Você está aqui: Página Inicial / Seminários / 06/11/2019 - A Hierarchical Model for Heterogeneous Reliabilty Field Data - Palestrante: William Q. Meeker (Iowa State University)

06/11/2019 - A Hierarchical Model for Heterogeneous Reliabilty Field Data - Palestrante: William Q. Meeker (Iowa State University)

Quando 06/11/2019
das 15h30 até 17h30
Adicionar evento ao calendário vCal
iCal

Seminário Conjunto UFSCAR/USP

Data e Horário:
06/11/2019 às 16h30

Local:
Sala 4-111 do ICMC-USP ( auditório Luiz Antonio Favaro)

Título:
A Hierarchical Model for Heterogeneous Reliabilty Field Data

Palestrante:
William Q. Meeker
Iowa State University

Resumo:
When analyzing field data on consumer products, model-based approaches to inference require a model with sufficient flexibility to account for multiple kinds of failures. The causes of failure, while not interesting to the consumer per se, can lead to various observed lifetime distributions. Because of this, standard lifetime models, such as using a single Weibull or lognormal distribution, may be inadequate. Usually cause-of-failure information will not be available to the consumer and thus traditional competing risk analyses cannot be performed. Furthermore, when the information carried by lifetime data are limited by sample size, censoring, and truncation, estimates can be unstable and suffer from imprecision. These limitations are typical; for example, lifetime data for high-reliability products will naturally tend to be right-censored.

In this paper we present a method for joint estimation of multiple lifetime distributions based on the Generalized Limited Failure Population (GLFP) model. This 5-parameter model for lifetime data accommodates lifetime distributions with multiple failure modes: early failures (sometimes referred to in the literature as ``infant mortality'') and failures due to wearout. We fit the GLFP model to a heterogeneous population of devices using a hierarchical modeling approach. Borrowing strength across sub-populations, our method enables estimation with uncertainty of lifetime distributions even in cases where the number of model parameters is larger than the number of observed failures. Moreover, using this Bayesian method, comparison of different product brands across the heterogeneous population is straightforward because estimation of arbitrary functionals is easy using draws from the joint posterior distribution of the model parameters. Potential applications include assessment and comparison of reliability to inform purchasing decisions.

This is joint work with Eric Mittman (Intel) and Colin Lewis-Beck (University of Iowa).



registrado em: