Professional Documents
Culture Documents
Despite ASME PCC-l offering a warning and guidance on the disassembly of thick flanges and gen-
era/ good practices in various training documents avoidable incidents still occur.
This paper presents jormal studies and case histories that can be used to formalize and strengthen
procedures and technician training. The paper will include scenarios o/flange disassembly both with
and without intemal pressure and wil/highlight Ihe potential for incidents that could arise from
them. Finite element analysis and case histories will be used where applicable to support or illustrate
the results.
Samuel Tait
Methanex NZ Ltd.
. -42in Pipe·Pipe,
1.0
lOOMP,sasket
12
•
.5 -... 42in Blind f1anse,
-2 2.0 no I,$let
•• 1.5
~ -18in el'lI 300
blind liallle, 100
S
1.0 MP,psket
.....-
- 18 1n elus 300
0.5 + 1 - - blind flall8e, no
s,$ltt
0.0 .... 18in elm 300,
Figure 2. Screenshot from thin flange unbolting o 2 4 6 8 10 12 14 16 18 20 22 24 100 MP, psk!l,
animation Remov,1 Sequtnct low eloMl
Figure 3. Bolt load ratio during disassembly of
Unexpectedly, the bolt loads in the thin flan ge 42 in. and 18 in. flanges
were signifi cantly higber than the bolt loads
seen in the thick flange . As can be seen on the
stud highlighted with a red circle in Figure I, All of the scenari os saw th e bolt load ratios in-
the thick flange, the stresses are mostly in the crease in a simila r pattern to that published in
yellow and orange areas of the scale (400MPa- the 2002 paper with the loads increasing in ste ps
500MPa). While in Figure 2, the thin flan ge, the to a maximum followed by a rapid drop in load
stresses are mostly in the red region of the scale for the last few bo lts. There were too many vari-
(600MPa) and the max imum stress is in the grey ables to draw any valid compari sons between
area (739MPa+), this is above the yie ld strength diameters and thicknesses.
of a standard B7 or B 16 stud. This result was
counter-intuiti ve and is anomalous to the current The mode ls were adjusted so that one variable
infonnation from ASME PCC- I 20 13 that bolt could be examined at a time. A 24in. flange
loads during disassembly increase with increas- which is the largest standard ASME B16.5 3
ing flange thickness. flange was modelled in stead of the 42in. flange
which was not a standard size. To compare di-
The investigation was expanded to try and un- ameters a standard ASME 81 6.5 12in. flange
derstand th e anoma ly from the original results was mode lled instead of the 18in. flange. The
where the bolt loads in the thin fl ange were model includes a gasket of standard ASME
higher than the bolt loads in the thick flan ge. 816.204 dimensions, the properties of the gasket
The original 42in. pipe to blind fl ange model are assumed to be the same as carbon steel
was expanded to include a pipe to p ipe j oi nt, based on the assumptions that the flange faces
and a gasket. An 18in. class 300 flange was also will pull up against the outer a nd inner ring of
modelled for comparison. As per the 2002 finite the gasket and that there is very little relaxation
element simul ation, the resul ts were plotted as a of the spira l wound part of the gasket during
bolt load ratio against the bolt removal se- fl ange di sassembly. Class 300, 600, 900 and
quence, the removal sequence modelled was the 1500 fl anges were modelled.
cross pattern from ASME PCC-I 20 13. The bolt
load ratio is the force on the bolt during disas-
i..... 1.S
o
~ 12in
- 1 1 i n Class 1 500
C lass 900
.~
~ '-+- 2 4in C lass 300
~ 1.0 ___ 2 4in Class 600
~
0.0
o 2 4 6 8 10 1 2 14 1 6 1 8 20 22 2 4
Removal Sequence
Figure 4. Bolt load ratio during disassembly of various classes of J2in. and 24in.jlanges
B
With the number of variables reduced the results
became more consistent. Whi le there was
remarkably little difference between the 12 inch
, .r
and 24 inch flanges, the result that stands out is
the 24 inch class 300 flange. This has
1 J .S f
~
• ,"
significantly higher bolt loads than all of the
other flanges as is particularly noticeable when -+-O.SI. cross. blind fl~nle
S
..... I.Ot. cross. blind Ilan,e
compared with the 24 inch class 600 flange that O.S
_ 1.St. cross. blind flanae
also has 24 bolts. This backs up the findings
~.
