Frequently Asked Questions (FAQs) about Flixborough

This webpage responds to questions raised about “Flixborough Revisited” (posted here in January 2005 and published in The Chemical Engineer in April 2005) and “Flixborough - Some Additional Lessons” (presented at an IChemE Symposium in December 1975 and reprinted in The Chemical Engineer in May 1976).

The event sequence appears as Fig. 1 of “Flixborough – Some Additional Lessons”. Within the Q&As, illustration references are to Figs. 1-6 of “Additional Lessons” and Figs. 7-10 of “Flixborough Revisited” and readers will find it helpful to have these to hand. The FAQs are grouped into four “topic areas”:

1          The main explosion

2          The fin-fan coolers explosion

3          Damage to the 8" line

4          How the 20" line failed

These FAQs have been written by chemical engineers for chemical engineers and contain no new metallurgical insights or mechanical failure theories. As chemical engineers we have tried to reconcile the findings of mechanical engineers and metallugists (on matters wholly within their expertise) with our findings on matters wholly within our expertise as chemical engineers.

Researchers wishing to receive further details may purchase a CD with many megabytes of photographs, reports and copies of selected transcripts from the proceedings of the Public Inquiry.

Topic Area 1: The Main Explosion

FAQ 1.1      

If there were flames at the gasket leak and from the creep failure in the 8" line and an explosion in the fin-fan coolers before the 20" line failed, why didn’t any or all of these ignite the escape from the two 28" nozzles?  Do the flash fire, explosion and soot fallout indicate that the cloud did not ignite until it reached the Hydrogen Unit?

Answer

The process fluid had an intrinsic probability of ignition as it was two-phase (96% cyclohexane and 4% water) – a potent electrostatic generator – and also carried an oxidation promoting catalyst.  Either or both, with ample air access, would have led to early ignition.  The 8" line hypothesis presumes that the escape from the gasket leak would be ignited, effectively instantaneously (possibly within milliseconds).

Whilst, in this scenario, the flame directed at the 8" elbow could be quenched when the 8" elbow burst, this also would reignite within milliseconds to burn as the jet fire that consumed the fin-fan coolers and, in turn, caused the explosion that brought down the 20" line.

This escape (from the two 28" nozzles), however, was so huge that any unquenched flames would be confined to the periphery of the cloud as it expanded.  So, whilst early ignition was a near-certainty, the rate of escape (from two opposing 28" nozzles) was so huge that a large expanding volume of cyclohexane in the centre was not consumed and created the air/fuel cloud that caused the main explosion.

FAQ 1.2      

How does this scenario differ from an oil or gas well fire? The volumes escaping from well fires can be many times greater than occurred at Flixborough – but, once ignited, these do not result in subsequent vapour cloud explosions.

Answer

A high-pressure escape normally results in a jet fire (not an explosion) when air is entrained with the fuel, close to the origin of the escape and the air, fuel and flames all travel in the same direction.

At Flixborough the two 28" nozzles faced each other and the two escapes collided to create a rapidly expanding cyclohexane cloud that, in less than a second, engulfed the space between the reactors. Unlike the jet fire scenario, the escape(s) originated wholly within the cyclohexane cloud and without any opportunity for air entrainment and mixing.

Eventually, after about 20 seconds, as radial expansion slowed, flames began to penetrate, accelerating as the temperature rose (the flame and pressure front would raise the internal pressures and, incidentally, marginally raise the UFL and facilitate flame propagation.) Then, possibly by now at sonic velocity, the flames met at the epicentre of the explosion. Within a relatively small volume – unconfined by walls but ‘confined’ by the advancing flame and pressure fronts, the central cyclohexane-air mixture exploded and generated the over-pressure blast that then exploded outwards in all directions.

If an explosion or major deflagration occurs during a fire, it will be due to a release of extra fuel (for example, from a ruptured vessel) that, although initially above its UFL, mixes rapidly with air to form an explosive mixture. This is not an unusual scenario and, in this respect, the Flixborough explosion was by no means unique.

FAQ 1.3

What is your evidence that flames existed before the main explosion?

Answer

Flames were reported by a majority of eye-witnesses, more than 20 being in a position where they could have seen an escape from the 28" nozzles – over 30 seeing flames beforehand (though some may have seen the fin-fan coolers explosion and/or an emission from the 8" line). This evidence for flames prior to the main explosion is overwhelming.

One coherent account from an ex-employee, 3½ miles away at Scunthorpe (location 32 on Fig. 4) describes “a fire burning, with smoke fairly high over Plant Section 25A” with “a very bright flame which seemed to be blowing slightly to left from what seemed to be a gas leak” (before the main explosion) then, "whatever was burning actually burst itself and dropped straight to the ground and spread out across the perimeter of the plant" (the main explosion).  He “did not see a flash from Hydrogen Plant” and believed he would have noticed it if this had occurred. (Note: as the blast wave would take 17 seconds to reach Scunthorpe, his evidence is wholly visual).

FAQ 1.4

Has this been demonstrated by a computational fluid dynamics (CFD) simulation?

Answer

CFD simulations were not then available but have since become the generally accepted method of explosion prediction for unconfined, congested, environments. They have demonstrated that high pressures can arise by the effects of congestion induced turbulence (far higher than conventional wisdom thought in 1974 – far higher indeed than is achieved in a wholly confined, but uncongested, explosion).  CFD has been applied to Flixborough assuming stoichiometric combustion conditions within the cloud created by the failure of the 20" line alone.  (see also FAQ1.7).

FAQ 1.5

What period of time elapsed between the first alarm and the main explosion?

Answer

            Eye-witness estimates of time varied between 10 to 60 seconds, probably because the witnesses heard or saw different ‘first events’. A reliable estimate of the minimum time was obtained by re-enactments of the time it took witnesses to move from where they heard/saw the ‘first event’ to where they were when hit by the blast of the main explosion.

Of these re-enactments, the most reliable were the seven laboratory witnesses (1-7 on Fig. 4 and 5A) who, on hearing the first noise, stopped, looked, left the laboratory through two swing doors and ran north; the two witnesses on fire duty that day (locations 13 and 14 on Fig. 4) who ran towards Section 25A; and the witness at location 15 who was on a roof when he saw a ‘disc’ flying in the air and had had time to descend and move a few feet from the bottom of a cat ladder before the blast wave.

