You are on page 1of 4

Process Safety Progress (Vol.19, N o.

2) Sum m er 2000 65
We’veall heard of Mu r p h y’sLaw. It hasbeen stated in
variousforms, generallysomethinglike“If somethingcan go
w rong, it will gowrong, and at theworst possibletime.” We
often quoteMu r p h y’sLaw often in a light-hearted or joking
context. But, do wereallybelieveit?What effect doesour
belief in Mu r p h y’sLaw haveon how weactually opera t e
and design chemical plants?Isour application of Mu r p h y’s
Law appropriate? How should webeusingMu r p h y’sLaw?
T h e re are m any statem ents of M urphy’s Law , as
w ell as a large collection of corollaries, com m entaries,
postulates and other variations of M urphy’s Law . A
s e a rch of the Internet w ith any search engine w ill
reveal hundreds of hits, a large percentage of w hich
appear to be related to com puters and softw are. From
m y ow n experience w ith com puters and softw are, this
is certainly not surprising! Som e com m on statem ents
of M urphy’s Law include:
• If anything can go w rong, it w ill.
•If som ething goes w rong, it w ill at the w orst possible
t i m e .
•If there are tw o or m ore w ays to do som ething, and
one of those w ays can result in a catastrophe, then
som eone w ill do it that w ay.
•If som ething just cannot go w rong, it w ill anyw ay.
•The probability of anything happening is in inverse
ratio to its desirability.
• N ature alw ays sides w ith the hidden flaw .
A nd then there is O ’Toole’s Com m entary on M ur-
phy’s Law : M urphy w as an optim ist.
M urphy’s Law has becom e a part of the English
language - nearly everybody understands w hat w e
m ean w hen w e refer to M urphy. W e frequently quote
M urp hy’s Law and its variations, often in a joking
m a n n e r, w hen som ething goes w rong. B ut, w as there
really a M urphy, did he really state this law , and w hat
did he have in m ind? D oes the universe really operate
in accordance w ith M u rp h y’s Law ? D o w e re a l l y
believe in M urphy’s Law , and does it have any im pact
on how w e behave? W hat relevance does M urphy’s
Law , or a belief in M urphy’s Law , have to the opera-
tion and design of chem ical m anufacturing facilities?
Yes, there really w as a M urphy [1]. Captain Edw ard
A. M urphy w as an engineer w orking on U nited States
A ir Force Project M X 981 at Edw ards A ir Force B ase,
C a l i f o rnia in 1949. The project objective w as to deter-
m ine how m uch sudden deceleration a person can
stand in a crash. In the testing program , a volunteer
w as strapped into a rocket propelled sled, and his
condition w as m onitored by a group of sensors in a
h a rness as the sled w as rapidly halted. Follow ing one
test, it w as found that no data w as recorded because
all of the sensors had been w ired incorrectly. Captain
M urphy cursed the responsible technician, saying, “If
there is any w ay to do it w rong, he’ll find it.”D r. John
Paul Stap p of the U S A ir Force, described as “the
fastest m an on earth”in new s accounts of the tim e,
w as the volunteer rider in the sled that day. At a press
conference few days later, D r. Stapp attributed the test
p rogram ’s good safety record to a firm belief in “M ur-
phy’s Law .”The press picked up M urphy’s Law , and it
has been a part of the English language ever since. D r.
Stapp continued to ride rocket pow ered sleds, includ-
ing a 1.4 second deceleration from 632 m iles per hour
in a 1954 test. H e subseq uently w orked for the
N ational H ighw ay Tr a ffic Safety A dm inistration and
did research to im prove the design of autom obile seat
belts. D r. Stapp died in N ovem ber 1999 at the age of
89 [2].
A ctually, as M atthew s [3] notes, M urphy had just
restated som ething w hich had been observed for
m any years:
Thebest laid schemeso’ micean’ men
Gangaft agley.
- Robert Burns, 1786
Was Murphy Wrong?
Murphy’s Law in Operation and Design of
Chemical Plants
Dennis C. Hendershot
Rohm and H aas Com pany, Engineering D ivision, P.O . Box 584, Bristol, PA 19007
I had never had a pieceof toast
Particularly longand wide
But fell upon thesanded floor
And alwayson thebuttered side.
