The application of probabilistic safety assessment in decision process
es depends to a great extent on the quality of the input data. The pla
nt specific acquisition of these data must be performed in a way which
allows a reliable prediction of system failures in the future. Comfor
table computer tools are available to facilitate data processing and t
o make use of data that are already available via the computerized pla
nt information system. The correct classification of failure occurrenc
es with respect to the PSA relevant failure modes requires thorough ev
aluation of all information available in maintenance records, test rep
orts, and licensing event reports. The flow of information relevant to
future data acquisition must be properly organized The application of
generic common cause failure data requires careful evaluation of the
underlying failure causes and of the level of defence achieved in the
plant under consideration to prevent too pessimistic analysis results.