Research, Based On Information Entropy Theory Of Structural Reliability Problems

This paper presents a new mathematical model for a class of structural optimization problems with reliability constrains. The main idea is to transform the reliability constrains into an infinite number of constrains valued in the close ball. The radius of this ball is equal to the value of the reliability index and the center of the ball is at the origin on the n -dimensional standard normal space obtained by transforming the basic random variables of the problem into the standard normal random variables. Thus we change the original problem into a semi-infinite programming. To solve this model, we give two methods: one is a so-called "outer approximation" algorithm and the other is the "aggregate function" method. Our numerical results support the viability of the former.