Uncertainties in the initial conditions and parameters of process syst
ems can lead to large variations in the predicted system performance,
sometimes with catastrophic consequences. For systems with non-linear
dynamics, often the only generally applicable approach to assess the i
mpact of these uncertainties on system performance is a search through
the operation and parameter range of interest by direct integration o
f the system equations. A methodology is presented that can be used to
assess process reliability and safety under such uncertainties much f
aster than direct integration. The methodology extends the capabilitie
s of a previously developed discrete state transition modeling approac
h to include capability for the determination of initial conditions th
at lead to desirable system operation. Implementation of the methodolo
gy on an example system taken from nuclear reactor dynamics shows that
: a) the methodology can handle both small and large uncertainties in
system parameters and initial conditions, and b) fast conventional app
roaches such as perturbation analysis may lead to the choice of more r
estrictive or narrower operational ranges for the system than that req
uired by reliability/safety considerations. (C) 1996 Elsevier Science
Limited.