Reliability-Based Design of Thermal Protection Systems with Support Vector Machines

Abstract

The primary objective of this work is to develop a computationally efficient and accurate approach to a reliability analysis of thermal protection systems using support vector machines. An adaptive sampling approach is introduced, which informs an iterative support vector machine approximation of the limit-state function, which is used for measuring reliability. The proposed sampling approach efficiently adds samples along the limit-state function until the reliability approximation is converged. This methodology is applied to two mathematical functions to test and demonstrate the applicability. Then, the adaptive sampling-based support vector machine approach is applied to the reliability analysis of a thermal protection system. The results of all three problems highlight the potential capability of the new approach in terms of accuracy and computational savings for determining thermal protection system reliability.