Page 33 - Failure Analysis Case Studies II

20
This paper is concerned with the first and second questions, and describes the hand calculations. A
second paper will describe the finite-element calculations and compare the results of the two analyses.

2. STRUCTURE OF THE B/C FIREWALL

The platform topsides between the 87 ft and 107 ft levels had four modules, identified by letters
A to D from south to north. The firewalls between the modules were intended as fire barriers, and
were not designed to resist blast. The B/C and A/B firewalls were similar in construction, but the
C/D firewall was thicker and more massive: the significance of this point is discussed below.
The B/C firewall was 6.35 m high, and extended across the whole 46.63 m breadth of the platform.
Its construction is illustrated in Fig. 1, which shows part of the wall, and details of the cross-section
and clamping.
The wall was composed of rectangular and square panels within rectangular frames, some 2438
by 1524 mm (8 by 5 ft “large frames”), others 1524 by 1457 mm (“small frames”), and a few smaller.
Each panel was bounded by a rectangular frame made of 50.8 x 50.8 x 6.35 mm (2 x 2 x 1/4 in.) steel
angle, welded at mitre joints at the corners. The frame held a sheet of fire resistant composite, an
8.5 mm asbestos-cement core faced on both sides by 0.5 mm galvanised steel sheets, pierced by
6.35 mm tined holes on a 17.5 mm square grid. The composite sheet was bolted into the frames by
9.53 mm (3/8 in.) Whitworth bolts (“panel bolts”).
The frames were bolted together by 9.53 mm bolts (“frame bolts”). There is conflicting evidence
about the spacing of the frame bolts, which gave values between I5 and 24 in., and the subsequent
analysis is based on an intermediate spacing of 18in. (457.2mm). Generally, but not always, the
lowest panel was large, the next panel small, and the top panel large.
At the foot of the wall the frames were welded to the deck plate. At the top the frames were
bolted to 76 x 51 mm angle cleats, which were bolted to 102 x 63 mm cleats, which were welded to
the ceiling plate.
The firewall lay on the south side of the heavy trusswork which formed part of the module
structural framing. The truss was a sequence of N-shaped panels. The wall was clamped to the truss
by two lines of clamps, one just over 2438mm from the floor Gust above the lower bolted join
between frames), and the other just over 3962mm from the floor. Each clamp held the wall against
the truss by two lcngths of 9.53 mm studding. An explosion in module C would push the wall to the
south, away from the trusswork, and would put the studding in tension.
There were minor variations in construction along the length of the wall, which included a door,
pipe penetrations of a few composite panels, and some smaller frames.

3. OVERPRESSURE LOADING OF B/C FIREWALL BY AN EXPLOSION IN
MODULE C

The inquiry report [I] examined the evidence for the size of the flammable cloud before the first
explosion. It concluded that the fuel involved was condensate, and that the mass of fuel within the
flammable part of the cloud was probably in the range 30-80 kg [l], Section 5.
Computational fluid dynamics (CFD) calculations were carried out by CMI [2]. The calculations
are based on a gas cloud containing 45 kg of condensate, homogeneous and at stoichiometric
concentration, and ignited close to the south wall. Figure 2 gives the calculated pressure history at
point PI, which is in module C, on the B/C wall near the west end. The maximum pressure reached
at P1 was 19.5kN/m2 (0.195 bar). The maximum pressures at other points within module C were
higher. The CFD calculation assumed that the wall broke up when the instantaneous pressure
reached 10 kN/m2.
In the subsequent dynamic analysis the pressure history is idealised as an asymmetric triangular
pulse with peak 19.5 kN/m2, rise time 81 ms and fall time 46ms. Figure 2 compares this idealisation
with the calculated history.