Estimating risk of failure of engineering structures using predictive likelihood

Colin C. Caprani, Dublin Institute of Technology
A. Bordallo-Ruiz
E.J. O'Brien

Document Type Conference Paper

In Third National Symposium on Bridge and Infrastructure Engineering in Ireland, eds. N.A. Ni Nuallain, A. O'Connor and K. Gavin, Trinity College Dublin and University College Dublin, p.139, 8 pp.


It has been common engineering practice to define characteristic values for loading and capacity of structures in order to assess the structural capacity of existing structures. This approach, yet practical and intuitive, lead to the comparison of deterministic values (characteristic values) that had to represent all the variability of the problem and is considered to be conservative, as usually loading is overestimated and capacity underestimated, yielding to calculations with high safety margins for the extreme events.

Probabilistics methods have tried to overcom e this limitation by computing the overall probability of failure (Pf) for the lifetime of the structure, taking into account the real probabilistic distribution of both loading and resistance.

In this paper, Predictive Likelihood (PL) is presented as a powerful method to determine the lifetime distribution for loading and resistance. From these lifetime distributions the probability of failure is computed. An example of the application of the proposed method is finally presented. The result obtained using PL is then compared with the numerical approximation for the exact lifetime probability of failure.