- Jam es Payn (Victorian satirist), 1884
So, is M urphy’s Law true? W hile w e often quote its
m any variations, m ost of us w ould adm it that it is not
really “true”if asked. W e w ould adm it that our “belief”
in M urphy’s Law is a result of selective m em ory. W e
easily rem em ber the tim es that things go w rong, but
all of the tim es that things go as expected are re a d i l y
forgotten. The w orld behaved in the w ay w e expected
(and w anted) it to behave, and that is not m em orable.
A ctually, it turns out that M urphy’s Law can be
show n to be correct in som e situations. In an intere s t-
ing article, Robert A . J. M atthew s has identified cases
in w hich the universe really is “against us”[3]. Som e
of these are a result of people’s poor understanding of
the law s of probability, and others re p resent true
physical behavior of system s. For exam ple:
• If all lines in the superm arket are assum ed to
m ove at the sam e average rate, w ith random variation
in the tim e to serve each custom er, there is only a 1/3
chance that the line you are standing in w ill be faster
than the one on either side of you. A nd, if you can
see 10 lines, there is only a 1/10 chance that yours
w ill be the fastest. So, m ost of the tim e, you w ill
observe other lines m oving faster than yours.
• Start w ith a draw er containing 10 pairs of socks,
and random ly lose socks. A probabilistic analysis
reveals that, by the tim e you have lost half of your
socks, it is four tim es m ore likely that you w ill have a
draw er full of odd socks, rather than five com plete
pairs. The m ost likely outcom e is tw o com plete pairs
and six odd socks.
• M atthew s analyzes Payn’s observation that the
toast alw ays lands butter-side dow n, and determ i n e s
that this is indeed the expected outcom e. The w eight
of the thin layer of butter is not significant, nor are
a e rodynam ic effects. The toast lands butter side dow n
because the torque induced by gravity as the toast
falls results in a rotation rate that causes the butter
side to be dow n w hen the toast is dropped from the
height of a typical table. A ll of this can be related to
basic physical constants of the universe, w hich deter-
m ine everything from the rotation rate of the toast to
the m axim um reasonable size for a m am m al living in
the earth’s gravitational field (and there f o re establish-
ing the typical height of a table). M atthew s’conclu-
sion: the toast lands on the butter side because the
universe is designed that w ay!
So, w e can conclude that, for at least som e cases,
M urphy’s Law does correctly predict the m ost likely
outcom e. But, these cases are really the exception.
M ost of the things w e attribute to M urphy really do
re p resent selective m em ory of bad experiences or
events in w hich a failure or m istake has occurred. But,
w hat is the usual outcom e of failures and errors? In
m any cases, there is no bad outcom e. The universe is
really a very benign and forgiving place, and allow s
us to get aw ay w ith errors and failures m ost of the
tim e. If this w ere not the case, it is unlikely that any of
us w ould have survived childhood. W e can all rem em -
ber dum b things that w e did as children, and m ost of
us survived those m istakes. If M urphy w as really right,
and the w orst possible outcom e occurred, m ost of us
w ould not be here.
Consider som e exam ples of failures in every day
life, in high risk system s, and in the operation of
chem ical plants:
• M any years ago, on th e w ay to w ork, I ran
t h rough a stop sign at a fairly busy intersection at
about 40 m iles per hour. I don’t know w hy - I know
the intersection w ell, and pass through it tw ice a day,
every w orking day. B ut, on this particular day m y
m ind w as som ew h ere else, and I did not stop .
A ccording to M urphy, I should have been hit by a
passing gasoline truck at about 60 m iles per hour and
killed in a large fireball. In fact, there w as no traff i c
com ing, there w asn’t even a policem an w atching the
intersection, and the only consequence w as that I got
to w ork a few seconds earlier.
• O n A ugust 23, 1962, a U nited States A ir Forc e
Strategic A ir C om m and (SA C ) B -52 bom ber arm e d
w ith nuclear w eapons got lost due to a navigational
e r ror w hile on a routine airborne alert route over the
A rctic O cean. The bom ber flew over 1,300 m iles on a
course w hich w ould have taken it into the air space of
the Soviet U nion. A bout 300 m iles from Soviet air-
space, ground control detected the error and ordere d
an im m ediate change of course. There is no indication
that the Soviet air defense system s detected the B-52
in this incident. It is frightening to think of the poten-
tial conseq uences had this error occurred a few
m onths later, in O ctober 1962, during the Cuban m is-
sile crisis. SAC w as still using the airborne alert ro u t e ,
w hich its review subsequently categorized as “high
risk,”through the first ten days of the m issile crisis [4].
