FLIXBOROUGH 25 years on
15 Jan 2000
The cyclohexane oxidation plant at Flixborough, near Scunthorpe in what was then Lincolnshire, was one of at least seven such plants built to DSM (Dutch State Mines) designs around the world to serve the nylon industry. Compressed air was bubbled through hot liquid cyclohexane under pressure at 155 degrees C to convert it to a mixture of cyclohexanol and cyclohexanone (for subsequent conversion to the nylon-precursor caprolactam). Conversion and reaction rates of the oxidation process were low and the inventory of hot pressurised cyclohexane was high (about 145tonnes).
At Flixborough six reactors were operated in cascade, with liquid passing from one reactor to the next at a slightly lower level through 28in flanged stub-pipes joined by 28in flexible stainless steel bellows. Compressed air passed in crossflow through the reactors via sparge pipes. All six reactors had powerful stirrers to improve the dispersion of air in the liquid. The bellows allowed for any thermal expansion movements.
The plant was run by Nypro (UK), which was jointly owned by DSM (55%) and the UK's National Coal Board (45%). It had started operations in early 1973 and ran without serious incident until November of that year when, because of the electricity shortages prevalent at the time, it was operated temporarily with the stirrers stopped to save power. On restarting the stirrers, that in No 4 reactor was found to be bent and it was removed.
The plant then continued to run with five stirred and one unstirred reactor until the end of March 1974 when a split occurred in one side of No 5 reactor next to one of the 28in pipe stubs. The plant was hastily depressurised and shut down while the crack spread further. Water was sprayed over the leak as the cladding was removed to reveal a 2m long crack.
The split reactor was removed and, before the cause of the split was established, it was replaced by a `dog-leg' 20in diameter pipe with 28in flanges at each end to connect it to the bellows on reactors No 4 and No 6.
DSM subsequently reported (after the disaster) that the split was caused solely by nitrate stress corrosion resulting from contaminated cooling water that had been sprayed over the reactor earlier. Meanwhile the plant was restarted with five reactors, one of which (No 4) had no stirrer. It ran for two months before being shut down again to repair a leak.
On restarting the plant hot cyclohexane was first circulated through the reactors until the whole system was full. Only then was compressed air to be introduced. The hot circulation started at 9.00am on 1 June 1974. By noon the plant had reached operating pressure and temperature but the introduction of air could not be started at that time and so hot circulation was continued.
At 4.52pm the 20in dog-leg pipe and the bellows attached to it ruptured, causing the entire contents of all five reactors to escape at sonic velocity to form a vast flammable vapour cloud the size of a football pitch. The vapour's ignition less than a minute later resulted in an `open flammable cloud explosion'. The explosion was the largest ever in peacetime Britain. It killed 28 workers (most of whom were in the plant control room), destroyed the entire works and office block, and caused widespread injury and property damage within a five mile radius.
The Government set up a public inquiry to investigate the disaster. Consulting engineers Cremer and Warner advised the court of inquiry, while Manderstam and Partners were engaged by Nypro. Although assigned to Manderstam to act on their behalf, I wanted, perhaps naovely, to investigate independently without direction from Nypro or its owners. Unfortunately, the NCB director concerned had the idea that a hypothetical earlier explosion outside the oxidation plant, caused by a supposed leak from an 8in pipe, had preceded the bellows and pipe failure.
The main events in the chain of disaster were the rupture of No 5 reactor in March 1974 and the rupture of the 20in pipe and bellows on 1 June. The second could not have happened without the first, and both needed proper investigation. However, so much of the inquiry's time had been spent on the 8in pipe theory that more likely causes received scant attention.
It was not until the inquiry's report was published that the officially accepted cause for the failure of No 5 reactor (nitrate stress corrosion) was reconsidered by the HSE. In 1976 it issued Technical Data Note 53/2 `Nitrate stress corrosion of mild steel'. This states that nitrate stress corrosion cracking only occurs in mild steel in areas subject to abnormal stress. A crack which starts, for whatever reason, in a pressurised mild steel vessel will continue to spread as the stress increases. If it spreads through an area in contact with water containing nitrates, the metal here may show symptoms of nitrate stress corrosion. So this does not show that nitrate stress corrosion was the cause of the crack in No 5 reactor. Rather, it suggests high local stress as the cause.
Because of the safety implications of this for plants of similar design in other countries, the current secretary of state for the environment Michael Meacher has recently instructed the HSE to look further into the reasons why No 5 reactor failed. This it is now doing.
HSE is also carrying out experiments (due for completion this August) at its Buxton site to demonstrate my `water theory', which was first published by PE in September 1975. The combination of the 20in pipe and bellows at either end had two weaknesses: each of the bellows was fixed only at one end; and because the bellows were off-set, the thrust forces on each end of the 20in pipe produced a considerable bending moment allowing the bellows to squirm.
The explanation for the bellows' failure accepted by the court of inquiry was `dynamic squirm' coupled with progressive deterioration of the highly and repeatedly stressed metal. The court admitted that it was not an entirely convincing explanation and there might be other unknown causes.
The essence of the `water theory' was that there had been a violent internal eruption starting in the unstirred No 4 reactor when the liquid interface between water which had settled in the reactor and hot cyclohexane above it reached a critical temperature of about 145 degrees C. This theory is in line with all relevant facts about the plant history and may partly explain the earlier failures of the stirrer in No 4 reactor and of No 5 itself. It may also explain the `rumblings' from the reactors heard by eye-witnesses just before the rupture and the unusual damage found inside No 6 reactor after the disaster. PE
Ralph King is a consultant chemical engineer and author of King's Safety in the Process Industries, the 2nd edition of which (co-authored with Ron Hirst) was recently published by Arnold.