| The
calculation of the failure of a system or
error in the context of a complex measure
is a combination of probabilities for elementary
that of the overall function represented.
In the case of a failure
of the trees, the probability of failure
is seen in the specialized tools as a deterministic
value, but in practice it is even a random
variable which seeks to publicize the dispersion
and especially upper limit (most likely
significant portion of the feared event).
In the case of complex
measures, the distribution of overall measurement
error is determined by simulation (Monte
Carlo method) or an analytical approach
that can be put in place for the laws of
elementary simple distributions (Gaussian)
propagating through the basic components
(addition, average ...). These 2 methods
are limited in practice because they can
not be applied to systems consisting of
multiple sources of errors or because they
lead to increased global uncertainty, which
leads to oversize the quality of the components
to make implemented.
The formalism developed
by RMS allows the calculation of the distribution
of the overall failure rate based on distributions
of probabilities of default basic components,
even when some distributions are strongly
non-Gaussian.
This formalism definitely
more on the following topics:
The company RMS has submitted
a
draft paper on this topic at the congress
Lambda Mu 2008. |
