Proceedings of the Eighth International Conference on Probabilistic Safety Assessment & Management (PSAM)
282 Determination of Reliability Index by Using Ants Colony (PSAM-0284)
Download citation file:
- Ris (Zotero)
- Reference Manager
Structural time invariant reliability assessment consists in modeling every design variable on which there are uncertainties by using a random vector. After that, a failure criterion is defined by a limit state function that defines the failure domain. To assess the structure reliability, one needs to know the joint density probability function of vector this vector. The failure probability will then be obtained by integrating the joint density probability over the failure domain.
For designers, it is not only important to obtain the failure probability but also to know the design point of the structure. Intensive researches have been carried out on these two points during the two last decades. The design point gives these two information, because it enables to calculate the failure probability by means of FORM approximation. Even if Monte Carlo simulation is an efficient approach when the limit state function is known, it is well-known that a large number of simulations is necessary to assess small failure probabilities (10−8 – 10−4 for example). This is a weakness of classical Monte Carlo simulations in the assessment of failure probability. Modern Monte Carlo simulation enables to fill this gap, but does not enable to obtain the design point. Thus, it is useful to develop methods based on global optimization to determine the design point and therefore, the failure probability.
Hasofer and Lind introduced a notion of reliability index that enables to obtain the structure reliability. To obtain the reliability index, one needs to solve a constrained optimization problem. Some methods are proposed in the literature based on gradient method such as Abdo-Rackwitz algorithm. In this paper it is proposed an evolutionary algorithm that enables to solve global search optimization problem: Ants Colony Optimization combined with GA. This algorithm is introduced take profit of the ability of Genetic algorithm to solve global optimization problem efficiently and ants colony to solve local optimization problem.
The results obtained on the tests performed on API show that it is not easy to improve significantly the result supplied by genetic algorithm. Many sensitivities analysis must be performed do to give better idea on how to tune the API parameters which are the global amplitude, the size and the exploration criteria. The results obtained are acceptable and constitute a good basis for further investigations.