Derailment investigation
Dr Mark Burstow, Network Rail Technical Fellow Principal Vehicle Track Dynamics Engineer
Providing technical leadership Overview
• History- when and how did derailment investigation really develop?
• Causes of derailment and contributory factors
• The investigation process o Evidence collection
o Analysis
Derailment Investigation December 2020 Page 2 How derailment investigation became a science
The 1960s became the era of the freight train derailment • Particularly, short wheelbase 4-wheeled wagons on plain track • A worrying trend o 11/2/1961: Between Rugby & Lutterworth o 1/4/1963: Weedon o 21/1/1966: Steventon o 31/7/1967: Thirsk
450
400 Passenger
350 Freight
300
250
200 Derailments Derailments 150
100
50
0 1950 1960 1970 1980 1990 2000 2010 2020 Year
Derailment Investigation December 2020 Page 3 Vehicle dynamics research
Newly formed BR Research department took on challenge of finding cause for these derailments • Developed theoretical and mathematical basis for wheelset dynamic instability (‘hunting’)
• Identified that some freight vehicles were becoming unstable at speeds as low as 20mph, depending on condition of suspension and wheel wear Understanding of vehicle dynamics provided better explanation of the causes of derailment • Separate group within BR Research formed to undertake investigations • Became part of AEA Technology Rail following privatisation, then DeltaRail
Derailment Investigation December 2020 Page 4 So, where do we come in?
Following the Hatfield accident in October 2000 • Railtrack created its own ‘in-house’ expertise in vehicle dynamics and vehicle/track interaction o Initially to investigate rolling contact fatigue (RCF) o Support the cross-industry Vehicle/Track Systems Interface Committee (V/T SIC)
In 2010 DeltaRail announced they would withdraw from track consultancy work • Including supporting derailment investigation (‘call-off’ contract with NR)
In 2011 Head of Track asked if NR’s wheel/rail interface team could provide a similar service
Derailment Investigation December 2020 Page 5 Our first derailment
Edinburgh Princes Street Gardens, 27th July 2011
Derailment Investigation December 2020 Page 6 Where have we been since then?
43 incidents in 9 years • Roughly 1 every 2½ months
Derailment Investigation December 2020 Page 7 Where have we been since then?
43 incidents in 9 years • Roughly 1 every 2½ months But it’s not as simple as that • October 2013: 2 Incidents on the same day
Derailment Investigation December 2020 Page 8 Where have we been since then?
43 incidents in 9 years • Roughly 1 every 2½ months But it’s not as simple as that • October 2013: 2 Incidents on the same day • January 2013: 4 incidents in 1 week
Derailment Investigation December 2020 Page 9 Where have we been since then?
43 incidents in 9 years • Roughly 1 every 2½ months But it’s not as simple as that • October 2013: 2 Incidents on the same day • January 2013: 4 incidents in 1 week And we’ve even been to the same place twice! • April 2014 • June 2015
Derailment Investigation December 2020 Page 10 So, how does it work?
Notifications of incidents obtained from National Operations Centre • Liaise with the Route/Region team regarding attendance Sometimes we get invited some time after the event • Not ideal! Our purpose • Impartial investigation • Evidence collection • Technical analysis to explain the mechanism and course of derailment • Assist any formal investigation with technical input What we don’t do.. • Find a scapegoat! • Lead/chair investigation panels • SPADs, collisions, level crossing incidents… We are not in competition with, or replace, RAIB • Usually work together • RAIB may ask for our assistance: at site or with analysis
Derailment Investigation December 2020 Page 11 The investigation process: 1- mobilise!
Derailment Investigation December 2020 Page 12 The investigation process: 1½- arrive!
Derailment Investigation December 2020 Page 13 The investigation process: 2- evidence collection
Two crucial questions to answer • What happened? • How did it happen?
Derailments very rarely (if ever) have a single cause • Result of a combination of vehicle and track contributory factors
Derailment Investigation December 2020 Page 14 What are the mechanisms of derailment? Gauge spread
Flange climb
S&C
Derailment Investigation December 2020 Page 15 1. Flange climb derailment
Process whereby a wheel climbs up the side of the rail and the flange tip runs across the head of the rail to drop onto the field side of the rail Flange climb is a risk when lateral forces (Y) on the wheel overcome the vertical forces (Q) • Quantified by Nadal’s limit Wheel • Nadal’s limit gives critical Y/Q ratio: Y/Q above this Wheel load, Q limit indicate derailment is a risk Rail • Nadal’s limit depends on Lateral force, Y Normal (reaction) o wheel/rail contact angle force, N Tangential (reaction) o friction between wheel and rail force, F
Contact angle, q
Derailment Investigation December 2020 Page 16 How do we get a high Y/Q?
