Aplicación de Redes Neuronales en la Detección de Regímenes Degradados en el Proceso Wedm

[EN] This article presents the results of a comparative study performed to select the most appropriate neural network configuration for Wire Electrical Discharge Machining (WEDM). The main objective is to detect instability trends that allow alerting to the increasing risk of wire breakage of the cu...

ver descrição completa

Detalhes bibliográficos
Autores: Portillo, E., Cabanes, I., Marcos, M., Zubizarreta, A.
Tipo de documento: artigo
Data de publicação:2009
País:España
Recursos:Universitat Politècnica de València (UPV)
Repositório:RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Idioma:espanhol
OAI Identifier:oai:riunet.upv.es:10251/145374
Acesso em linha:https://riunet.upv.es/handle/10251/145374
Access Level:Acceso aberto
Palavra-chave:WEDM
Electroerosión
RNA
Redes neuronales artificiales
Perceptrón Multicapa
Elman
Electro-discharge machining
ANN
Artificial Neural Network
Perceptron Multilayer
Descrição
Resumo:[EN] This article presents the results of a comparative study performed to select the most appropriate neural network configuration for Wire Electrical Discharge Machining (WEDM). The main objective is to detect instability trends that allow alerting to the increasing risk of wire breakage of the cutting tool: the wire. The wire breakage reduces the process productivity and the required accuracy. Considering the results of previous works of the authors, in which different types of degraded behaviors were identified, a comparative study that considers different aspects has been performed. Among them, the evaluation of classic neural architectures stands out, in particular, the static architecture MultiLayer Perceptron (MLP), and the recurrent architecture Elman. The main conclusion of this work is that the Elman architecture is the most approppriate for detecting the degradation of the cutting process.