Most global optimization systems provide little output beyond an estim
ated optimum point or a failure message. Since nonlinear global optimi
zation problems can be quite complex, more information is needed if th
e solution is to be trusted or if the model requires modification. We
describe a system which provides measures of confidence in the returne
d solution, and helpful information in the case of infeasibility, unbo
undedness, or early termination. (C) 1997 Elsevier Science Ltd.