This research considers the problem of determining parametric bounds for the performance criterion of perturbed nonlinear programming problems. Initially special problem structure, such as separability will be exploited, the aim however is to apply the results to so-called "factorable functions" which occur in a wide variety of problems. Thus the eventual outcome of the research could have wide applicability. The central idea is to use branch-and-bound techniques to generate regions in which sequences of approximating convex programming problems can be used to get bounds, then piece together the information from different regions to create global bounds. The goal of the research is to produce an algorithm for general nonconvex parametric problems. The methods will be implemented on practical applications and user friendly code will be produced. This research considers the problems of producing parametric bounds for the performance criterion of perturbed nonlinear programming problem. It is expected that the research will lead to techniques and computer implementations of direct use in computing the solution to difficult nonlinear optimization problems.