- 2.Ot. cross, blind Ilanae
from the original model of the 42 inch flange
that a thinner flange produces higher bolt loads
0
0 , • , 8
" " ..
Removal Sell"ence
16 18
" "
during disassembly. Notably, the peak load ratio
is 2.24 times the original bolt load meaning Figure 5. Boll load raflo durmg disassembly of
there is a real possibility of yielding the stud. 24in. class 300 pipe to blindjlange with varying
thicknesses
There are geometrical differences between the
class 300 and class 600 flanges other than just The same simulation was run for a pipe to pipe
the thickness. To focus purely on the thickness flange joint with similar results. The half thick-
the 24 inch class 300 flange was modelled with nes s flange had unexpectedly low bolt ratios,
varying thickness while keep all other dimen- but the remainder of the thicknesses fo llowed
sions constant. The thicknesses modelled were the expected pattern. It was noted during the
original thickness (I), 0.51, 1.5t, and 2.0t. The model that the half thickness fl ange deflected to
pipe to blind flange model demonstrated the ex- the point where the two flange faces contacted
pected behaviour with the bolt load ratio de- each other, reducing the energy stored by the
creasing with increasing flange thickness. fl ange and explaining the unexpectedly low bolt
ratios.
_ 2.0t, cross, pipe-pipe Investigation into the incident identified fai ling
o
o 2 , 6 8 10 12 14 16 18 20
\ to follow the flange un-bolting procedure as a
22 24 contributing factor. The procedure requires the
Removill Sequence
Figure 6. Bolt load rallo durmg dlsassemhly of 24m. flange seal to be broken while there are four
class 300 pipe 10 pipejlal/ge with varying thicknesses. studs remaining in the joint. This helps mitigate
any unexpected release of energy. In this in-
In the 2002 finite element simulation it was stance it would not have prevented the leak, but
identified that removing the studs in rotational it would have reduced the risk to the craftsmen
order rather than a star pattern resulted in lower disassemb ling the joint.
bolt load ratios. The 24 inch class 300 flange
simulation seen in Figure 5 was run again re- As was reported at the 2014 Ammonia Safety
moving the bolts in rotational order. The results Symposium workshop, another incident in-
confirmed those of the 2002 study with the max- vol ved a major girth flange sealed with a lip seal
imum bolt load ratio seen only 1.51 compared gasket that was being disassembled while inter-
with 2.24 for the star pattern. nal pressure remained. The nature of a lip seal
gasket allowed significant unbolting to occur
1.6 without loss of containment. However the lip
,.. seal gasket had distorted to the point where it
required replacing which is not a tri vial task,
1.2 making thi s a significant incident. While the
flange was being disassembled the wamjng
O.st, clockwise, blind
flange
sign s that internal pressure remained were pre-
E 0.8 sent to the technicians, but were not recogni zed .
••
~ 0.6
...... 1.Ot, clockwise, blind
flange The warning signs reported were the remaining
S - l .St, clockwise, blind
0. 4 flange
studs becoming progressively harder to remove
- 2.Ot, clockwise, blind as the flange was disassembled.
0.2 - '---"'n'!""
!l,',___---'
o Case History - Near Miss, Methanex New Zea-
o 2 4 6 8 10 12 14 16 W 20 22 M land Ltd
Removal Sequence The bottom flange on a vertical heat exchanger
Figure 7. Bolt load ratio during disassembly of 24in. was being di sassembled on a plant that had been
class 300 pipe to blind flange, studs removed in rota-
shut down for four years. The nuts were being
tiona/ order
gas cut as they were seized, while cutting the
fourth to last stud the nut rapidly and unexpect-
Case Histories
edly di slodged and fell to the ground twenty feet
There have been several incidents due to flanges below.
be ing di sassembled when the system was not
completely pressure free.