During the proceedings, estimates of between 25-45 seconds were suggested and the Court (Paragraph 115) gave as their opinion: “The general picture is a clear one and is of a main explosion preceded by about half a minute by some noisy event which was followed immediately by rumbling”.

FAQ 1.6

What was the volume of cyclohexane that escaped?

Answer

The Court’s favoured hypothesis required an escape from the 28" nozzles for 30 seconds (from when eye-witnesses were first alerted until the main explosion) and for 40-60 tonnes of cyclohexane to escape before the ignition. This is reasonably consistent with our calculated ~1.8 tonnes/second combined discharge from the 28" nozzles.

The 8" line hypothesis, by identifying the first noise as the burst from the 8" line elbow, used up 10-15 seconds before the hypothesised explosion under the fin-fan coolers brought down the 20" line. This halves the time for a major escape from the 28" nozzles (reducing this to 20-30 tonnes) but also differs in that there was also an escape from the 8" line (perhaps 2-3 tonnes) and the fin-fan coolers and, in this alternative scenario, a proportion of this cyclohexane was consumed on the outside of the burning cloud throughout this period.

A recent (2003) paper, “A Re-analysis of the Atmospheric and Ionospheric Effects of the Flixborough Explosion” re-examined the ionospheric record and suggests that 14±2 tons of cyclohexane was involved in the actual main explosion, precisely the value determined by the SMRE and DSM in 1974. In either of the two scenarios examined by the Court in 1974, this is less than the quantity estimated to have escaped – and so consistent with either hypothesis.

FAQ 1.7

Were cloud dispersion simulations conducted to check whether the size of the cloud corresponded to the soot fallout from the flash fire? Is the footprint of soot fallout from the flash fire reliable as locating the boundary of the pre-explosion cloud?

Answer

            The simulations that were conducted indicated a cloud dispersion consistent with the soot fallout to the south and west - but that the actual soot boundary extended further north than could be explained solely by emissions from the 28" nozzles. Because soot fallout from a flash fire is very fine and easily identified, it provides – except where subsequent fires consume the soot – a very good indication of the extent of a pre-explosion cloud (q.v. Fig 4).

FAQ 1.8

What is the evidence that the explosion itself was generated within a relatively small volume?

Answer

The forensic investigators from the Safety in Mines Research Establishment (SMRE) deduced the direction of the main blast from the alignment of damaged street lighting posts (and other artefacts not constrained by structures).  Plotted on the site plan, these directions pointed back very precisely to a remarkably confined epicentre between the Office Block (destroyed from above) and the end of Section 25A (the Plant Unit that housed the 8" and 20" lines and fin-fan coolers - q.v. Fig. 4). This is shown by the large red star (below).

 
Court Plate 3 (plus annotations in yellow and red)

The yellow dotted curve illustrates the flight path (presumed) for Fan Rotor 9.

The larger red star is the location of the epicentre of the main explosion

The smaller red star is a suggested secondary epicentre

The SMRE investigators deduced overpressures from observations of crushed skirts and sheared bolts and other easy-to-measure cold-formed damage and plotted these on the site plan. This procedure, incidentally, provided an independent check of their inferred location for the blast epicentre and revealed that the blast wave was of similar strength in all directions – with just a slight hint of greater strength in the direction away from Section 25A.

If the explosion resulted from a delayed ignition when the cloud reached the Hydrogen Plant, the consequence should have been a deflagration that passed through the cloud causing a blast wave substantially more destructive to the south east. Although there was an indication that the blast was slightly stronger in this direction, this is attributable to some blast energy being reflected to the southeast and masked to the northwest by the obstacles of Section 25A.

FAQ 1.9

Was all the damage within Section 25A consistent with these blast directions deduced for the rest of the site and beyond? How do you explain any anomalies?

Answer

Unlike the stand-alone artefacts, the columns in Section 25A were constrained by major structures and could not fall unimpeded. Columns C4 and C5 and the fin-fans assembly were installed at an early stage of construction and appear, from a photograph, to be linked structurally. C4 fell towards Reactor 2 (R2) at about 75° to the left of the main blast direction as, like C5 it would have been impeded by the structure that housed the fin-fan coolers. Most columns deviated by less than 45° - which is consistent with their having started to fall in the primary blast direction but being impeded by the adjacent structures.

The glaring anomaly was the flight path of Fan Rotor 9 (yellow dotted line on Court Plate 3). To explain this anomaly, the Court’s advisors early on did suggest that there could have been a secondary epicentre, shown by the smaller red star on Court Plate 3 above – though not mentioned by the Court in its published report.

The housing of the fin-fan coolers was crushed by Column C5 (double blue lines above and below) and its shattered stainless steel finned tubes were found under (not on) the crushed structure. Because Fan Rotor 9 landed west of Reactor 1 (R1), its flight path (and that of Fan Rotor 8 - Fig. 5A) is only explicable if they occurred before Column C5 fell. As the skirt of C5 was cold sheared from its foundation ring, its fall must have been initiated (if not completed) at the time of the main explosion – after the fan rotors had already taken flight. (Note: Colums C4 and C5, here and in papers by J Venart, correspond to Columns C2525 and C2527 respectively on the constructors' drawings. Papers written in 1974/75, notably by Dr Keith Gugan, use the constructors' nomenclature.)

Court Plate 6

Photograph of pre-construction model of Section 25A plus annotations for

Columns 4 and 5 (fall directions) and the positions of the 8" and 20" lines.

Topic Area 2: The fin-fan coolers

FAQ 2.1

How did the Court explain the flight of Fan Rotors 8 and 9? Did they agree that these took place after the main explosion?

Answer

The Court made no specific claim that the fin-fan coolers shattered after the main explosion and did not take up the suggestion of their own technical advisors that a secondary explosion caused the flight path. They dismissed the entire issue in two enigmatic paragraphs within a detailed multi-page rebuttal of the 8" line hypothesis:

Fire in area of fins and embrittlement of enclosed pipes

164    Since the 8-inch hypothesis assumes that the explosion bringing the 20 inch assembly down and blowing the coolers up occurs within about 10 seconds of the rupture (of the 8 inch line) this section of the (8-inch line) hypothesis requires that within 10 seconds:

(i)      A fuel source (presumably cyclohexane from the 50 inch rupture in the 8 inch pipe) is provided in the area of the fins.

(ii)     The fuel should have been ignited in some way.

