Modelos de previsão aplicados ao controle de qualidade com dados autocorrelacionados

• Ano de defesa: 2009
• Autor(a) principal: Klidzio, Regiane
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Santa Maria; BR; Engenharia de Produção; UFSM; Programa de Pós-Graduação em Engenharia de Produção
• Palavras-chave em Português: Séries temporais; Modelos lineares e não-lineares; Autocorrelação; Previsão; Gráficos de controle; Time series; Linear e non-linear models; Autocorrelation; Forecast; Control charts; CNPQ::ENGENHARIAS::ENGENHARIA DE PRODUCAO
• Link de acesso: http://repositorio.ufsm.br/handle/1/8128

This research has a topic forecast models applied to industrial productive processes with the objective of verifying the stability of the process through control charts applied to the residues originated from linear and non-linear model. In the presence of autocorrelation data, it was necessary to look for a mathematical model which are produce independent and identically distributed residues. This investigation about the stability of the process goes by the verification of the volatility is influence in the detection of points that are capable to affect the productive process performance. This fact shows the existence of the volatility in productive processes, which it is just used until now in economic variables. The data used for analysis belong to three different industries in different segments. The mathematic models were used multivariate dynamic equation, ARIMA and ARIMA-ARCH model. According to the control charts the statistical techniques used to eliminate the serial autocorrelation was statistically adequate comparing to the classic model used by each industry analyzed. Besides, it was verified, in the period that the volatility occurs corresponds to the period the shows a lack of stability detected by Shewhart control charts. The mathematic models were able to represent the productive process, facilitating understands the behavior of the variables, and help to accomplish the forecast and monitoring the process.