Abstract: The stress concentration factors (SCF) is one of the most important parameters in reliability analysis of mechanical structures, because the precision of the SCF predicted would affect the computation results of reliability analysis directly. Generally, the SCF is difficulty to be obtained adopting theoretical method, and at the present time some experimental methods are usually adopted. Although the numerical solution of the SCF can be obtained accurately applying the experimental methods, the relation between the basic random variables and the SCF can’t be given directly and which will bring some difficulties for the further reliability analysis. As an alternative approach, the finite element method (FEM) is used to gain the SCF database, and the neural network (NN) has been developed to simulate the explicit expression of the SCF based on the database obtained by FEM.Then the back-propagation(BP)artificial neural network are used modeling the relation of structural input and output,using this model.Monte Carlo and the reliability theory have been combined to solve the structural reliability problems in the case of the stress concentration.

Key words: stress concentration factors, FEM, neural network, reliability analysis