(iii)    The fire thereby produced should have heated the fins to 419°C so as to melt the zinc onto the stainless steel tubes.

(iv)    The fire thereby produced should have raised the stainless steel tubes to a temperature of not less than 800°C so as to cause zinc embrittlement.

(v)     Despite this fire an unignited vapour cloud was building up at the same time so as subsequently to ignite and produce an explosion that would exert a downward force on the 20 inch assembly and an upwards force on the fan motors.

No evidence of any kind was given to suggest that this was a reasonable possibility much less a probability. We regard it as most unlikely.

Whereas sub-paragraphs (i) to (iv) accurately represent what advocates of the 8" line hypothesis submitted, no one had suggested an “unignited cloud building up at the same time (as a fire)” – this was pure fantasy (q.v. FAQ4.4). What was suggested was that the shattered finned tubes would release superheated liquid cyclohexane that, as it fell and vaporised, on re-ignition by the flame from the 8" line, would generate the pressure wave that caused the flight of Fan Rotors 8 and 9.

Having dismissed its own fantasised scenario, the Court continued (next paragraph):

Explosion exerting downward force on the 20 inch assembly and at the same time an upward force on the motors sufficient to blow them up and to the west

165  In view of the fact that pressure waves caused by deflagration or explosion produce unpredictable effects we would not regard the difficulties envisaging the required explosion as a serious impediment to the acceptance of the 8-inch hypothesis if the remainder of it were acceptable. We therefore take no further time on it.

These two paragraphs demonstrate that the Court found no flaw in the actual submissions in respect of the flight of the fan rotors and its consequences for the 20" line. For their part, the Court did not even attempt to explain how such a flight path was possible after the main explosion.

FAQ 2.2

Given that it only needs a few seconds at 800ºC for galvanised stainless steel to become embrittled and shatter (if under moderate pressure), was there enough time for the explosion in the fin-fan coolers to have taken place after the failure of the 20" line had begun – yet before the main explosion?

Answer

Although 10 seconds would be sufficient time to cause zinc embrittlement, the challenge is to explain how such a fire could exist within an expanding cloud of cyclohexane as it engulfed all of Section 25A and beyond whilst burning only at its periphery. So the answer to FAQ2.2 is: “No - there was no opportunity for a suitable fire to exist at this location during the first 10 seconds after the 20 inch line had failed (or started to fail)”.

The undisputed facts are that the finned stainless steel tubes of the southernmost fin-fan coolers shattered into 2-3 ft. lengths through zinc embrittlement and the two end rotors landed on waste ground. These two were covered by soot from the flash fire and their bearings’ grease had been partly melted whilst still rotating. The paint on the rotors also showed evidence of a fierce fire (over 800°C) and the soot had fallen on the pre-bubbled paint. The only way the rotors could have got to the waste ground was upwards (through an already shattered bank of tubes). It was accepted by everyone who ventured an opinion in 1974 that these events must have occurred before the main explosion.

 Faced with these facts, the Court needed to argue that these events must have occurred after the (for them first event) failure of the 20" line, yet before the main explosion. If this was their thinking, their enigmatic paragraphs 164 and 165 could be the discarded skeleton of a case the Court had started to formulate to reconcile the 20" hypothesis with the pre-explosion flight of the fan rotors.

One suspects that the Court, already committed to the no-ignition-until-the-cloud-reached-the-Hydrogen-Unit scenario, realised that they could not reconcile this with a fire in the fin-fan coolers as it became envloped by the (un-ignited) cloud* and then unconsciously transposed their scenario for their 20" hypothesis to the scenario they thought was needed for the 8" line hypothesis – thus creating the fantasy of their sub-paragraph 164 (v).

*Note: A fire and explosion in the fin-fan coolers could not occur within an expanding cloud only burning at its periphery. In this scenario, the fire in the fin-fan coolers has to take place before the escape from the two 28"nozzles (FAQ4.4).

FAQ 2.3

So what, according to the 8" line hypothesis, did cause the fire that caused zinc embrittlement in the finned tubes and how did this lead to an explosion that sent the fan rotors flying?

Answer

The Court accepted (Para. 213) that: “It is plain that a relatively small but fierce fire can, if there is a source of zinc nearby cause a sudden catastrophic failure.” Such a fire could have resulted from the 50 inch rupture in the 8 inch pipe. Its emission would have been forward and upwards from this split (as illustrated) and could have engulfed the stainless steel (southernmost) fin-fan coolers (the four that shattered).

Damage to 8" line showing 50" split (splayed out petals indicating a burst under pressure).

The emission would spray in many directions - but with most going upwards and forwards.

One of the two loose bolts is visible in the bottom left corner of this photograph.

The presumed primary orientation of the emission from the 50 inch rupture is illustrated on the photograph. The elbow was located roughly 10 metres below, west and south of the structure housing the fin-fan coolers. When the elbow failed from creep failure, it would release 100-300kgs/sec cyclohexane upwards into the Section 25A structure as a 25-45 metre long jet. As the air intake of 2000cubic feet/sec/fan would have drawn virtually all of the cyclohexane vapour through the fans, this would quickly ignite and burn as a jet fire within the fin-fan coolers. It then would take only a few seconds for zinc embrittlement and the finned tubes would have been shattered as found.

In consequence of the shattering of the finned tubes, superheated liquid cyclohexane would be released from the open tubes and, initially, snuff out the top of the jet flame from 8" elbow. This would re-ignite virtually instantaneously with a pressure wave certainly of sufficient strength to send the fan rotors flying and, simultaneously, knock down the 20" line.

FAQ 2.4

Surely a flame of this size would have been seen by many eyewitnesses?

Answer

            Various statements mentioned flames, before the main explosion, shaped like “a poplar tree”, a “pillar”, “a cone of flames,” “an ice-cream”, etc. On re-enactment it was noteworthy that all witnesses whose viewing angle discriminated between the fin-fan coolers and the 28" nozzles located the flames closer to the former. Any or all of these eye-witnesses may have seen the flame that caused the explosion under the fin-fan coolers.

FAQ 2.5

Is there any witness evidence consistent with an explosion in the fin-fan coolers before the main blast?

