Reliability of Stiffened Cylindrical Shells Using Random Polar Sampling Technique

Abstract

Rational design of stiffened cylindrical shell structures calls for the calculation of reliability. Random variables occur in modelling loads and strengths. The reliability can be evaluated with ‘Monte-Carlo Simulation’ (MCS) which consists of obtaining cumulative distribution functions for each and every random variable and simulating the ultimate strength of stiffened shells for combinations of random variable values. However for MCS to be successful, the sample size should be very large. Hence methods have been proposed to reduce the sample size without however sacrificing any accuracy on reliability. ‘Point Estimation Method’ (PEM), ‘Response Surface Technique’ (RST), ‘Importance Sampling Procedure Using Design points’ (ISPUD), ‘Latin Hypercube Sampling’ (LHS) etc., are some of these methods. In this paper, a method based on ‘Random Polar Sampling Technique’ (RPST) is proposed, in which combinations of variates are obtained using a polar sampling of Latin Hypercube sampled values. A typical example has been worked out using this method.