Vol.10, No2, 2010, pp.143-148
UDC  621.01


Philippe Jodin

LABPS, Université Paul Verlaine – Metz & ENIM – France


Since several years, the trends to designing structures are towards probabilistic design more than deterministic design. This is based on the fact that a null risk does not exist. Engineers now prefer to speak in terms of probability of failure. This means that a given risk of failure is accepted. In counterpart, a cheaper structure is expected.

The problem is in determining the pertinent failure criterion and the corresponding random input variables. If the laws of distribution of mechanical properties are generally well known, it is more difficult to have those representing the type and size of defects, or loading. Some experiments should be done to obtain the desired data.

There are several reliability indexes that are presented in this work. We will focus with the following on the Hasofer and Lind’s index, that exhibits some theoretical advantages in conjunction with easiness of computation.

Four methods of computation are presented: a “black box”, associated with finite elements software, a numerical method implemented with the help of symbolic computation software, the same but operating with a spreadsheet and the well-known Monte-Carlo method. A calibration test and an example have been treated with these methods and the results are compared in terms of precision, speed, easiness of programming and cost.

Keywords: structures, reliability, reliability indexes, numerical methods, probability of failure, Hasofer-Lind’s index

full article (220 kB)