Increase ‘Y’ (lateral) force • Tight radius curve • Lateral alignment track faults/variations in combination with tight radius curve Decrease ‘Q’ (vertical force) • Track twist • Slow speed- cant excess • Vehicle loading • Vehicle suspension o Twisted bogie frame Many derailments occur on twist faults on tight radius curves with freight vehicles at low speed • Combines many of the features above!
Derailment Investigation December 2020 Page 17 2: Complete wheel unloading- Cyclic top
Vehicle dynamic behaviour is excited by the top profile of the track • Derailment becomes a risk when vehicle’s natural ‘bounce’ or ‘pitch’ frequency matches, or comes close to, the frequency of dips in the rail • Depends on wavelength of track features and vehicle speed
Excitation of vehicle in vertical plane can fully unload the wheels • Wheel ‘bounces’ onto or over the rail
More likely as speed increases as there is more energy to drive the behaviour
Derailment Investigation December 2020 Page 18 Cyclic top track geometry and vehicle response
The response of each vehicle varies with wavelength • The wavelength at which the most unloading occurs varies with speed, suspension design and loading condition • The maximum amount of wheel loading also varies
0.9 Freight 1 (Laden) Freight 1 (Tare) ) 0.8 Q
/ Freight 2 (Laden) Q 0.7 Freight 3 (Laden) 0.6 Freight 3 (Tare) Freight 4 (Laden) 0.5 Freight 4 (Tare) 0.4 Freight 5 (Laden) 0.3 Freight 5 (Tare)
0.2 Wheel unloading Wheel unloading (Delta 0.1
0 40 50 60 70 80 90 Speed/mph
Derailment Investigation December 2020 Page 19 3: Gauge spread
As curve radius reduces • Lateral forces on leading wheelset increase and act to push the rails apart • Derailment occurs when sleepers/fastenings are no longer able to support the lateral forces • Usually outer rail of curve moves, allowing wheel on inside of curve to drop in to derailment
50
45
40 High PYS, Wset 1
35 High PYS, Wset 2
30 Low PYS, Wset 1
25 Low PYS, Wset 2
20
15
Gauge spreading Gauge force/kN 10
5
0 0 500 1000 1500 2000 Curve radius/m
Derailment Investigation December 2020 Page 20 4. Derailments on switches & crossings (S&C)
Switches & crossings provide a real ‘assault course’ for a railway vehicle • Need to provide guidance • Rapid changes in curvature, alignment and rail profiles Different wheel profiles make contact with the switch at different points • Although some wheels can negotiate the switch ok, others may result in the wheel flange lifting on to the top of the switch rail
New wheel profile Worn wheel profile
Derailment Investigation December 2020 Page 21 Poor switch rail profiles
Worn profile of switch blade shows a shallow angle • High friction (poor lubrication) and low contact angle reduces Nadal’s limit (Y/Q limit) • Easier for a vehicle to generate a Y/Q which exceeds Nadal’s limit and flange climb
0
700 500 300 100 800 600 400 200
Derailment Investigation December 2020 Page 22 Derailments on S&C: Crossing issues
Guidance at crossings • Crossing nose geometry and poor track quality: allows wheel to climb onto crossing nose
Derailment Investigation December 2020 Page 23 Challenges on site
Derailment Investigation December 2020 Page 24 The investigation process: 2- evidence collection
How do we separate damage as a consequence of derailment from that which was causal?
Derailment Investigation December 2020 Page 25 Point of derailment: flange climb marks
Derailment Investigation December 2020 Page 26 Flange climb marks: what can we learn
Long flange mark: Short flange mark: Low lateral forces acting- flange climb due to reduced High lateral forces forced wheel over rail vertical load: possible track twist or vehicle imbalance?
Derailment Investigation December 2020 Page 27 Point of derailment: wheel tread corner marks
Derailment Investigation December 2020 Page 28 Sources of evidence: track
Develop a keen eye for detail!
It can sometimes be useful when parts of the train are blocking access to the PoD
Derailment Investigation December 2020 Page 29 Sources of evidence: damage to track
Broken rail- cause or effect? It’s vital that you find all the ‘bits’!
Derailment Investigation December 2020 Page 30 Absence of evidence?
Derailment Investigation December 2020 Page 31 Sources of evidence: damage to track components
How many wheelsets derailed and where did they run?