Answer

            The witness at location 15 (the Oleum plant, Fig. 4) was on a roof top and could not see Section 25A itself. Not being distracted by what was happening to the structure and columns, he only describes the flames, smoke and debris flying above. He saw something consistent with the flight of Fan Rotor 9 with a cone of flames underneath. As he saw the flying ‘disc’ from the roof top and did not experience the blast until he was on the ground, we may reliably deduce that this event (whatever it was) occurred some 25 seconds before the main blast.

            Witnesses 1 and 3 at location 33 (Flixborough village) both report seeing something consistent with the fin-fan coolers having already shattered when they first looked at the plant. They were over ¼ mile distant from the plant and unlikely to have heard the 8" line burst. As neither moved more than a few feet, we did not re-enact the interval of time between their first event and the main blast.

            Additionally, an amateur film taken from the south within 1-2 minutes of the main explosion shows a vertical flame consistent with that to be expected from the 8" line (and seen in later press photographs) but does not show Columns C4 or C5 or the fin-fan structures still standing.

FAQ 2.6

Could these reported flames have been an upward component of the collision between the escapes from the R6 nozzle?

Answer

            A directed jet from within the expanding cloud would, on emerging, be likely to burn as a jet fire whatever else was happening whilst within the cloud. If, as noted in FAQ 2.3, the escape from the failed 8" elbow had an upwards component of 25-45 metres, this certainly would explain the emergence of a pillar of flame in the locations reported. (Note: a ‘cone-shaped’ flame often describes a jet flame whose upper region has become obscured by smoke and haze).

The orientation of a flame from the collision of the R4/R6 escapes could include an upwards component but would have been more diffuse and, since the R4 nozzle was slightly higher than the R6 nozzle, probably had an initial net southerly direction (before buoyancy took over). It does not seem so likely that such a flame would be likened to a “poplar tree” or a “pillar” though it might just pass for a “cone” and, other than those witnesses viewing in a north-south orientation, all the reports located the flame too far north to have originated from the R4/R6 nozzles.

FAQ 2.7

Isn’t this two-explosion theory a bit over-complicated? Wouldn’t the pressure pulse from the initial ignition of the escape from the 8” elbow be sufficient to blow down the 20” line?

Answer

Possibly “yes” – but, in such a scenario there would be no time to shatter the finned tubes by zinc embrittlement and there would be no explanation for the flight of the fan rotors. It is more credible to argue that the same pressure pulse that caused the flight of the fan rotors (an incontrovertible fact) also brought down the 20" line.

Whatever caused the flight of Fan Rotor 9, there must have been at least 20psi overpressure under the rotor. We do not know the exact distance of this explosion, below and presumably a little to the northeast of the original position of the fan rotor, but it seems entirely credible that its blast wave still exceeded 0.4psi by the time it reached the 20" line (the overpressure all the mechanical engineering experts at that time agreed would have caused the 20" line to jack-knife). 

Topic Area 3: damage to the 8" line

FAQ 3.1      

So what was the damage to the 8” line?

Answer

The most obvious feature was the 50" split with the splayed out petals that showed (without recourse to the detailed metallurgical investigations) that the line had burst under considerable pressure. The previous photograph also shows a non-return valve downstream (left) of the elbow, with an undistorted bolt cage. There is also an indication, easier to appreciate on the next photograph, of a 40° kink that had been entirely accommodated at the failed elbow.

The forensic metallurgical evidence was more than adequately reported and analysed Prof. Ball in his authoritative paper to the IChemE symposium in December 1975 (reprinted in The Chemical Engineer, April 1976, pp. 275-277) and, insofar as detailed metallurgy is not within the expertise of a chemical engineer, is not repeated here. Readers wishing to study metallurgical detail are advised to procure copies of the papers of Prof Ball and the Cottrell and Swann report (The Chemical Engineer, April 1976).

FAQ 3.2      

What impact did the main explosion have on the 8" line? Could this have caused all the damage?

Answer The photograph shows the 8" line viewed from the south in the debris after the main explosion (and fire). The walkway (left) has been pushed ~16" laterally by the collapse of Column C4 (q.v. Court Plate 6 reproduced earlier, causing the line to kink and twist in several places. The most relevant from our standpoint is a 40° twist at the failed elbow – wholly accommodated by the 50" split failure without any distortion to the vulnerable bolt cage of the non-return valve. This is significant. If the elbow was intact at the time of the explosion and the fall of C4, the bolt cage would have been its weakest point and most certainly would have distorted to accommodate the kink. Even without the results of the metallurgical investigations (confirming the absence of torsion at the elbow at the time of failure) it was obvious that the creep failure, whatever its cause, must have preceded the explosion. Furthermore, if this single elbow did fail after the main explosion – uniquely amongst the seven on this line alone and perhaps a few hundred on the plant - it is necessary to postulate the unique circumstances that made it possible for these temperatures to be achieved at this single location.

FAQ 3.3      

So what caused the caused the creep failure in the 8” line?

Answer

The metallurgical findings require there to have been a small flame (no more than a few inches wide) impinging on the intrados of the elbow for at least four minutes (allowing for warming up) with the line still at its normal internal pressure. If the flame had impacted on the outside of the elbow, this would have failed first and if it had been wider than a few inches, the temperature profile around the elbow would have differed from that illustrated on Fig. 6.

A post-explosion general fire cannot explain these metallurgical phenomena, there has to be a small flame directed at the elbow from a short distance. The only plausible source of such a directed flame is the joint associated with two loose bolts (not merely loose – as can happen in fire - but undone by several turns) – a hypothesis formulated in detail by Dr. Gugan. Analysis of pipe pressure stresses at the 8" elbow reveals they were lowest along the line of the intrados (the precise line where creep failure occurred); not a location therefore of likely failure in general fire even if its relative protected geometry is ignored; all other positions on this pipe including other elbows were either more exposed or more vulnerable.

The only way in which failure could have occurred here was by a directed flame supported by a flammable escape from local source. Other than the wholly inadequately bolted flange and the lagging box no other source was found or has been suggested to satisfy this criterion.

FAQ 3.4

Could the creep failure have occurred immediately after the main explosion?

Answer

As well as the so-called “petal” and “3 inch” cracks (Fig. 6) – that must have occurred whilst the line had an internal pressure – there were many obliquely-angled smaller cracks that all investigators agreed were caused by minute droplets of liquid zinc that were thrown on the line from the south-east (i.e. the direction of the main blast) whilst the line was at 800°C (or higher) with a moderate internal pressure.

