Return to Archive -by date - by topic.
Three progress reports later and we still have more questions than answers as retribution takes priority over cause.
It is all very to criticise the Health & Safety Executive over its handling of the Potters Bar derailment inquiry, but the railway inspectors at the sharp end are working under considerable handicaps. In effect, they are third in the pecking order.
Leading the inquiry are the British Transport Police for whom the accident is a potential crime, with, at the least, a prosecution for manslaughter in the offing. Next in the queue is the HSE for whom the accident is a potential breach of Health & Safety at work regulations. Only then do we get to the Railway Inspectorate who are trying to find out the cause. And they are under the oversight of an Investigation Board, none of the members of which have first hand experience of working railways.
In terms of accident investigation it is about the worst system you could have. It is why we need the Rail Accident Investigation Branch and why we should all support Caroline Griffiths, the Branch's new Chief Inspector Designate in the turf wars to come. Be assured that she has no doubt as to the challenge she faces.
Is it really a problem? Yes, and the third Progress Report published in May provides further confirmation. Here's a quote from this latest document. ‘As the BTP are in the lead, HMRI inspectors have concentrated on reviewing records and analysing information arising from the BTP investigation and the Health & Safety Laboratory(HSL) work'. As readers will recall, the crossover which included points 2182A which failed, was cut out and taken to HSL at Buxton for examination. The resultant lifting may have contaminated the evidence.
Technical investigation led by HSL has also involved ‘outside experts', such as AEA Technology Rail who were responsible for derailment analysis for Railtrack following separation from British Rail in 1994. ‘The HSE has <ital>also involved<ital> (my italics), railway inspectors from HMRI with specialised knowledge about various aspects of the derailment'. Anyone who has read the commendably detailed report into the Great Heck derailment will appreciate the waste this sentence implies.
Thus, we come to the report of the Investigation Board, whose members are the Deputy Chief Inspector lf Nuclear Installations (Chairman), HSE Director Scotland, the former Group Technical Director of the British Airports Authority who retired in 2001 and an aviation engineer who became Development Director for British Gas and retired in 1995. And this august body seems to have spent much of its time studying the wood, when accident investigation is all about trees, or even the twigs and leaves thereof.
More of the wood another time. This month I will concentrate on the cause of the accident.
Quite‘Whereas the immediate faults were obvious, and the derailment sequence can be determined from the physical evidence, establishing how and when points 2182A came to be in the condition in which they were found has proved to be complex. This is especially so as we are informed that the BTP has so far been unable to obtain a full picture from speaking to those who may have been able to shed some light on how the points came to be in their pre-derailment condition'. Third Progress Report |
In general, the report confirms the condition of the points as described in this column in March. The most significant development is that the flangeway gap – the space between the right hand switch rail and the right hand stock rail, had a minimum width of 30mm just beyond the rear stretcher bar (tiebar 2 on my diagrams). However HSL reminds me that the inner nuts on the right hand end of this stretcher were so loose that they could have moved ‘significantly' during the accident.
On the basis of the ‘shiny thread', these nuts are believed to have been screwed outwards from their correct position, closing the flangeway gap from its nominal 50mm. These was speculation that this gap could have been small enough to squeeze passing wheels.
In this event, flange contact with the switchrail would have resulted in the wheels of passing trains applying high frequency repeated cyclic forces to the lock stretcher. This would have accelerated its fatigue failure.
But HSL also tell me that no contact marks were found on the back of the flanges of the right hand wheels of the first two coaches of the Class 365 involved. However there were evidence marks on the back flange of the leading right hand wheel of the leading bogie of the third coach.
This suggests that the flangeway was wide enough for wheel flanges to pass without contact until the lock stretcher, which was all that was holding the switch blades together, failed as the leading bogie of the third coach was passing through the flangeway. As the broken lock stretcher worked out of its insulating sleeve it released the right hand switch blade which sprang out and hit the back of the leading wheel flange, which then held the flangeway open for the trailing wheel on the bogie to pass through.
Of the remaining bogies, two derailed in a straight ahead direction, but the trailing bogie of the fourth coach became airborne and the left hand wheels dropped to the left of the switch rail and took up a normal running position on the diverging rail. This allowed the right hand wheels to run round their switch blade, swinging the back of the coach to the left, broadside onto the direction of travel.
While it appears that there was no contact between flange sand switchrail before the failure, the report reveals that there was damage on the right hand outer corner of the rail head of the right hand switch rail. This suggests, says the report, that at some stage in the life of the points there had been flange contact in the flangeway, sufficient to squeeze wheels upwards. At which interesting point the HSE explains that ‘further investigation of this feature would have involved destructive testing and was not therefore carried out for reasons of evidence preservation'. Just take a rant for granted here to save space.
So the crux of the matter remains the true condition of the points in the weeks leading up to the derailment.
HSE says that ‘no evidence has yet been established' that sabotage or deliberate unauthorised interference was the direct or root cause of the derailment. All the evidence indicates that the condition of the points was so poor that they were not ‘fit for purpose'.
