INFEASIBILITY AND NON-OPTIMALITY TESTS FOR SOLUTION SPACE REDUCTION IN DISCRETE OPTIMAL-DESIGN

Techniques employed for nonlinear design optimization with discrete va riables are studied in the context of a particular problem arising fro m the design of a gear train. The mathematical model formulation was p resented in an earlier article. In this sequel, a solution derivation is described, patterned as a multistage process. After After certain r eformulation and relaxation, a variety of infeasibility and non-optima lity tests are performed, greatly reducing the size of the space conta ining the global optimum. Methods used to investigate the remaining sp ace do not guarantee a global optimum, but could be replaced by more c ostly methods that do provide such guarantees. A global infimum is gen erated, bounding any improvements on the best known solution.