An assessment of risk and safety in civil aviation - Industrial Centre

48 M. Janic / Journal of Air Transport Management 6 (2000) 43}50
Fig. 2. Dependence of the average fatal accident rate on the number of airline #ights (1970}1997).
Source of data: Internet (1998).
The risks associated with di!erent aircraft types
are found by regressing the number of accidents per
aircraft type AR
, on the number of #ights per aircraft
type F
, and the average age of particular
aircraft type E
. The data cover the period from the
entry into service of a particular aircraft type to 1992.
The aircraft types covered are Fokker F28, Fokker
F70/F100; Airbus A300, A310, A320; Lockheed L1011;
British Aerospace BAe146; Boeing B727, B737-1/200,
B737-3/4/500, B747, B757, B767; McDonnell Douglas
DC9, MD80 (Internet, 1998; Walder, 1991). The equation
used is
AR "1.206 #1.743F #0.900E
(0.692) (6.355) (3.887),
R"0.929; F"84.640; D="1.823; N"16. (4)
The overall equation and individual coe$cients are signi"cant
at the 1% level. The equation may o!er an
explanation of the past but this information was not
available in advance to those involved. The equation
indicates that the cause of air accidents could stem
from the existence of geriatric factors that escalated
faster as aircraft are utilized more and age. Because
the operational hypothesis is that aircraft accidents happen
as random events, the equation does not imply that
more used and older aircraft have been less safe. It
indicates that the risk of traveling in these aircraft is
higher.
5.2. Probabilistic assessment
The probabilistic assessment of accidents uses
a sample of 259 accidents over the period 1965}1998. The
distribution of time intervals between these events is
shown in Fig. 3. A simple calculation provides an estimate
of the average accident rate: λ7.818 accidents per
year or λ0.020 accidents per day. An analysis of the
time intervals between accidents, independent of aircraft
type, indicates they have been independent and exponentially
distributed (a χ test con"rms the hypothesis
matching the empirical and theoretical data, Ang and
Tang, 1975). This o!ers con"rmation that the observed
pattern of accidents can be treated as Poisson process.
Using the exponential distribution seen in Fig. 3, it is
possible to assess the probability of an occurrence of an
air accident. If there is unlikely to be any improvement in
safety features then this distribution can be used for
assessing the probability of future events. Fig. 4 illustrates
the probability of the occurrence of at least one air
accident per period t. This probability rises over time
until the event. For example, the probability of at least
one accident by the following day from now is about 0.02,
by next month 0.45, by six months 0.97, and by next year
0.999.
5.3. Assessment of deaths and injuries
To complete the analysis of past accidents, the distributions
of air accidents, fatalities and survivors per aircraft
category can be separated. Aircraft can be treated as
turbojets, turbo-props and piston-engine; see Table 2.
The largest number of accidents involved turbo-prop
aircraft but the greatest number of people involved in
The accidents involved aircraft types, Boeing B727, B737, B747,
B757, B767, B777 (no event); MD80, DC10, MD11; Lockheed L1011,
Airbus A300, A310, A320, A330 (no events), A340 (no events); Fokker
F28, F100; British Aerospace BAe 146; Embraer EMB-110 Bandeirante,
EMB-120 Brasilia; Dorrnier 228; and Saab 340.