Proposta de uso de distribuições finitas de probabilidade e distribuição T^2-Hotelling para o monitoramento de processos de múltiplos fluxos

• Autor(a) principal: Vicentin, Damaris Serigatto
• Data de Publicação: 2021
• Tipo de documento: Tese
• Idioma: por
• Título da fonte: Repositório Institucional da UFSCAR
• Texto Completo: https://repositorio.ufscar.br/handle/ufscar/14570
• topic: Mistura de distribuições; Gráficos de controle; Processos de múltiplos fluxos; Gráfico de T^2-Hotelling; Setor Alimentício; Mixture distributions; Control chart; Multiple-stream processes; T^2-Hotelling chart; Alimentary sector; ENGENHARIAS::ENGENHARIA DE PRODUCAO::GERENCIA DE PRODUCAO

A topic still little researched is the monitoring and control of multiple stream processes (Multiple Stream Process - MSP). These are cases in which the equipment has several modules and each module produces the same parts, components and materials with different parameters of means and standard deviations, constituting subpopulations. Within this scope, the objective of this thesis was to propose a finite mixture control chart model for monitoring MSP, which is still rarely found in the literature, and to compare it with the control chart T^2-Hotelling. To achieve the intended objective, was used the quantitative and qualitative approach based on modelling, through the application in an illustrative case with real data collected during the production of pastry dough from a food industry. A comparative analysis was performed regarding the statistical and operational aspects between the finite mixture control charts and T^2-Hotelling. It was observed that the approaches differ in terms of their assumptions, as in their mathematical and operational complexity. In terms of performance, the T^2-Hotelling charts were better at detecting problems (special causes of the process) and easier to apply due to their lower mathematical complexity. On the other hand, the finite mixture graph is innovative, it is applicable, it considers all the flows and the variability between them in a single control graph, it is easier in terms of the sampling procedure, but it brings greater mathematical complexity to its application.