Metodologia para ajuste e comparação de modelos em confiabilidade

AUTOR(ES)
DATA DE PUBLICAÇÃO

2004

RESUMO

This investigation presents a methodology for fitting models in reliability and comparative studies involving two or more samples. The first stage is the selection of the most appropriate distribution to determined failure processo This selection is carried through Probability Plots and the Coefficient of Determination. After that, the parameters of the distribution are estimated through Maximum Likelihood Method. For the Exponential distribution, there are direct expressions that supply the estimates for the parameters. For the Normal and Weibull distributions, the equations to obtain the estimates for the parameters are solved numerically using the Newton-Raphson Method. Due to inherent uncertainty to each estimate, confidence intervals approached by the Normal distribution are calculated and the approximated sample size is supplied in function of an interval with a specific magnitude and confidence leveI. With the models completely defined for each sample, hypothesis tests are carried out with the objective to verify the existence af statiscally significant differences and the Likelihood Ratio Test is used. The developed methodology was applied to an example of consulted literature and three real cases had been analyzed

ASSUNTO(S)

estimativa de parametro confiabilidade (probabilidade) verossimilhança (estatistica)

Documentos Relacionados