In order to reduce the computational expense in identifying limit state functions, an efficient adaptive convex hull strategy is presented in this paper. High fidelity simulations are performed at selected subset of an initial design of experiments. A cluster of unwanted behavior is identified and a convex hull is wrapped around it. The faces of the hull form the limit states. The hull is progressively updated with limited additional sample evaluations around the boundary of initial hull. Once the boundaries or areas converge, accounting for uncertainties and estimating reliability indices are straight forward. © 2012 by Palaniappan Ramu.