STRUCTURAL RELIABILITY-ANALYSIS USING NEURAL-NETWORK

In estimating reliability of a structural system, a limit-state functi on is needed to relate the structural state (failure or safety) to ran dom variables of the system. However, it is not easy to obtain such an explicit function for complex structures. As a consequence, structura l analysis must be performed repeatedly to check the structural state, which is very expensive. We develop an approximate limit-state functi on by using a neural network. Orthogonal factorial designs are selecte d as learning data for the network. An ''active learning algorithm'' i s proposed to enable the network to determine important failure region s by itself and also to do further learning at those regions to achiev e a good fitness with the real structural state there. The validity of the method is illustrated through numerical examples.