Structural reliability analysis due to the great number of random variables or large number of simulations needed may result in a high computational cost. Two techniques largely used for structural reliability assess are Monte Carlo Simulation and the analytic methods FORM/SORM. These may present some inaccuracy in the assessment of the probability of failure. The Monte Carlo Method is easy to implement and absolutely general, but the great number of required simulations may result in high computational cost making the application impracticable. This work used a trained neural network to substitute the structural analysis needed for each Monte Carlo Simulation, in order to reduce the computational cost. The applications produced good results with low computational cost, certifying its application viability.
Structural reliability; Monte Carlo Method; Neural Networks
