Approximate dynamic fault tree calculations for modelling water supply risks
Artikel i vetenskaplig tidskrift, 2012
Traditional fault tree analysis is not always sufficient when analysing complex systems. To overcome
the limitations dynamic fault tree (DFT) analysis is suggested in the literature as well as different
approaches for how to solve DFTs. For added value in fault tree analysis, approximate DFT calculations
based on a Markovian approach are presented and evaluated here. The approximate DFT calculations
are performed using standard Monte Carlo simulations and do not require simulations of the full
Markov models, which simplifies model building and in particular calculations. It is shown how to
extend the calculations of the traditional OR- and AND-gates, so that information is available on the
failure probability, the failure rate and the mean downtime at all levels in the fault tree. Two additional
logic gates are presented that make it possible to model a system’s ability to compensate for failures.
This work was initiated to enable correct analyses of water supply risks. Drinking water systems are
typically complex with an inherent ability to compensate for failures that is not easily modelled using
traditional logic gates. The approximate DFT calculations are compared to results from simulations of theorresponding Markov models for three water supply examples. For the traditional OR- and AND-gates, and one gate modelling compensation, the errors in the results are small. For the other gate modelling compensation, the error increases with the number of compensating components. The errors are, however, in most cases acceptable with respect to uncertainties in input data. The approximate DFT calculations improve the capabilities of fault tree analysis of drinking water systems since they provide additional and important information and are simple and practically applicable.
dynamic fault tree (DFT)
monte carlo simulation