Structural optimization based on second-order approximations of functions and dual theory (Q579038)

Based on the dual theory of nonlinear mathematical programming and the second-order Taylor-series expansions of functions, an efficient algorithm for structural optimal design has been developed. The main advantages of this method are the generality in use, the efficiency in computation, and the capability in identifying automatically the set of critical constraints. On the basis of the virtual work principle, formulas expressed through element stresses for the first- and second- order derivatives of nodal displacement and stress with respect to design variables are derived. By introducing Saint-Venant's principle, the computational effort involved in the Hessian matrix associated with the iterative expression can be significantly reduced. This method is especially suitable for optimum design of large-scale structures. Several typical examples have been optimized to test its power.