Publicação

Predicting product quality from operating conditions based on multinomial logistic regression

Ver documento

Detalhes bibliográficos
Resumo:The main objective of this study is to evaluate the influence of several industrial operating conditions in the occurrence of the defect (air bubbles) in a wet bonding process of two pieces (a glass and an electronic display). The study of the influence of operating conditions in business targets is of great interest to the community of process engineers, to ensure the long-term stability of the process, as this involves complex factors. It is also relevant for the development of Data-driven approaches to support Real-Time System Monitoring. In this work, we propose to build a multinomial logistic regression model that can assess the probability of a bonded kit (glass and electronic display) having a defect in the filling material, taking into account process operating conditions. The study was conducted in a retrospective evaluation and consisted of 659 bonded kits randomly selected from a reference period. We evaluated multiple factors in the construction of a multinomial logistic regression model with 3 logit functions. The technique used to select variables to be included in the model, was the stepwise technique by choosing the smallest p-value for the variable entering in the model. Of the modelling process through multinomial logistic regression have resulted five operating conditions that have statistical significance and that can contribute as an auxiliary for the prediction of a defect.
Autores principais:Vaz, Pedro
Outros Autores:Braga, A. C.; Carvalho, Maria Sameiro; Menezes, Gerardo
Assunto:Logit Multinomial logistic regression Process improvement Process monitoring
Ano:2021
País:Portugal
Tipo de documento:comunicação em conferência
Tipo de acesso:acesso restrito
Instituição associada:Universidade do Minho
Idioma:inglês
Origem:RepositóriUM - Universidade do Minho

Registos relacionados

Registos relacionados