基于MCIS抽样的SRAM失效概率的估计-概率论与数理统计专业论文.docxVIP

I I 基于 MCIS 抽样的 SRAM 失效概率的估计 中文摘要 中文摘要 小概率事件的发生概率很小，但估计其发生概率具有重要的实际意义，它在很多 领域得到推广和应用，比如航天工程、保险、医疗等领域.随着亚微米技术的发展， 估计 SRAM 失效概率变得越来越重要.针对 SRAM 失效这一特定的小概率事件，目前 的研究大多集中于正态分布的情形，本文将其推广到一般分布. 在这篇文章中， 我们首先介绍了小概率事件模拟的相关理论（传统的MC抽样、 重要抽样、Gibbs抽样的原理）.随后，我们介绍了MCIS抽样原理，并将其运用到SRAM 失效概率估计的研究中.最后，基于文中提出的模拟精度及模拟效率的刻画标准，通 过SAS软件模拟，考察了阈值电压V 分别服从多元正态分布、多元 t 分布时的失效概 率估计的情形.计算表明：为了得到可靠估计，MCIS抽样方法所需的模拟次数比传统 抽样方法所需的模拟次数少，这表明我们的方法是有效的. 关键词： 小概率事件，MCIS，SRAM失效概率估计，模拟精度，模拟效率. 作 者： 俞丽丽 指导教师： 汪四水 II II Abstract Efficient SRAM Failure Rate Estimation via Markov chain important sampling Efficient SRAM Failure Rate Estimation via Markov chain important sampling Abstract The occurrence probability of the rare event is extremely small, but rare-event simulation is very important. It has been widely applied in many fields, such as space engineering, insurance, medicine. As the development of submicron technology, SRAM failure rate estimation is of great importance. As to this special rare event, the study has always focused on the normal distribution recently. In this paper, we study the simulation in general distributions. In this paper, firstly, we introduce some theories of rare-event simulation ( na?ve simulation：Monte Carlo、Importance Sampling、Gibbs Sampling). Secondly, we introduce Markov chain Monte Carlo Sampling, and then apply these theories in the SRAM failure rate simulation. Finally, according to the rules of accuracy and efficiency of the results referred in this paper, through some related simulations in SAS, we inspect the results of SRAM failure rate simulation in different distributions( multivariate normal distribution、 multivariate t distribution). All results indicate that to get a reliable estimation, the simulation frequency of MCIS Sampling is much lower than that of MC Sampling. Therefore, our simulation is feasible. Keywords： Rare event, MCIS, SRAM failure rate simulation, Accuracy, Efficiency. Written by Yu Lili Supervised by Associate Prof. Wang Sishui 目 录 第一章 引 言