Improvement of Design Reliability for Buckling Load Maximization of Composite Cylinder Using Genetic Algorithm with Recessive-Gene-Like Repair.

Abstract

A genetic algorithm is applied to obtain the optimal stacking sequences for maximization of the buckling load of composite cylinders. In the problems, three constraint stacking rules are implemented. The first is that the angle plies of laminates have to be balanced. The second is that more than four contiguous plies of the same fiber orientation are not allowed. The third is that the difference in fiber orientation between adjacent plies must not be more than 45 degrees. The difficulty of handling the combinatorial constraints in genetic optimizations is overcome by adoption of a new repair system. When a chromosome violating the stacking constraints is decoded to a stacking sequence, the new repair system operates. The new repair system does not alter genes of the chromosome but changes only the decoding rules to introduce the constraints. This is similar to recessive genes in biology. By using this new repair system, a chromosome of {2 2 211 1 0 0 0} is decoded to [90/90/90/45/90/- 45/- 45/ 0/45] S. The design reliability of the genetic algorithm with the new repair system is examined in detail by comparing the reliability with a conventional penalty method. As a result, the genetic algorithm with the new repair system is shown to provide higher design reliability.