Page 91 - Plant design and economics for chemical engineers

GENERAL DESIGN CONSIDERATIONS 69
Since process and operating procedure changes are often made during or
shortly after plant startup, it is strongly advised that hazard assessment not stop
after startup. Rather, periodic hazard-assessment studies should be used to
define the hazard potential of such changes throughout the life of the facility.
The average time between reviews is about three years; more hazardous
facilities are reviewed more frequently.

Fault-tree Analysis
The fault-tree analysis (FTA) is primarily a means of analyzing hazardous
events after they have been identified by other techniques such as HAZOP. The
FIA is used to estimate the likelihood of an accident by breaking it down into
its contributing sequences, each of which is separated into all of its necessary
events. The use of a logic diagram or fault tree then provides a graphical
representation between certain possible events and an undesired consequence.
The sequence of events forms pathways on the fault tree, provided with logical
And and Or gates. The And symbol is used where coincident lower-order events
are necessary before a more serious higher-order event occurs. By multiplying
the probabilities of each event in this set, the probability of the next higher-order
event is obtained. Correspondingly, when the occurrence of any one of a set of
lower-order events is sufficient to cause a more serious higher-order event, the
events in the set are joined by an Or gate’ and the probabilities are added to
obtain the probability of the higher-order event. Probabilities of the various
events are expressed as a yearly rate. For example, a 1 X 10m3 chance occur-
rence per year would represent an event that average-wise would occur only
once every 1000 years. Estimation of failure rates with any precision is generally
difficult because of the limited prior data. In such cases, information from
various sources is used and then revised to incorporate information that is
site-specific.
Once a fault-tree analysis has been completed, it becomes rather easy to
investigate the impact of alternative preventive measures. For example, in the
developmenf of a FTA for Fig. 3-1 and its associated HAZOP study presented
in Table 4, Ozogt has determined that the most probable event is a liquid
release from the storage tank (Event 6) due to overfilling. However, by adding
an independent high-level shutoff to the tank-truck unloading pump, the proba-
bility of a liquid release by this event is significantly reduced and Events 12 or
13 (PV-2 closed) become the most probable events. The probability of these
events, in turn, could be reduced by the installation of an independent low-pres-
sure alarm to the tank. This process of reducing the probability of the most
probable event could be continued until an overall acceptable risk level is
eventually achieved.

W. Ozog, C/tern. Eng., 92(4):161 (Feb. 18, 1985).