Page 37 - Dynamic Loading and Design of Structures

Page 24

years, further uncertainties are introduced, and should be taken into account as far as possible
through appropriate judgement. In the case of time-varying loads, these uncertainties may
have both systematic and random components. The former can be particularly important for
some man-made loads, such as traffic loads on bridges, whereas the latter may include poorly
understood environmental influences as well as purely random effects.
The characteristic value of a time varying load Qk is normally chosen so that events during
which the observations exceed the characteristic value are fairly rare. Typically, characteristic
values in Eurocode 1 are prescribed for an exceedance probability p=0.02 and a reference
period t =1 year (European Standard, 2000; Eurocode 1.1 Project Team, 1996). Thus, the
r
characteristic value Q may be estimated from
k

In the above the distribution for the annual maximum is used and the reference period is also
1 year. If, for example, the distribution for the monthly maximum was available instead, then
for the same criteria (i.e. 2 per cent exceedance probability during one year) and providing
monthly maxima were mutually independent, the characteristic value could be estimated from

Note that using a distribution based on observations from a shorter time unit results in a much
higher fractile required for estimating a characteristic value based on the same criteria as
before. Clearly, many more observations would be needed for the monthly distribution to be
sufficiently well described at a 99.8 per cent fractile, than the annual distribution at a 98 per
cent fractile. On the other hand, much longer (unit and total) observation periods are
associated with the estimation of the distribution of annual maxima. The message is that
predictions cannot be improved simply by changing the basis of the distribution used. The
most appropriate observation period should be determined on the basis of the characteristics
of the action being modelled and the capabilities of the devices/methods used for
measurement.
A useful concept in the treatment of time varying loads is the return period T defined as the
average time between consecutive occurrences of an event. Again assuming independence
between events, and denoting with p the probability of occurrence of the particular event
considered, the return period may be determined from

(1.29)

that is, the return period is equal to the reciprocal of the probability of occurrence of the event
in any one time interval. In many cases, the chosen time interval is 1 year and p is determined
as the probability of occurrence during a year, so that the return period is the average number
of years between events.