• A reactant w as incorrectly charged to a batch
re a c t o r, and the result w as an unanticipated runaw ay
reaction. The engineers and chem ists w orking on the
p rocess w ere unaw are that the reaction could occur.
N ote that this occurred before H AZO P w as com m only
used - the reaction should have been identified fro m
the H A ZO P deviation “M O RE REACTAN T”. In a w orld
ruled by M urphy, the reactor w ould have blow n up.
In the real w orld, the em ergency relief system on the
reactor w as large enough for the incorrect charg e ,
even thou gh it had no t been designed for the
unknow n reaction.
• Because of confusing control system displays, an
operator m oved a group of rem ote control valves on
Reactor Train A, thinking he w as operating the corre-
sponding valves on the identical Reactor Train B. W ith
som e operations in this equipm ent, this error could
have resulted in a serious incident. In fact, w hen it
w as done, a batch w as in a vacuum distillation step,
and the consequence w as a purge of the batch to the
reactor overhead, and no release of chem icals [5].
• A flam m able solvent leaked from a pre s s u r i z e d
reactor into a building and caused a explosive m ixture
66 Spring 2000 Process Safety Progress (Vol.19, N o. 2)
Process Safety Progress (Vol.19, N o.2) Sum m er 2000 67
in the building. M urphy w ould predict that the cloud
w ould find an ignition source and there w ould be a
l a rge, catastrophic explosion, but, in this case there
w as no ignition and the cloud dispersed.
The often forgiving nature of the w orld has an
adverse im pact on people’s perception of risk - w e
begin to expect the w orld to alw ays behave that w ay.
Each of our individual experience w ith the w orld is
lim ited, and serious incidents really are relatively rare
events. M ost of the tim e, w e w ill get aw ay w ith at risk
behaviors. M any years ago, at an A IChE short course
on statistical design of experim ents, one of the princi-
ples discussed by the instructor, D r. Stuart H unter, w as
“R a re events do happen, but not to m e”[6]. That prin-
ciple m ay be appropriate for interpreting the results of
a group of experim ents, but it is not a good rule for
d e t e rm ining w hat is safe behavior. W e expect injuries
and chem ical incidents to be extrem ely rare. Just
because w e get aw ay w ith a certain behavior m any
tim es does not m ean that the behavior is acceptable.
Each of us, as an individual, has too sm all a data set
of experience to reach a m eaningful conclusion on
this issue based on our ow n individual experience. If
the unsafe behavior persists, eventually the rare event
(an incident or injury) w ill occur to som ebody, and
perhaps to m e.
U nfortunately, w e often do reach the conclusion
that a particular activity is safe based on our ow n past
experience w ith no adverse outcom es. The operator
goes up to the fifth floor of a process rack to take a
sam ple and realizes that he has forgotten his splash
goggles. N ot w anting to go dow n five floors to get the
goggles, and then back up again to take the sam ple,
he goes ahead and takes the sam ple carefully, and
nothing goes w rong. H e files that aw ay in his m em ory
and the next tim e he forgets the goggles, he takes the
sam ple w ithout goggles again. Soon that behavior
extends to other safety rule violations, and other peo-
ple begin to assum e that they w ill also be safe even if
they don’t follow the rules. O ur daily experience w ith
e r rors and failures w ill often lead us to believe that
nothing bad w ill happen. O ur behavior is show ing an
unconscious belief in M urphy’s Law - w e believe that
if som ething can go w rong, it w ill. Since nothing did
go w rong, w e begin to behave as if nothing can go
w rong, and continue the at risk behavior. Eventually,
som ebody w ill get caught and there w ill be an injury.
