Exposé 1 (14h)

• Titre : Efficient Methods of Sampling Uncertain Variables in Risk Assessment Models
• Intervenant : Nicola Pedroni (ECP)

Models for the assessment of the risk of engineering systems are affected by uncertainties, due to the randomness of several phenomena involved and the incomplete knowledge about some of the characteristics of the system. Such uncertainties have to be (i) quantitatively represented coherently with the information available on the system and (ii) accurately and efficiently propagated onto the model output(s) (whose uncertainty must be estimated for a realistic assessment of the system failure/safety behavior). The seminar mainly addresses issue (ii) above. Monte Carlo Simulation, based on the repeated random sampling of possible model inputs and the running of the system model for the different inputs sampled, offers a powerful means for propagating uncertainty. A very large number of random realizations of the uncertain input parameters are typically necessary for a deep exploration of their ranges and a robust estimation of the model output uncertainty; for each input realization sampled, the model simulating the system behavior must be run: the resulting computational cost may be very high and at times impractical. This calls for advanced simulation techniques that allow performing robust uncertainty propagation while reducing as much as possible the number of input samples drawn and the associated computational time. Some advanced simulation methods are described and their efficiency is evaluated with respect to the difficult task of estimating very small failure probabilities of engineering systems.


Exposé 2 (15h)

• Titre : Non-Homogeneous Redundancy Allocation in Series-Parallel Multi-State Systems
• Intervenant : Yan-Fu Li (ECP)

The multi-state system (MSS) and its components have a range of performance levels from perfect functioning to complete failure. The series-parallel architecture is among the most popular models of reliability based design. Identical redundant elements are considered in order to achieve a desirable level of system availability. Redundancy allocation is a family of well-known reliability/availability optimization problems. The non-homogeneous type of redundancy allocation in series-parallel multi-state systems is among the most difficult ones. This presentation will give an overview of the redundancy allocation in series-parallel MSS, the existing approaches to solve this problem and the recent developments by the presenter and his colleagues including the optimization under uncertainty.