[ABE-L] Convite Ciclo de Seminários DEST - Palestra Adiada

[Andressa Siroky]

O Ciclo de Seminários do Departamento de Estatística (DEST) da UFRN apresenta uma série de 3 seminários ministrada pelo professor Geoff Vining do Departamento de Estatística da Virgínia Tech.

O seminário que ocorreria dia 27 de junho ocorerrá em 05 de julho de 2022, às 15 horas, no auditódio do CCET da UFRN. O último seminário desta série ocorrerá dia 11 de julho. Todos serão transmitidos ao vivo e ficarão disponíveis no canal do YouTube do DEST UFRN

 https://youtube.com/channel/UCr_8R_aiS69hroG4-Yvzv4Q

Resumo:

This seminar series consists of three approximately one hour lectures in English introducing Weibull regression and its application within reliability studies.  Weibull regression is extremely important for accelerated life tests where the assumed failure mechanism is “failure due to the weakest link.”  The final lecture summarizes an actual NASA study that studied the expected failure rates of composite overwrapped pressure vessels very commonly used on spacecraft.



The first lecture provides the origins and the foundations for the standard two-parameter Weibull distribution.  It derives the distribution and its various forms, in particular the hazard function, which is extremely important for explain the risk of failure for the system over time.  It then presents a high level overview of estimation and inference of the important parameters.  It then introduces the concept of censoring.



The second lecture covers two major topics.  The first is the power law model critical for accelerated life tests.  Many systems have quite long expected lifetimes.  As a result, engineers often “stress” the system to produce failures in a much shorter period of time.  The power law is the most common empirical approach to such testing.  The second major topic is proper residual plots for assessing the assumption of the Weibull distribution.  Some of this work appears to be novel.



The third lecture summarizes the NASA study.  It highlights the importance of the proper analysis of the data, especially when dealing with engineers who do not understand proper data analysis.  How does the data analyst deal with an engineer who has planned the perfect experiment to test everyone of her/his theories, and the data contradict all of them?



Traditionally, mathematical statisticians tend to dominate the field of statistical analysis of reliability data.  These people often have little appreciation for linear models and their analyses.  Power law models ultimately are linear models that require likelihood based estimation and inference.  This series of lectures illustrates how to combine the mathematical statistics and the linear models theory to enhance the understanding of the experimental results.

Atenciosamente,
Andressa N Siroky
Professora Adjunta do DEST-UFRN
-------------- Pr?xima Parte ----------
Um anexo em HTML foi limpo...
URL: <http://lists.ime.usp.br/pipermail/abe/attachments/20220625/91627278/attachment-0003.htm>