科学研究

Seminar 首页 - 科学研究 - 学术活动 - Seminar

【Seminar】Modelling and analyzing d ependent failure behaviors using stochastic hybrid automata

Speaker


A/Prof.Zeng Zhiguo (曾志国 教授) 
巴黎中央理工大学


Host


Prof. Guan Xuefei  (关雪飞 研究员)


 

Date&Time


 July 17, 2019 (Wednesday)    10:00~11:00


Venue


C405 Conference Room, 4rd Floor


Abstract 



On the evening of 31 May 2009, flight AF447 took off as usual in the beautiful sunset of Rio de Janeiro, supposed to provide its 216 passengers with a nice and comfortable journal to a city of love and romance, Paris – But the love and romance never arrive. Shortly after taking off, the plane crashed into the Atlantic, killing all the 228 people (including 12 crew members) onboard. The tragedy shocked the entire world, as modern aircrafts like AF447 have been designed with sufficient numbers of safety systems in a “defensive-in-depth” manner: such accidents could occur only when all the safety systems fail, which is considered highly unlikely under the commonly-used engineering assumption that the safety systems are independent from each other.


What happened to this doomed flight, then? Can we learn something from it to better understand and characterize the failure behaviors of modern engineering systems? In this talk, we start from a brief analysis of the accident, through which we show that the “Achilles’ Heel” of modern highly complicated engineering systems is the dependent failure behaviors that are often neglected in the analysis. In the following parts of this talk, then, we discuss some of our recent works on modelling and analyzing dependent failure behaviors. A new framework, based on stochastic hybrid automata, is developed for modelling and analyzing dependent failure behaviors. To reduce the computational burdens in the analysis, we also develop a semi-analytical method for reliability analysis of dependent failure processes. The semi-analytical method is also applied to describe dependent failure behaviors in the system’s level, i.e., the common cause failures. Finally, based on the stochastic hybrid automata model, we develop online reliability assessment and remaining useful life prediction methods for dependent failure processes. Applications on different engineering systems are performed to demonstrate the developed methods. The main results presented in this talk can provide more insights on how to better understand and characterize dependent failure behaviors in modern engineering systems.



About the Speaker


曾志国,198953日出生。2011年、2016年分别于北京航空航天大学获得工学学士、博士学位。2017年至今在巴黎中央理工大学担任助理教授。他的研究方向包括不确定性量化与分析方法、确信可靠性理论、动态风险分析、工业工程连续性建模与分析方法等。他在各类期刊与国际会议上发表了30余篇论文(其中,SCI检索19篇)。


All interested are welcome!