Given that zinc embrittlement needs a clean oxide-free surface – which is unlikely to persist for even a few seconds into a general fire – it is clear that these droplets of splashed on to the line within seconds of the main explosion and that, at that time, its internal pressure had already been relieved. It is also noteworthy that these obliquely orientated zinc-induced cracks had been propagated from the outside except on the petal itself - where propagation was initiated from the inside (indicating that the petal was already open when the zinc arrived).

FAQ 3.5

The Court quoted a passage from the Cottrell/Swann report in support of their contention that the 8” line failed during the general fire. How do you explain this?

Answer

This passage was a misguided (and, in the event, futile) endeavour by Sir Alan Cottrell to distance the metallurgists from the by-now very public controversy over causation. He did this by rephrasing the metallurgical findings in ambiguous wording that did not distinguish between post- and pre-explosion phenomena and, within this passage, not mentioning those specific findings that unequivocally favoured the 8" line hypothesis.

The Court interpreted his insertion as referring exclusively the post-explosion period and, with spin, quoted the passage to support for their conclusions. The key passage began: “the temperature of the 8 inch pipe rose, during the disaster” (note the neutral wording in respect to post-explosion or pre-explosion phenomena) and continued, rewording the metallurgical findings in a five paragraph neutral context that notably included:    

4      Over a period of many minutes at these high temperatures, the sustained pressure caused the red hot parts of the pipe to swell by creep deformation

5      At the elbow G, where the temperature may have exceeded 900°C, this creep deformation became sufficient to cause failure by w-type cavitation and the 50 inch split then formed.

When questioned on the details, Sir Alan confessed to having no suggestions for how such temperatures could be achieved in a general fire or why they should be confined to the intrados of the single elbow and could not suggest a source for the local flame necessary to produce the temperature profile. The passage quoted by the Court was an ambiguously worded attempt to distance the authors from the causation controversy whilst the report, in respect of its genuinely metallurgical findings, was acclaimed and endorsed by all parties involved in the investigations.

The key issue is to explain why and how, in a general fire whose temperature did not heat any other equipment above 750°C, the 8" line experienced sustained heating to around 950°C at the intrados of this elbow for at least 4 minutes. This requires there to have been a flame, not more than a few inches wide, directed at the intrados for at least 4 minutes and, simultaneously, (to explain the pattern of zinc-induced cracks) within a few seconds of the onset of the general fire.

Neither the Court, nor anyone else, has ever attempted to reconcile these timings or to explain how, in a general fire, this would be possible.

FAQ 3.6

Could the flame from the R6 nozzle have caused the creep failure or, alternatively, from the pool fires that existed within the general fire?

Answer

Since posting “Flixborough Revisited” (January 2005), it has been claimed that “the 8" temperature distributions could be explained by the R6 jet fire discharge” (after the explosion) and the writer, in subsequent e-mail correspondence, appears unshaken in his belief – though, to us, it seems more than incredible.

If the jet fire from the R6 nozzle was the source of this post-explosion flame (initially 28" across but widening as the entrained superheated cyclohexane vaporised), it would have to go right at Reactor 4, then left and then (ignoring buoyancy) down and past the elbow, do an about-turn within the space between the elbow and the non-return valve (!) and, without apparently affecting the non-return valve or the pipe hanger, become a stable flame no more than a few inches wide precisely targeting the intrados for at least 4 minutes in the midst of a general fire (competing for the oxygen necessary to maintain the flame pattern) and with the escape pressure (driving the flame) falling throughout this period.

An alternative suggestion is that the failures could be explained by a pool fire burning from below the elbow after the main explosion. In addition to the objections raised above and in the immediately preceding FAQs, this scenario would have heated the extrados of the elbow more than its burst intrados.

Neither explanation is even remotely credible.

FAQ 3.7

Could a small flame from the lagging box, even if directed at the intrados, have raised the temperature to 950°C?

Answer

It was clearly essential to demonstrate that an external flame could heat the intrados of the elbow to 950°C whilst the superheated cyclohexane was circulating. As the Court was not prepared to check the 8" line hypothesis, the Insurers arranged independent tests at a commercial fire-testing establishment (Yarsley Laboratories). These were reported formally to the Court and the test report introduced in evidence.

At that time it was felt that demonstrating the principle would be sufficient; accordingly a test rig was constructed in which the 8" pipe run was fabricated to one quarter scale in stainless steel pipe. The test fluid was water recirculating by pump through the test region from a supply reservoir to a receiver. The flow rate was also scaled and the water temperature was maintained close to 100°C – thus emulating the prototype. Burners directed flames at the intrados and red heat was achieved there in the flowing system.

Subsequently there was criticism that the test conditions were too mild and therefore produced a misleading result. This necessitated a thorough review of the fundamental thermodynamics and fluid flow criteria. A paper dealing with these and other issues was subsequently published by the IChemE (The Chemical Engineer, May 1976, pp.341-352) in which it was demonstrated that not only were the test conditions not too mild but that they were more onerous by more than a factor of two beyond what the plant would have experienced.

That is to say, not only did the test demonstrate the possibility of the alleged failure mechanism; it confirmed that failure was inevitable given a directed flame at that location.

FAQ 3.8

What was the origin of this small directed flame? Why did it remain stable for at least 4 minutes?

Answer

The presumed source of the flame was a gasket leak at the downstream flange of non-return valve. This would have pressuriaed the annular space between the adjacent pipe and lagging (detatching straight sections of lagging) whilst leaving the strapped lagging box intact. The escape was then redirected by the intact lagging box, as crudely illustrated by Fig. 6B and in several other papers dealing with the controversy and makes the simple assumption that the most intense burning would be around the point of impingement of the jet on the pipe.

During the cross-examination of this expert evidence, the Court poured scorn on the very concept of an off-port flame, asserting that the flame would consume the lagging box well within 4 minutes. They appeared to be genuinely unaware of the sophisticated engineering design that ensures that domestic gas appliances burn from the port or, as visually evident from videos of gas well fires, that off-port flames are quite normal. As these FAQs are written for chemical engineers, no response to this ignorance is needed.

FAQ 3.9

Does any of the eye-witness evidence confirm that the 8" line burst before the explosion in the fin-fan coolers and the jack-knife collapse of the 20" line?