Derailment Investigation December 2020 Page 32 Sources of evidence: track geometry
Derailment Investigation December 2020 Page 33 Sources of evidence: rail profiles
Direction of travel
Sleeper 7 Sleeper 20 Point of derailment Sleeper 50 Sleeper 100 Sleeper -5
Derailment Investigation December 2020 Page 34 Sources of evidence: wheels
Derailment Investigation December 2020 Page 35 Sources of evidence: wheel profiles
Wheelset 5 Compliant with allowable wear limits? Uneven wear could be an indicator of vehicle suspension problem? Contact angle with measured rail profile? Wheelset 6
Wheelset 7
Wheelset 8
Derailment Investigation December 2020 Page 36 Sources of evidence: vehicle condition
Derailment Investigation December 2020 Page 37 Sources of evidence: vehicle load
Uneven loading can make a vehicle more susceptible to flange climb on a track twist
Derailment Investigation December 2020 Page 38 Sources of evidence: other sources of data
CCTV footage OTDR data: train speed and handling
Wheel impact load detector sites: vehicle loading 25% 20% Left rail Wheelchex 15% Gotcha 10% 5% 0% -5% -10% -15% Wheel load variation/% -20% -25%
Axle number
Derailment Investigation December 2020 Page 39 The investigation process: 3- analysis and interpretation
40 30 Direction of travel Track geometry data 20 10 • Variations in crosslevel, track twist 0 and curvature -35 -30 -25 -20 -15 -10 -5 -10 0 5 10 15 20 25 -20 • Compliance with standards, Crosslevel/mm -30 Crosslevel Hand survey -40 any features to present an increased Level survey 4-hole joint Crossing nose Sw itch tip -50 derailment risk? -60 3m Distance relative to PoD/m60 • Track twist can reduce load on outer 5m 2m 40 wheel Hand measurements (3m) Hand measurements (5m) 20 o What is the actual twist seen by the vehicle, not Hand measurements (2m) just that over 3m? 0 -35 -30 -25 -20 -15 -10 -5 0 5 10 15 20 25 Twist/mm • Effect of the combination of track -20
curvature/alignment and vertical profile -40
on vehicle behaviour? 15 -60
Distance/m 5m 10m 10 20m
Curvature/1/km 5
0 -25 -20 -15 -10 -5 0 5 10 15 Distance from PoD/m
Derailment Investigation December 2020 Page 40 The investigation process: 4- vehicle dynamics simulation
Demonstrate/validate the mechanism of derailment
Amber trolley 35 Spot measurements • Need a ‘model’ of the vehicle- include all its suspension stiffness and Sw itch tip 25 Rail joints damping properties 15 Lubricator or checkrail Miniprof 5 Crosslevel/mm Explore and understand the contributory factors -5 0 10 20 30 40 50 60 70 80 90 100 -15 • Speed, wheel/rail friction, track gauge/alignment, vehicle loading, Distance/m wheel profile
Derailment Investigation December 2020 Page 41 Vehicle dynamics modelling: Flange climb on a curve
Testing the effects of different wheel and rail profiles • Helps to explain why that vehicle derailed
18 16 14 12 Measured w heel & rail 10 P8 & measured rail rail 8 Measured w heel & 113A rail 6 4 2 Wheel Wheel lift,1, wheelset high 0 -25 -20 -15 -10 -5 0 5 10 15 20 25 Distance from PoD/m
Derailment Investigation December 2020 Page 42 Flange climb example: Sensitivity to friction
• Very dry rail- 18 rough worn surface 16 • Very dry wheel- 14 Tread friction=0.4 12 newly turned mu=0.1 10 mu=0.2 mu=0.3 rail 8 • High friction reduces the critical mu=0.4 6 mu=0.45 Y/Q ratio, increases derailment risk mu=0.5 4 2 Wheel Wheel lift,1, wheelset high 0 -25 -20 -15 -10 -5 0 5 10 15 20 25 Distance from PoD/m
Derailment Investigation December 2020 Page 43 The investigation process: 5- report and recommendations
Derailment Investigation December 2020 Page 44 So, what can we conclude?
No two derailments are the same • Treat every day as a school day!
You can’t collect too much data • You only get one chance before the site gets cleared up • …and when it’s gone….it’s gone
Never stop asking ‘WHY?’ • Keep an open mind o Don’t jump to conclusions • Assess ALL the evidence, not just the bits that fit your theory! o Assess everything yourself o Think, think, think!
Derailment Investigation December 2020 Page 45 Providing technical leadership