Metodologias alternativas aos gráficos de controle na caracterização de processos univariados

• Autor(a) principal: Gonçalves, Thiago da Costa
• Data de Publicação: 2008
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: LOCUS Repositório Institucional da UFV
• Texto Completo: http://locus.ufv.br/handle/123456789/4019
• topic: Gráficos de controle; Redes neurais artificiais; Análise multivariada; Control charts; Artificial neural networks; Multivariate analysis; CNPQ::CIENCIAS AGRARIAS

The companies have invested more in quality, to find more survivor in the market, to win more satisfied customers and to increase productivity and profitability. To get it with more success with the aid of statistical control, can be used control charts that monitor the process and indicate whether is need to fix it, so that the final product can be within the standards required by consumers. However, alternative methods to the control charts, also can be used to classify or discriminate if the process is on control or not. In this research, the following were applied: analysis discriminant, logistic regression and artificial neural networks. To apply each one of the proposed methods were used simulated data, where the measures of comparison between them were based on incidences of false and true alarms on the classification these values into or outside the control statistic. Were simulated normal values and independently distributed under control statistics. Subsequently, were imposed some changes to the values set in the middle or the end of data, left the control or to submit autocorrelations. The artificial neural networks and logistic regression were able to replace the best types of control charts in signaled out of control points located in the middle or the end of data set under different distances from the average of control and distributed independently or not.