A neural network classifier for the die casting industry

• Tipo: Artigo em Livro de Atas de Conferência Internacional
• Data: 2000

Autores

João Cardoso (Autor) FEUP; Nuno Fidalgo (Autor) FEUP; Manuel Matos (Autor) FEUP

Ata de Conferência Internacional

Título: Proceedings of the 11th Portuguese Conference on Pattern Recognition

Páginas: 43-48

11th Portuguese Conference on Pattern Recognition

Porto, Portugal

Classificação Científica

FOS: Ciências da engenharia e tecnologias > Engenharia electrotécnica, electrónica e informática

CORDIS: Ciências Físicas > Ciência de computadores > Design de sistemas > Redes neuronais

Outras Informações

Abstract (EN): This article describes an application of neural networks for on-line classification of defective parts on the die casting industry. The productive process is described with special emphasis on online data collection and treatment. Feature selection is a two step process. The first step uses the sequencial forward search strategy with a KNN classifier as the feature quality measure and the second step uses a neural network to asses the first step selected features relevance. The feedforward neural network has one hidden layer and uses the backpropagation learning rule. The neural network topology is determined using the error rate on a validation dataset. Analysis and results from a particular die cast real data are presented.

Idioma: Inglês

Tipo (Avaliação Docente): Científica

Nº de páginas: 6