PROCESS SAFETY · BYPASS

A temporary pipe
killed 28 people.

Six reactors in a line oxidised cyclohexane on the way to nylon, each a step lower than the last so the liquid overflowed into the next. In March 1974 reactor 5 cracked and was lifted out. A temporary 20-inch pipe bridged the 28-inch nozzles of reactors 4 and 6; the height difference forced a dog-leg, with a bellows at each end. Nobody calculated it. The only drawing was chalk on the workshop floor. Two months later, on 1 June 1974, it failed: the cyclohexane flashed into a cloud, the cloud found a flame, and the explosion wrecked the site.

The reactor line

R1R2R3R4R5R628 in+ bellowscracked, March 1974Six reactors, each 14 in below the one before. Liquid overflows from one to the next through 28 in pipes.
REACTORS IN SERVICE6 of 6
THE LINK BETWEEN 4 AND 628 in pipeswith expansion bellows (para 41)
THE DRAWINGEngineereda well designed and constructed plant (para 209)
OPERATING CONDITIONS155 °C8.8 kg/cm² gauge, about 863 kPa (para 28)

Section view, not to scale: the 14-inch steps and the pipe diameters are drawn to one scale, the vessels to another. The real reactors were stirred and fed with air and catalyst; those lines, the off-gas line and the after-reactor are left out. The flow is slowed for readability.

MARCH 1974

Reactor 5 has cracked and is leaking. Production is waiting. Lift it out and bridge the gap.

Six reactors oxidise cyclohexane at 155 °C and 8.8 kg/cm² (para 28). Each sits 14 inches below the one before, so liquid overflows from one to the next by gravity through 28-inch pipes with expansion bellows (para 41). In March 1974 reactor 5 was found cracked and leaking. It was taken out of the line.

Assumptions and limits

The drawing follows the report's description (paras 28, 41, 61, 62 and 68): six reactors each 14 inches below its predecessor, 28-inch connecting pipes with inset bellows, and a by-pass of three lengths of 20-inch pipe welded into a dog-leg, bolted to the flanges of the two bellows that had connected reactors 4 to 5 and 5 to 6. The scaffold poles that held it up are not drawn. Lengths are schematic.

THE CALCULATION NOBODY DID

Put it under pressure

P × A = 17.8 tonnes-forceP × A, pushing backturning moment P × A × 2eshear F on the bellowsshear FR4R628 in bellows28 in bellowsmitre bends: the bending moment peaks here8.8 kg/cm² (863 kPa): the operating pressure (para 28)20 in bore, bellows 28 in. Rotation and squirm are exaggerated for legibility.
PRESSURE8.8 kg/cm²863 kPa gauge · 1 kg/cm² = 98.0665 kPa
THRUST F = P × A174.9 kN17.8 tonnes-force on the 0.2027 m² bore
THE COURT'S FIGURE38 tonnesthe Court's figure for the thrust on the bellows at working pressure (para 62)
WHERE YOU AREOperating point

A = π × 0.254² = 0.2027 m² for the 20-inch bore (0.508 m). The Court's 38 tonnes is the thrust on the bellows, whose effective cross-section is larger than the pipe bore. The Court recorded that no calculations were done to ascertain whether the bellows or pipe would withstand these strains (para 62).

PRESSURE IN THE BY-PASS

Take it from empty to the operating point, then on to the relief setting. Where does the only test it ever had sit on this scale?

F = P × A with P in kPa and A = 0.2027 m². Tonnes-force = kN / 9.80665.

At 8.8 kg/cm² (863 kPa) the pressure pushes on the 0.2027 m² bore of the 20-inch pipe with 174.9 kN, about 17.8 tonnes-force, at each end. The two pushes are 14 inches apart in height, so they twist the dog-leg and load the bellows sideways. The Court of Inquiry put the thrust on the bellows themselves at some 38 tonnes at this pressure (para 62; Appendix I para 3 gives the shear on each bellows as about 2.8 tonnes). No calculations were done to ascertain whether the bellows or pipe would withstand these strains (para 62).