Answer

The laboratory workers were near enough to hear the 8" line elbow burst and, from the end window, could see the scaffolding that supported the 20" line. None saw anything remotely consistent with the first event being the jack-knife collapse of the 20" line or the massive escape of cyclohexane that would have been visible from witness locations 1 and 2 (Fig. 5A).

The one account that may be a reference to the jack-knife collapse describes debris being “spat out” from between Reactors 4 and 6 (Witness 7) – but, he could not have seen Reactors 4 and 6 until he was leaving the laboratory, at least 10 seconds after the events reported by the other laboratory witnesses

Neither the fin-fan coolers nor the 8" line was visible from the laboratory. If, as is probable, the emission initially quenched the flame that caused the burst, the audible sounds and visible events (from the end window) would have been:

ü The sound of the burst, following by the sound of a high-pressure escape.

ü A cloud issuing from beneath the plinth next to Reactors 2 and 3 as superheated cyclohexane vaporised close to the burst and began to condense on contacting air.

ü The sound of the siren as the hot cyclohexane vapour actuated sprinkler sensors and the fire alarm system.

ü The sound of rumbling in vessels as the drop in internal pressure caused boiling.

ü The sound when the jet fire ignited and/or the explosion that sent the fan rotors flying and caused the 20" line to jack-knife.

ü Flames on the plinth from liquid cyclohexane escaping from the 8" line.

ü As the witnesses ran away and looked back, they would have seen flames high up and around the periphery of the expanding cloud, prior to the main explosion.

The laboratory witnesses were heard at the start of the Public Inquiry, before the Court’s 20" line simulation test. Their evidence persuaded Dr. Cox that the jack-knife collapse of the 20" line must have been preceded by at least one other event – though, as he wrote his notes (below), its nature had not yet been determined.

His notes of the laboratory witness evidence are exactly as written by in 1974 - save that location numbers on Figs. 4 and 5A replace witness names and references are to numbers on this website.  Passages in italics are clarifications added in 2005. Seen in retrospect, these original notes are wholly consistent with the 8" line hypothesis (though, at the time, there had been no forensic studies of the finned tubes or the 8" line and the hypothesis had not been formulated).

Although the accounts differ in respect to timings and locations, the witnesses are self-consistent in reporting an initial noise followed by the emission of a low-level cloud and not seeing the 20" line collapse. They diverge as they describe what they saw as they ran from the laboratory and clearly witnessed different events.

Witness 1 was in the middle of the laboratory on the North Side, by the calculator (See Fig.5A) standing, facing east. He heard quite a loud rumble (like thunder from a distance) which quickly (within one second) gave way to or perhaps was drowned by a hissing which, within perhaps another second, grew louder than the rumble and then stayed at the same level of volume and tone. He looked and saw an off-white vapour cloud emerging under pressure from beneath the plinth of reactor R4 extending from ground level (where it may have been more dense) to about five feet in height, giving an apparent "impression" of coming from the south end of this gap (earlier statements referred to a large cloud and being able to see at least two reactors in front of it).  The vapour seemed to go 20 to 30 feet past the edge of the (north to south) road.  Almost immediately after seeing the vapour, heard siren (had not yet moved) but no recollection of hearing it stop.  Someone said it was dangerous and we had to get out.  A matter of a second or so and was running.  No recollection of anything while running.  Did not hear tannoy.  First out.  Person following [M*** who did not survive – JC] ran off at an angle outside lab. Ran to T-junction at NW corner of Plant Section 25B (i.e. about 100 yards from lab exit). Stopped.  Turned almost 180° and saw Hydrogen Plant ignite: that is, saw flames at base of reforming furnace (did not see flames elsewhere).  Noise like a "whoosh" and flames shot down and came out and black smoke went up. Flames around Hydrogen Plant. Turned and ran again. After not more than eight yards (Note: checked on-site) blown down by main explosion (though unaware of an explosion). Total time estimated 20 to 30 seconds. (Re-enacted 25 seconds).

Witness 2(Shift leader). Sitting by (same) calculator, facing south.  Heard rumble for a second or two; did not hear hissing or the siren; does not remember vibration.  Saw six feet high cloud "creeping" westwards at steady pace (from same position as indicated by Witness 1), thought leading edge took 1-2 seconds to pass (Note: vague on time) came out at right angles; glanced a second or two, shouted "Everybody out". Thought rumbling had stopped when saw cloud. Made off (same route as Witness 1); with group when running; then blown down end of Plant Section 25B.  Estimated 20-30 seconds: re-enactment 30 seconds. (Note: this witness lost a close relative in the explosion who had called in to see him only minutes before.  He has made a conscious effort to forget as much as possible. It is of interest that his early statement to Nypro, corroborated by others, shows him to have acted very responsibly.  He gave the call to leave and, at the foot of the stairs, called up to Witness 4 to leave. After the explosion he went back for Witness 5 and made efforts to find M***. Witness 2's statement to Nypro was fuller than that to the Court but confirms what he said there.  He is believed to be a thoroughly reliable witness for the information given.)

Witness 3 In NE Corner of laboratory, sitting facing east. Heard s1ight rumble then high-pitched hissing (between a whistle and a shriek) almost immediately, with no gap between.  The windows started to vibrate.  General commotion in laboratory. Within few seconds got off stool and moved west (into the labora­tory) to look. Could not see Plant Section 25A.   (Note: position of windows on Fig. 5A) but saw 6ft. high vapour cloud still under pressure streaming 30-40 ft. across vacant plot at ground level. Seemed to come from north end of Plant Section 25A.  Watched for few seconds.  Someone said "Get out". Third out after Witness 1 and M*** (who ran diagonally to left on leaving laboratory towards the Reform Plant). Did not hear fire siren, saw no debris, etc.  Turned by end of Plant Section 4 and saw mass of flames (yellow in earlier statement) above Plant Section 25A, nothing at Hydrogen Plant.  Continued running, knocked down alongside Plant Section 25B towards intersecting road at the bottom (i.e. no more than 30 yards further along). Estimated 15-30 seconds overall.  The re-enactment took 14 seconds.  - (Note: this time is considered to be physically impossible; it takes so long merely to run up the road. We think this re-enactment allowed no time for actions in the laboratory).

