Design of dynamic systems based on dynamic fault trees and neural networks

Reliability design of static systems has been widely developed in recent years. Fault trees (FT), genetic algorithms (GA) and neural networks (NN) are among the most common methodologies used in such tasks. However, because of the complex behavior of dynamic systems, less attention has been paid on their design. With the thorough research into dynamic fault trees (DFT) and neural networks, which could be used in reliability analysis of dynamic systems, the design of dynamic systems becomes possible. In this paper, a hierarchically modular design method based on DFT and NN are proposed to solve this problem. Fault tree of the system is constructed, a linear-time modular method is performed to find out all the static and dynamic subtrees and the reliability demand of each subtree could be determined. Then the static subtrees are optimized using traditional methods, and each dynamic subtree are mapped into feed-forward recursive neural networks, which could be trained to obtain the optimal design parameters.