The minimum weight design of a composite stringer-stiffened panel is performed by a genetic algorithm search. Design variables include the stacking sequence of three laminates as well as four dimensions of the hat-stiffened panel geometry. Buckling and strength analyses were performed by the PANDA2 program. Because genetic algorithms require a very large number of analyses, response surface approxiamtions of the critical failure margin are used during the optimization. The approxiamtions are quadratic least-squares-fit polynomials obtained based on PANDA2 analyses from previous generations of designs. For these approximations, the effects of changes in stacking sequences are represented by five lamination parameters per laminate. With a practical limit of about 40 hours of computer time, the approximations permit the genetic search to proceed further to convergence. As a result, the designs obtained by the genetic algorithm with approximations are significantly lighter on average than those obtained without the use of approximations.
展开▼