Witness 4 Upstairs in laboratory, facing south straight out of window. "Uncertain whether there was a muffled thump at the start but it was the rumbling which made me look up", after about 2 seconds. (It sounded like a previous experience he had of a rupture). Saw a relatively small escape from plinth level (semi-translucent) centred around gap between R3 and R4.  It fell to point about 2/3 across road and in 2-3 seconds crossed about 2/3 of the rough ground, travelling about 14 MPH, fanning out at a slower speed at edges and expanding into a cloud. The point of escape seemed to be above the plinth and it seemed to be transparent at first and then became cloudy like steam. Did not see any debris.  Turned, picked up safety helmet, moved pretty fast but did not run.  (In Nypro statement mentions being called by Witness 2 and that he heard the siren. Neither point emerged from his examination in Court).  Outside, "the sound had softened somewhat.  The pitch was lower. There probably was a hissing but I do not recollect that.  I still took it as a reverbatory sound". Ran to junction of Plant Sections 4 and 25B. Thought explosion came from Hydrogen Plant.  Timed at 45 seconds; he thought this accurate to within 5 seconds.

Witness 5 Bench south side, downstairs, looking south. Saw vapours escaping. Remembers hearing "the rumble that was previously mentioned (i.e. in earlier evidence). However I cannot recall the time when I heard it - whether I heard it while I was in the laboratory or after I had left the laboratory". Saw white cloud and a jet which he believed was the point of escape.  Cloud came from between R4 and R6, about half way up reactors. Jet was fairly small, apparently creating the cloud which moved westwards, at right angles to plant.  (No mention of this cloud dropping or rising) but thought that he couldn´t see under it.  Watched for 5-10 seconds. Heard someone say that we should get out. Heard siren when leaving laboratory, and then tannoy.  Was one of the last out.  At first no panic, then, about halfway to Plant Section 25B, somebody said "It’s going to go up", then "ran like hell".  Fell at T-junction, Witness 2 came back to help.  Timed as 57 seconds (from first sight to explosion).  Thought it was probably longer in reality.

Witness 6 Working between benches middle west of laboratory heard muffled (mediocre) explosion and almost immediately windows rattled (the explosion was more like a 2 second rumble) accompanied by a hissing like a background noise.  Glanced at top of reactor for not more than a second; saw colourless fumes coming from mid­way to top of reactors somewhere between R3-R6. (Note: that his angle of vision would enable fumes in front of R4 to be seen as a shimmering against the background of R6 - as he demonstrated in Court).  He did not look at ground level. Shimmering looked like heat around tops of reactors.  Did not see any of things mentioned in statements of earlier witnesses. Said "let's get out of here".  Amongst the last to leave the laboratory: heard siren as leaving (probably while in corridor). Saw Witness 4 leave laboratory while outside (Note therefore, must have looked back but not questioned on this). Thought smelt cyclohexane when by Plant Section 4. "I think I started running then, because I thought if this catches fire the place will go up".  Then glanced to left, saw a ball of fire already ignited (seemed to be between the Hydrogen Plant and laboratory).  Could feel the heat (at this point at junction of Plant Sections 4 and 25B, about 40 yards from laboratory).  Blown over before reaching T-junction at end of Plant Section 25B.  Estimated 30-45 seconds; not timed; thought most of the time was in the laboratory (building) going through the swing doors.

Witness 7 Standing far west of the laboratory by fume cupboard facing east wall. Heard muffled bang. Had "half second" view through end window of a slowly rising vapour escape (like white smoke). The escape came out like a gas jet; a wide jet, not forceful.  Once it made contact with the air it went upwards. It was 4ft long, 2ft. diameter and after going up for almost 10 ft. it dissipated. Debris was being "spat­out" from between R4 and R6 higher than about half way up the reactor. Did not look long enough to see debris land. Ducked as he swung to right and north, this action being a continuation of his action when the bang first came. (Note All this was seen in the first "half-second" view and the above description of this movement appears incompatible with a view to the south. [Note added as clarification in 2005:  from the end window he could see the roadway opposite R4/R6 but not the reactors: from the adjacent window he could see reactors R1/R3. The only time he could have seen R4/R6 was at least 10 seconds later as he left the laboratory and passed Locations 1 and 2]. This view would have been possible only after he was leaving when he reached the north of the laboratory. This may be checked by laying a ruler across Fig. 5A).  After the first bang, windows seemed to shake; building vibrated, then died down, then all of a sudden increased.   Most of colleagues gone (i.e. when he looked up and around after ducking down) saw Witness 5 looking pretty shocked.  Saw someone running across open space while someone else still in laboratory with him. Began to leave laboratory.  When opposite calculator saw (for half a second) yellow/orange/red flames (no smoke visible) on reactor plinths (R4 to R6) (apparently "chasing" him along). Flames as high as reactors.  Impression of something ignited on plinths.  (This was a long time after someone had shouted "Let’s get out of here").   Vibration like approaching underground train, like flame fed by air; ran faster.  Heard siren on leaving but while still in laboratory, getting louder as he went.  Just outside building heard tannoy message "Fire on 25, Fire on 25".  Did not see Witness 4 leave building.  Saw group a long way off and then a group of which he was the last.  Did not run fast.  Explosion occurred about level with C402 (20-25 yards from laboratory door). In a police statement he suggested that the blast came from the Hydrogen Plant).  Estimated time 40 seconds.

Topic Area 4: How the 20" line failed

FAQ 4.1

So why did the Court prefer the 20" line hypothesis?

Answer

            This was publicised as the probable cause before the Court convened and, in advance of the hearings, the Court commissioned a test rig to simulate the anticipated jack-knife failure of the 20" line. This test took place a few days into the hearings and did not reproduce a jack-knife failure. Until then, the Court had not thought about an alternative to their preferred hypothesis and, even then, did not commission forensic studies of the 8"line.

            The alternative causation theory would have obliged the Court to highlight the vulnerability of the Nypro plant to two loose bolts and other shortcomings (excessive inventory, zinc embrittlement, jet fires, etc.) and, faced with this stark choice, their obligation to allay public concerns took precedence over scientific objectivity.

For further details go to "Flixborough Revisited"

FAQ 4.2

What is your opinion about the “water theory” advocated by Ralph King?

Answer

Ralph formulated this theory to account for the failure of the Court’s simulation test on the 20" line. It addresses the specific problem posed by the simulation test – that a rise in its internal pressure provided insufficient energy to cause a jack-knife failure. If there had been no creep failure in the 8" line and no pre-explosion in the fin-fan coolers, this would have been a front-runner amongst the many hypotheses suggested to explain a spontaneous collapse of the 20" line.