Four main factors contributed to the failure:The poor condition of the backdrive and other compoments; the loss of nits from the right hand end of the rear stretcher bar; the loss of nuts from the left hand end of the front stretcher bar; and the fracture and disengagement of the lock stretcher bar. It is probable that the lock nut on the outer right hand end of the rear stretcher bar had not been tightened against the outer main nut before the incident. It is probable that the outer insulating bush had been missing from the left hand end of the front stretcher bar for some time before the failure occurred; it would not have been possible to accommodate both outer nuts on the end of this stretcher bar. The lock stretch fracture was due to fatigue cracking which occurred progressively over a period of weeks or months. This set of points had been poorly maintained and was out of adjustmenmt in some respects; this is likely to have increased the possibility of vibration induced damage. Other points in the area were found to have similar maintenance deficiencies, although none were in such a poor condition as 2182A. This indicates that a wider problems existed. HSE Investigation Board conclusions |
Jarvis, in a formal comment on the report remained adamant that it ‘maintained the points in accordance with long-established industry procedures' and claimed that the report ‘gives no evidence of a failure to maintain the points within the required maintenance regime'.
On this issue, Jarvis says ‘We have provided statements, documentary evidence and witnesses to the investigating authorities which show that the points were maintained in accordance with laid down standards. The maintenance of the points was carried out by trained and competent personnel, in accordance with industry guidelines'. The company concludes ‘the investigation has yet to establish what happened in the seven weeks between the inspection and May 10. Therefore it is still uncertain as to what happened to cause the points to fail so catastrophically. We note that the report does not draw any conclusions on ultimate cause'.
HSE confirms that the Omnicom inspections in April and July 2001 showed 2182A to be in a satisfactory condition. The front stretcher bar (Tiebar 1) was centred and not fully over to the right hand side as found. The assumed escapement gap, found in the back drive rod from this stretcher, was not present.
We now jump forward to a formal joint inspection by Jarvis and Railtrack engineers on March 20 when the points were found to be in good condition. But, as a respected manager remarked to me ‘have we forgotten what “good” is'.
There is an issue of perception here. When I was briefly in the construction industry, I could not imagine how houses got built in the muddy chaos of construction sites, compared with the factories I was used to. Similarly, point mechanisms, exposed to the elements, atmospheric pollution, oil from old diesels and flying ordure from train toilets, always looks to me like something out of a scrapyard compared to, say, exposed bit under my occupational therapy car.
So what should a ‘good condition' set of points look like? I am assured that the 20 March inspection was made by experienced engineers. And, significantly, it is now SOP for Jarvis track teams to take digital before and after photographs of S&C worksites, which are even encrypted to prevent tampering.
But, seven weeks after this inspection, HSE made 83 ‘observations or findings' about the state of points 2182A before they were dismantled for investigation. Of these, 15 were classified as having made a major contribution to the derailment ranging from missing nuts to ‘indicators' of poor maintenance or incorrect setting up or both. Three observations referred to the rogue escapement in the back drive. As you will recall, one theory for this is that it was needed to maintain detection when the tie rods were wound out in the Network Rail and Jarvis tests on a spare set of points.
And on the subject of tampering, where Network Rail and Jarvis reckoned 20-30minutes to get a correctly set-up set of points in to the pre-derailment condition – and I timed a video showing it being done, HSE reckons it can be done in 6min under ideal conditions.
Compared with the relevant standards and drawing, around half these initial observations, that's 40-ish, were interpreted by HSE as non-compliance with standards, codes of practice and drawings. They included installation faults, maintenance problems and incorrect adjustment.
So once again, we have to ask whether the 20 March inspection team knew what ‘good' was because their report did not mention any signalling faults or defects. Given that this was Railtrack on site with its contractors I would have expected them to be ultra picky
Also in March 2002, says HSE, a planned joint per way/signalling maintenance visit to 2182A ‘appears to have taken place without permanent way personnel being present'. Did this mean that faults outside the immediate concern of the S&T technicians were not noticed or went unrecorded? Since servicing the points includes checking detection with feeler gauges while cranking the points over manually surely a partly dismantled set of points would have been noticed?
Certainly, on 1 May, a signalling technician on a planned maintenance visit recorded ‘2182A points needs a new rubber bush on the back lock stretcher'. Note the imprecise terminology – there is only one lock stretcher. There was no reference to the state of the nuts on the stretcher bars.
And, despite 12 months' investigation, ‘no records have yet been identified' detailing remedial work required or carried out additional to planned signalling related maintenance work on 2182A. Nor has any maintenance backlog been identified although there is ‘some evidence', that not all maintenance work recorded was completed'.
Finally, there was the weekly visual inspection by the track worker the day before the derailment which did not register any problems. Correction, there was also the report of ‘rough riding' over the points south of Potters Bar by a WAGN supervisor that evening.
Commuters had been aware of a rough ride approaching Potters Bar for some time. But it seems unlikely that the state of the points would have been a factor when set ahead.
Readers who have followed this column's concerns over safety critical communications will not be surprised that the initial message – reported to a booking office clerk at Stevenage , was ‘misunderstood and then became confused' in the course of being relayed onwards. As a result the per-way inspection team was sent out to the wrong section of track and then carried out the inspection under red zone conditions without the required protection.
