A Combined Local-Global Active Set Strategy for Nonlinear Design Optimization

Local and global knowledge in rule-based forms is combined in an active set strategy for nonlinear programming design models. The combined strategy is implemented with a reduced gradient algorithm for solution of the equality constrained sub-problem. The strategy unifies previously separate local and global strategies. Algorithmic implementation and performance are discussed together with examples. This article is a sequel to a previous one describing a linear programming application of the strategy.