Exploitation of Sensitivity Derivatives for Improving Sampling Methods

Many application codes, such as Ž nite element structural analyses and computational  uid dynamics codes, are capable of producing many sensitivity derivatives at a small fraction of the cost of the underlying analysis. A simple variance reduction method is described that exploits such inexpensive sensitivity derivatives to increase the accuracy of sampling methods. Five examples, including a Ž nite element structural analysis of an aircraft wing, are provided that illustrate an order of magnitude improvement in accuracy for both Monte Carlo and stratiŽ ed sampling schemes.