Working on a 115mile/h section of railway in the dark with no pee-wee and no temporary speed restriction to 20mile/h with lookouts posted is not conducive to a long hard look at anything.
In general the technical content of this latest report is inadequate. For example, when the points were dismantled at Buxton the torque required to unfasten the nuts still in place on the connecting rods and stretcher bars was measured. The range was 5ft lb to 200 ft lb. Now 5ft lb is the sort of torque you would use on a small holt, say holding down a cover on a car engine, going into aluminium where you might strip the threads if a bit cack-handed. On the other hand the biggest nut I have done up on my occupational therapy car needed 130ft lb and because, at that time, I was recovering from a detached retina I got eldest son to tighten it and he huffed and puffed a bit.
So why no diagram of the points, with the nuts numbered and a table with the torques required? Or even a note of the optimum torque expected?
Similarly, why no diagram of the points as examined with the 83 observations identified and listed. All we get is the same generic diagram as before with three sets of lost motion escapements, including one in the front connecting rod in the very position described by the report as ‘erroneous'. This column is still the only source of a diagram showing the condition of the points immediately before the derailment.
Crucially we would be nearer to an answer to how long the points had been in the pre-derailment condition if we knew how long it would take the lock stretcher to suffer a fatigue failure with one of both of the stretcher bars ineffective due to the outer nuts coming off.
Metallurgical examination showed that failure of the lock stretcher had been preceded by fatigue or corrosion assisted fatigue under reverse bending stress conditions. Fatigue crack growth had been occurring for some time, but the last period of growth was quite recent ‘having taken place no more than a few days prior to the derailment'. There was a small area of ‘overload fracture' which was where the stretcher bar finally gave way when the crack had grown to the extent where there was insufficient metal to bear the loads imposed by the missing stretcher bar nuts. The small size of this area, says HSE, indicates that the nominal stress was low.
So just as the unauthorised adjustment or tampering theory was looking increasingly untenable we learn that while fatigue cracking had been going on for some time, the final cracking had happened ‘no more than a few days' before the accident. So something had changed recently.
HSL also reassembled the points with a new lock stretcher and stretcher bars (presumably less the missing nuts) but the same back drive. The lock stretcher was strain gauged and the outputs measured when the points were operated. Fatigue calculations based on these figures show that the cyclic stress caused by the changing levels of force as the points are motored across was above the threshold level for fatigue crack growth.
Obvious question. Why is there not a set of points, with an identical configuration to 2182A, even now being motored to and fro, to see just how fast cracks propagate in the lock stretcher?
According to HSE, it was appreciated at ‘an early stage of the investigation', that the role of vibration in the failure needed to be considered. A working group set up by HSL agreed a three part programme including the design and construction of a laboratory test rig to carry out 'simplified vibration experiments' on stretcher bar assemblies. When the third Progress Report was being written construction of the rig was ‘underway'.
Hell's teeth. On 10 January 1954 de Havilland Comet ‘Yoke Peter' exploded in mid air and plunged into the Mediterranean . Metal fatigue was suspected and a another Comet was put in a pressurisation test rig at Farnborough. While these tests continued Comets went back into service on 23 March. Two weeks later ‘Yoke Yoke' was lost.
During the summer, the test rig produced a fatigue failure and in August a piece of wreckage from ‘Yoke Peter' recovered from the sea bed provided conclusive proof. The formal legal inquiry into both accidents began two months later and the report was published in February 1955.
Compare and contrast, as they say. In seven months in 1954, the cause of a major accident at the leading edge of technology was demonstrated and confirmed. In 2003 we have no real understanding of the failure of a piece of long established basic railway kit. And vibration testing can wait until we can explain the history of those fatigue cracks which led to the failure, not to mention that evidence of an earlier flangeway closure
After a year?The Board notes that [the HSE] investigation into management systems and responsibilities associated with the condition of points 2182A still have some way to go. Owing to the primacy of the manslaughter investigation, this work has been restricted so far largely to examining records and standards to assist the investigating BTP to prepare for further lines of inquiry and interviews. We (the Investigation Board) also note that this process was held up by the difficulties experienced by the investigating officers in examining all the expected relevant files and documents'. Are you surprised?Establishing how and when Points 2182A came to be in the condition in which they were found has proved to be complex. This is especially so as we are informed that the BTP has so far been unable to obtain a full picture from speaking to those who may have been able to shed some light on how the points came to be in their pre-derailment condition. Furthermore some of the records are not as full as the investigators expected.' But not in a name and blame cultureThere is a power (under Section 20 of HSAWA) to require an employee whom an [HSE] Inspector has reasonable cause to believe to be able to give information relevant to the investigation, to answer questions truthfully. The answers to such questions cannot be used in evidence against that employee. This is an important tool in the management of an employer's safety management and systems when responding to incidents. The legal basis for the work of the RAIB and its protocols with other investigators will need careful preparation in this context if the new institution is to maximise its contribution (my italics RF)
All Quotations from the HSE Investigation Board Report |
Meanwhile, the final HSE report will not be published until after the conclusion of any legal proceedings or after they have been rule out.