This unconscious belief in M urphy’s Law is also the
reason w hy w e som etim es find safety devices not
w orking or bypassed. I w as in a superm arket and
found a fire exit padlocked shut. W hen I told the
m a n a g e r, he said that the building had excellent fire
p rotection system s, there had never been a fire, the
exit w as not really needed, and it w as a w ay that
unauthorized people could enter the building. H e w as
exhibiting a belief in M urphy’s Law - if the w orst thing
that could have happened did happen, and a fire had
not occurred, a fire m ust not be possible and he could
lock the fire escape. I suggested that som ething bad
m ight occur soon - the local Fire M arshall m ight visit -
and the next tim e I w as in the store the lock had been
rem oved. H ow m an y custom ers w alked by th at
locked fire escape and said (or noticed) nothing
b e f o re this trouble m aking process safety engineer
m ade a fuss? H ow m any people in chem ical plants
develop a sim ilar acceptance of unsafe conditions or
failed devices for sim ilar reasons?
W e m ust all w ork to elim inate the effects of this
unconscious belief in M urphy’s Law on our behavior.
M aintaining good operating and safety discipline
re q u i res continual vigilance and w ork. W e m ust con-
tinually look for unsafe conditions and behaviors and
m ake sure they are elim inated. M ost of the tim e, these
unsafe conditions and behaviors w ill not result in an
incident. This tem pts us to ignore them , but, if w e do,
eventually M urphy w ill re t u rn and som ebody w ill get
W hen M urphy and his colleagues at Edw ards A ir
F o rce B ase, stated his law , they really had in m ind
that it is a tool for the design of devices and pro c e-
d u res. It is a plea for inherently safer design. W hen
designin g som eth in g, the d esigner m ust alw ays
assum e that “if it can go w rong, it w ill,”and design his
device or procedure to account for errors and failures.
C aptain M urphy’s son states the case for M urphy’s
Law as a design tool:
“I w ould suggest, how ever, that M urphy’s Law
actually refers to the certainty of failure. It is a call for
d e t e rm ining the likely causes of failure in advance
and acting to prevent a problem before it occurs.”[7]
Captain M urphy believed that any failure w hich
could occur, eventually w ould occur. As a designer, it
w as his responsibility to anticipate these failures, and
m ake sure his system s w ere sufficiently robust to
respond safely w hen the failures eventually did occur.
For exam ple, how could Captain M urphy have elim i-
nated the problem of incorrectly w ired sensors in the
incident w hich prom pted him to state his law ? O ne
possibility w ould have been to design a w iring system
w hich could only be assem bled in the correct w ay. If
the error had been safety critical, rather than re s u l t i n g
in the loss of data only, perhaps this w ould have been
done, as it has in m any safety system s. For exam ple,
th e d esign o f m o dern polarized electrical p lugs
re q u i res that they be inserted into the socket the cor-
rect w ay.
In the chem ical industry, the H azard and O perability
Study (H AZO P) has been described as an application of
M urphy’s Law . H AZO P is a system atic and logical w ay
of exam ining a process to identify as m any failure sce-
narios as possible, to understand the existing safe-
guards, and to identify the need for additional pro t e c-
tion. Sim ilarly, Failure M odes and Effects A nalysis
(FM EA) is based on M urphy’s Law . In FM EA, a specific
f a i l u re m ode of a device or process is assum ed to occur,
and the effects of that failure on the system containing
the device or process are determ ined. In general, m ost
chem ical process hazard analysis techniques can be
c o n s i d e red to be based on M urphy’s Law .
Inherently safer design is the m ost robust approach
to accom plishing the goals C aptain M urphy had in
m ind w hen he stated his law . In the chem ical indus-
try, this m eans elim inating hazards in the p ro c e s s
rather than controlling them . According to M urphy, all
hazard control features w ill ultim ately fail if w e w ait
long enough. W hile that risk of failure m ay be sm all,
and tolerable to society, it w ill never be zero. There-
f o re, w hen w e do a H A ZO P or other process hazard
study, our first question w hen w e identify a hazard
should be “Is there any practical w ay to elim inate this
hazard?”O nly w hen w e are satisfied that there is no
feasible w ay to elim inate the hazard should w e begin
to focus on hazard control.
