Cross-Entropy-Based Adaptive Importance Sampling Using Gaussian Mixture

摘要: Abstract Structural reliability analysis frequently requires the use of sampling-based methods, particularly for situation where failure domain in random variable space is complex. One most efficient and widely utilized methods to such a importance sampling. Recently, an adaptive sampling method was proposed find near-optimal density by minimizing Kullback–Leibler cross entropy, i.e. measure difference between absolute best one being used In this paper, approach further developed incorporating nonparametric multimodal probability function model called Gaussian mixture as density. This fit complex shape functions including those with multiple important regions. An procedure update toward using small size pre-samples. The needs only few steps achieve density, shows significant improvement efficiency accuracy variety component system problems. far less samples than both crude Monte Carlo simulation cross-entropy-based employing unimodal function; thus achieving relatively values coefficient variation efficiently. computational are not hampered level, dimension space, curvatures limit-state function. Moreover, distribution parameters densities obtained help identify areas their relative importance.