Assumptions and limits

The thrust shown is P × A on the 20-inch bore, a simplified pressure-thrust and bending picture. The Court's Appendix I treats the two bellows as very flexible, so the only horizontal force across each is the thrust P × A with A the bellows' effective cross-section; the two thrusts are offset by 2e, the 14-inch drop, and make a turning moment P × A × 2e, balanced by shear forces F = PAe/L on the bellows, with the bending moment largest at the mitre joints. Under operating conditions the Court put P × A at about 38 tonnes and F at about 2.8 tonnes (Appendix I paras 1 to 3). Rotation and squirm on screen are exaggerated and qualitative. The exact sequence of the failure is still argued: the Court found the bellows squirmed and the pipe jack-knifed under pressure near 155 °C; later writers, Cox (2007) among them, have put a fire on a nearby 8-inch line first, and Venart has defended the 20-inch account. The bellows were original equipment; the 20-inch pipe was the only new part. The quantity released and its TNT equivalent are estimates and are left out of this page.

MANAGEMENT OF CHANGE

What a modification needs

  • A dimensioned drawing of the modification, checked before anything is cut.

    What the by-pass had (para 62)Chalk on the workshop floor. No other drawing of the pipe was made.
  • The forces the change creates under pressure: the thrust, the turning moment, the bending at the bends, the load on the bellows.

    What the by-pass had (paras 61 and 62)Only that a straight 20-inch pipe would pass the flow and hold the pressure. Nothing for the dog-leg, the thrust or the bellows.
  • A check against the piping standard and the bellows maker's designer's guide.

    What the by-pass had (paras 62 and 63)Neither BS 3351 nor the designer's guide was consulted. The guide says two bellows must not be used out of line without supporting the pipe.
  • A hydraulic strength test to the standard, above the relief valve setting, before the system goes back into service.

    What the by-pass had (paras 70, 72 and 73)A nitrogen leak test to about 9 kg/cm². Never a strength test, never above the 11 kg/cm² relief setting.
  • Someone with the authority and the engineering to check the design, the construction, the test and the supports before start-up.

    What the by-pass had (paras 71 and 210)No overall control or planning of the design, construction, testing or fitting, and no check that the work had been properly done. The works engineer post was vacant.

The five items are a teaching checklist distilled from paras 61 to 73 and 209 of the report, not any company's procedure. Real management-of-change procedures also cover supports and anchors, materials, hazard review, who signs, and training. The Court's second recommendation was inspection and test by a competent person after any significant modification and before the system goes back into use (para 209).

YOUR CHANGE

Tick what your modification has. Then see what the real one had.

0 of 5items a modification should meet before start-up

0 of 5. An assembly of completely unknown strength (para 63). This is where the real by-pass stood when it went back into service on 1 April 1974.

Assumptions and limits

HSE's case study summarises the failing in one line: a plant modification occurred without a full assessment of the potential consequences. The Court's first recommendation is that any modification should be designed, constructed, tested and maintained to the same standards as the original plant (para 209). Neither this page nor the report says the engineers were careless; the Court found that the desire to resume production led everyone to overlook that connecting two reactors was an engineering problem at all (paras 56 and 57), while the key post of works engineer was vacant (para 210).

WHAT THE CHALK MEANT

The pipe was built well.
The change was never engineered.

The welding was sound and the dimensions were accurate (para 64). What was missing was everything around the pipe. Calculations were made for the flow and for a straight pipe under pressure; none for a dog-leg, because nobody yet knew the pipe would have one (para 61). Nobody saw that two offset thrusts make a turning moment, or that the thrust on the bellows was some 38 tonnes at working pressure. No standard was consulted, no designer's guide, no drawing but chalk, no pressure test before fitting (para 62). The only test afterwards was nitrogen to 9 kg/cm², a leak test, when the standard asked for a hydraulic test above the 11 kg/cm² relief setting (paras 70 and 73).

Under pressure the picture is simple enough to draw. Each bellows feels P × A along the pipe. The two forces are 14 inches apart in height, so they twist the dog-leg; the bellows take a shear they were never designed for and can squirm, and the bending moment peaks at the two mitres (Appendix I). A hydraulic test to the standard would almost certainly have failed the assembly on the test stand, and the disaster would have been averted (para 73).

The Court's first recommendation is the whole lesson: a modification must be designed, constructed, tested and maintained to the same standards as the original plant (para 209). HSE puts it in one line: a plant modification occurred without a full assessment of the potential consequences. A temporary pipe is still a pressure system. Chalk is not a drawing.

FROM THE FLASK TO THE PILOT

Taking a process
beyond the lab?

Jon and I are building RARE Labs toward an autonomous pilot plant builder with AI and robotics. Join the waitlist and tell us what you are trying to scale.

Join the RARE Labs waitlist ↗

FOLLOW THE EVIDENCE

Sources and model limits