讲座题目:用于可靠性数据分析的贝叶斯方法
Title:Bayesian Methods for Reliability Data Analysis
时间:2018年12月18日16:00 - 17:00 地点:嘉陵江路校区教学楼A414
摘要:贝叶斯推理具有先进计算能力以及能够处理有限数据的优点,因而被越来越多行业从业者用于可靠性分析。本次报告首先介绍贝叶斯推理的动机并比较其与频率统计推理的优缺点。其次,对包括先验启发式和先验分布、抽样模型、后验推理和仿真方法等贝叶斯推理方法作一个基本而全面的介绍。在介绍各种贝叶斯推理方法的同时,将间歇通过可靠性数据分析实例和使用各种类型可靠性数据(如二进制‘失效/通过’数据、全生命周期数据、退化数据等)的应用实例加于说明。此外,本次报告还使用常用的统计数据分析软件 – 《R》软件演示数值分析案例。本次报告将向大家全面介绍贝叶斯可靠性分析方法,以及如何使用开源统计分析软件 - 《R》软件进行贝叶斯可靠性分析评估。
Abstract:Bayesian inference has been increasingly accepted and used by industry practitioners for reliability analysis due to advanced computational capabilities as well as its advantages of dealing with limited data availability. The motivation of Bayesian inference and its pros and cons versus the frequentist statistical inference are first discussed and reviewed. Then, the talk provides a basic and comprehensive introduction of Bayesian inference methodologies covering topics such as prior elicitation and prior distribution, sampling models, posterior inference and simulation methods. The methodological introduction of Bayesian inference is accompanied with reliability data analysis examples and applications using various types of reliability data, e.g., binary fail/pass data, lifetime data, and degradation data. In addition, numerical examples are also demonstrated through the commonly used statistical data analysis software of R. In summary, this talk presents to the audience a comprehensive understanding to the Bayesian reliability analysis methods, and the software implementation of Bayesian reliability estimation using the open source R software.
主讲人简介
李钊军(史蒂文)现为位于美国春田市(Springfield)的西新英格兰大学工业工程与工程管理系教授。李教授的研究兴趣主要集中在产品设计,系统工程及其在新产品开发、复杂工程系统诊断和预测以及工程管理应用中的可靠性、质量和安全工程。李教授在华盛顿大学获得工业工程博士学位,他是经ASQ认证的可靠性工程师和卡特彼勒Six Sigma Black Belt (SSBB)。李教授曾担任过美国大型工程机械公司-卡特彼勒铁路分部的可靠性团队负责人,负责支持该公司四冲程发动机和燃气-柴油双燃料发动机开发。李教授目前担任《IEEE Transactions on Reliability》副主编,《International Journal of Performability Engineering》主编。他是国际行业协会IISE和IEEE的资深会员,曾在IEEE可靠性协会董事会任职两届,并将于2019年担任IEEE可靠性协会出版副总裁。