Certain properties of geometric programming are used to develop a method for generating information about the optimum design for problems involving the minimization of a real valued function subject to inequality constraints. This information includes a lower bound on the global minimum value of the merit function, details of which constraints are active at the optimum and values of the characteristics associated with non-active constraints. An active set strategy is then presented to obtain the optimum design without invoking a formal search procedure. Approximations will usually be required for the merit function and constraints, and accuracy and convergence are considered. Application of these methods to problems associated with the minimum weight design of aerospace structures is demonstrated.
Read full abstract