Birkett [8] describes several “M urphy incidents”in
chem ical laboratories w hich, upon closer exam ina-
tion, turn out to be accidents w aiting to happen. For
exam ple, a graduate student w as condensing boranes
using liquid oxygen. W hen the cold finger containing
the liquid oxygen broke, it created a vessel full of an
explosive m ixture. W hile you m ight consider this a
M urphy incident (the cold finger broke at the w orst
possible tim e), Birkett asks “w hat the idiot w as doing
bringing boranes into close proxim ity to liquid oxy-
gen?”W hile I haven’t run into anything quite this
spectacular in a chem ical plant environm ent, I have
e n c o u n t e red situations w here a highly corrosive and
w ater reactive m aterial w as being cooled in a vessel
w ith coils containing cold w ater. W hile w e can take
m any m easures to ensure the integrity of the cooling
coils, a better solution w as actually im plem ented - a
cooling fluid w hich w ould not react w ith the vessel
contents w as identified and used.
M urp hy’s Law has its p lace in the d esign of
devices, processes, system s, and pro c e d u res. It forc e s
us to assum e that failures w ill occur, to identify and
understand the consequences of those failures, and to
design system s w hich w ill be tolerant of those fail-
u res. This w as w hat C aptain M urp hy had in m ind
w hen he and his colleagues first stated his law . M ur-
phy’s Law is the basis of m any design safety tools. It
encourages the system designer to search for inher-
ently safer system s - those w hich elim inate hazards
entirely or m inim ize them sufficiently that failures can-
not cause an unsafe outcom e.
H ow ever, a belief, m ost likely unconscious, in M ur-
phy’s Law in the operation of a facility or device can
lead us to believe that actions are safer than they real-
ly are. The chem ical industry, and life in general, is
very safe, and the w orld tends to be forgiving of
unsafe acts. M ost of the tim e w e w ill get aw ay w ith
unsafe behavior or system failures. This m ay lead to
com placency and an expectation that w e w ill alw ays
get aw ay w ith the behavior, or that it is not necessary
to fix the failed piece of equipm ent because nothing
bad happened w hile it w as broken. H ow ever, our
expectations for safety are high, and our personal
experience is not statistically significant. Eventually
w e, or som ebody, w ill get caught, and an incident w ill
occur if the unsafe behavior persists or the unsafe fail-
ure condition is not repaired.
In conclusion, w e m ust encourage the application
of M urphy’s Law in the design of system s. H ow ever,
w e m ust also w ork hard to ensure that w e and our
colleagues are alw ays alert for unsafe conditions and
behaviors, even in the absence of unsafe outcom es
from those conditions and behaviors. W e m ust be vig-
ilant in identifying and elim inating these situations,
and not fall into the trap of believing that w e m ust be
safe because no incidents have actually occurred. The
best w ay of ensuring a safe design is to design inher-
ently safer facilities w hich elim inate the hazard entire-
ly, or reduce its m agnitude sufficiently that it is not
capable of doing serious harm . Again quoting Captain
M urphy’s son [7], describing the engineers at Edw ards
Air Force Base:
“They w ere not content to rely on probabilities for
their successes. Because they knew that things left to
chance w ould definitely fail, they w ent to painstaking
efforts to ensure success.”
W e m ust do the sam e in the ch em ical pro c e s s
1. B e a r, T. “M urphy’s Law W as B orn H ere.”T h e
Desert Wings. Edw ards Air Force Base, CA, (M arch
3, 1978).
2. Martin, D. “John Paul Stapp, 89, Is D ead; ‘The
Fastest M an on Earth’.”The New York Ti m e s .
(N ovem ber 16, 1999), p. B13.
3. Matthews, R. A. J. “The Science of M urp hy’s
Law .”Scienti fic American, 2 7 9 (4), pp . 88-91
(April, 1997).
4 Sagan, S. D. “The Lim its of Safety.”Princeton U ni-
versity Press, Princeton, N . J. (1993).
5. Hendershot, D. C., and G. L. Keeports. “B a t c h
P u rge Caused by Confusing Control System D is-
plays.”P rocess Safety Pro g re s s 1 8(2), pp. 113-114
(Sum m er, 1999).
6. H u n t e r, J. S. “Statistical D esign of Engineering
Exp erim ents.”A IC hE Today Series. A m erican
Institute of Chem ical Engineers, N ew Y ork (1970).
7. Murphy, E. A., III. “M urphy W as a Perf e c t i o n i s t . ”
Scientific American 279(8), p. 8 (August, 1997).
8.Birkett, D. “The M urphy Reaction.”C h e m t e c h 2 8
(3), p. IBC (M arch, 1998).
68 Spring 2000 Process Safety Progress (Vol.19, N o. 2)

Related Interests