A gradient-based approach for discrete optimum design

In this paper, structural optimization with discrete variables in engineering design is modeled and investigated as a zero-one programming problem. The zero-one programming problem is first reformulated as an equivalent continuous problem through replacing the zero-one constraints by complementarity constraints, then as an equivalent ordinary nonlinear programming problem with the help of the NCP(nonlinear complementarity problem) function. Furthermore, an aggregate function method is introduced with the aim to simplify computation in the following augmented Lagrangian method. Numerical experiments on discrete optimum design showed the proposed method is promising.