The underlying science of Ralph’s hypothesis is wholly proven and has application to other scenarios (for example, the “Caustic Layering” phenomenon outlined elsewhere on this website). The shortcomings of his hypothesis are that it relates solely to the trigger for the 20" line failure and takes no account whatsoever of any other phenomenon or evidence.

FAQ 4.3

What is your opinion about the theory formulated by Professor Jim Venart that explains the failure of the 20” line by a complex multi-step mechanism over a period of more than 20 seconds? (see PSEP, 2004,82(B2): 105-107)

Answer

Whereas Ralph King suggested an event that could have triggered the 20" line failure, (accepting all else in the Court’s version of events), Jim Venart analysed the mode of failure and tried to accommodate it with some eye-witness accounts. Like Ralph King however, he did not try to explain the metallurgical evidencerelating to the 8" line or the flight of the fan rotors (being unaware, when he formulated his theory based on the Court’s version of events, of any of this additional evidence).

Jim Venart accepts that the B4 bellows failed before the B6 bellows (as did the Court) but (unlike the Court) postulates a delay of up to 25 seconds before the B6 nozzle failed. He presumes that an unignited vapour cloud could be created from a single nozzle escape whilst fires were burning on the plinths – a fundamental flaw undoubtedly recognised by the Court when they considered the possibility of a multi-stage mode of failure (q.v. FAQ2.1). On this count alone (there are others), his theory is not credible.

The merits of his scenario include that he does not claim, as did the Court, that the system pressure had been allowed to rise above normal and that, accepting that “a two-stage event was certainly experienced by most witnesses”, he tries to reconcile (some) eye-witness accounts before the main explosion. Nevertheless, his scenario is hard to reconcile with what was actually seen and heard (FAQ3.9).

In his scenario, the laboratory witnesses would have been alerted by the B4 bellows rupture and should have seen the scaffolding demolished as an already jack-knifed 20" pipe swung violently about a still-attached B6 bellows. In reality, as the PSEP paper notes: “five were specific in the description of a slowly developing vapour cloud moving to the west over the time period of their observation, ~8-12s … about 1.5-2m high.” This clearly describes a relatively small initial escape from under the plinths - with nothing whatsoever being observed at the scaffolding and 20" line.

During this same period, witnesses from a distance might have seen a huge jet flame from the B4 nozzle (or at least a major escape) probably with a southerly component: instead, several reporting seeing a vertical flame above the location of the 8" line before the main explosion (FAQ2.4).

In short, Jim Venart’s theory:

a.       Ignores what happenned to the 8" line and fin-fan coolers (FAQs 3.6 & 3.2)

b.      Assumes an unrealistic long time between the bellows’ failures (FAQ4.4)

c.       Relies on an unproven fatigue-induced initiating event (FAQ4.5)

The theory also does not mention the fire that caused zinc embrittlement in the fin-fan coolers prior to the main explosion (FAQ2.2) and or how it was possible for a 2" flame to impinge on the intrados of the 8" line elbow for at least 4 minutes after the 20" line had failed whilst the line remained at its normal working pressure (FAQ3.6).

FAQ 4.4

What is your opinion about the timing of the successive steps of 20” line failure as advanced by Professor Jim Venart?

Answer

His timings are scientifically arbitrary, having been chosen to correspond to his personal interpretation of selected eye-witness evidence. So, whereas his sequence is an integral feature of the mechanical engineering analysis, his timings may be altered without invalidating his hypothesised mode of failure.

The fundamental flaw in his timing is that it requires all the cyclohexane for the main explosion to come from the R4 nozzle. But, without a simultaneous escape from the R6 nozzle, the R4 nozzle escape would burn as a jet fire and (even if it didn’t and ignition was delayed for 30+ seconds as the Court hypothesised) insufficient fuel would escape in his time-frame to explain the force of the main explosion.

To be consistent with the evidence and the science of flames and explosions, both nozzles must be open to create an expanding cloud of unburnt fuel (FAQ1.2), for 15+ seconds before the main explosion. In this modified scenario, successive steps in his hypothesised mode of failure may be telescoped to 1-4 seconds and incorporated within the 8" line hypothesis scenario - retaining his mechanical engineering analysis and wholly conforming with the eye-witness evidence, the science of explosions and  flames and with the forensic evidence relating to the 8" line and fin-fan coolers..

FAQ 4.5

What is your opinion about the initiating event that led to the failure of the 20” line? Does the credibility of this event depend on the presumed mode of failure?

Answer

The essential facts of the failure are that it was initiated by a small tear in the B4 bellows, that the B6 bellows failed after elongation and that the line ended up in the jack-knifed orientation between the reactors. This initial tear was examined by more than one investigator for signs of fatigue but none were identified.

In the Court’s scenario, the initiating cause was internal overpressure. When their simulation test showed otherwise, they argued that the conditions were not quite right whilst conceding that a small external overpressure would cause a jack-knife.

The 8" line hypothesis, as formulated in 1974, accepted that an explosion in the fin-fan coolers would initiate the collapse and/or that internal boiling in the system, resulting from the burst 8" line and the shattered fin-fan coolers, may have been the trigger (as argued by Ralph King in his “water theory”).

In contrast, Jim Venart speculates that, despite the observed lack of forensic evidence for fatigue, that this caused the initial tear in the B4 bellows. Given the evidence for a pressure wave (from the explosion in the fin-fan coolers) and internal instability (from the burst 8" line and the shattered fin-fan coolers), there seems no reason to prefer his speculation over the investigators hard forensic evidence.

FAQ 4.6

What, in your opinion, was the trigger for and mode of failure of the 20” line?

Answer

Whereas modes of failure are for mechanical engineers to determine, it seems intuitively improbable that the R6 bellows could stretch with the R4 bellows in situ - as suggested by the Court (Fig. 3). In this respect, the mode of failure suggested by Professor Venart is obviously the more credible and, additionally, has the merit of explaining the substantial dents in the Reactor R6 shell and plinth. Telescoped into a believable 1-4 seconds, his mode of failure would appear to be the more credible. 

All parties and independent investigators agreed that the dog-legged 20" pipe would fail if subjected to a moderate external overpressure or from internal instability and, as both would have occurred when the fin-fan coolers shattered and an explosion sent the fan rotors flying, this seems the most likely trigger for the 20" pipe failure.

Last updated: April 2006