Avaliação da qualidade do processo de lingotamento contínuo na presença de correlação cruzada

• Autor(a) principal: Mezzomo, Meire
• Data de Publicação: 2013
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Manancial - Repositório Digital da UFSM
• dARK ID: ark:/26339/001300000h6fz
• Texto Completo: http://repositorio.ufsm.br/handle/1/8290
• topic: Lingotamento contínuo; Correlação cruzada; Modelos ARIMA; Gráfico de controle multivariado T2 de Hotelling; Componentes principais; Continuous casting; Cross-correlation; ARIMA models; Principal components; CNPQ::ENGENHARIAS::ENGENHARIA DE PRODUCAO

In the current competitive market, a great part of companies has as the main goal the search for continuous improvement of their products and services. Therefore, the application of statistical methods has great relevance in the quality evaluation, helping in the understanding and monitoring of the processes. In such context, the present study concerns to the use of multivariate control charts in the evaluation of the productive processes in the presence of cross-correlation, which the objective is to verify the continuous casting process stability in the production of still billets by means of Hotelling's T2 multivariate control charts applied in the estimated residual mathematical linear models. Initially, the existence of data autocorrelation was verified, it is necessary the ARIMA modeling, because when it happens, it is necessary to determine the residues and apply multivariate control charts to the residues and not on the original variables. The existence of correlation showed to be meaningful among the variables, being one of the assumptions for the statistical application T2. When the T2 chart instability is verified, it was necessary to identify the variable or the set of variables of steel temperatures in the distributor and in the distributor weight, which are responsible for the instability. Later, the estimated residues were decomposed into principal components, and with the help of the correlation of the original variables and the principal components, the variables which most contributed to the formation of each component were identified. Therefore, it was possible to detect the variables which caused the system instability, once for the steel temperature in the distributor were the T4 and T5, followed by T6, T3, T7 and T2 and for the weight of the distributor, PD4, PD5, PD3, PD6 and PD2, respectively. This way, the estimated residues from the mathematical models, the use of multivariate chart control Hotelling's T2 and the decomposition into principal components which were able to represent the productive process. This methodology allowed the understanding of the behavior of the variables and helped the monitoring of this process, as well as, in the determination of the possible variables which caused the instability in the continuous casting process.