The driver reduced the brake demand to step two and the wheels stopped sliding indicating that the adhesion had dropped below that adhesion demand level. The driver was then able to slow the vehicle further under normal braking. The red dotted line on the chart has been calculated from the original data to give an indication of the actual deceleration of the train in step three braking. In this particular example ( Brakestep Time) 3 Re quested Deceleration = = 7. 5% g Total Time ( Change in Speed) Actual Deceleration = = 3. 61% g Time This shows that the vehicle achieved a low level of deceleration for the particular event that was significantly less than the driver requested. AEATR-VTI , Issue 1 68 June 2003