Evaluation of neural models to estimate the roughness of advanced ceramics in surface grinding
Nenhuma Miniatura disponível
Data
2015-01-01
Autores
Nakai, Mauricio Eiji [UNESP]
Guillardi, Hildo [UNESP]
Aguiar, Paulo R. [UNESP]
Bianchi, Eduardo Carlos [UNESP]
Da Silva, Paulo Sérgio [UNESP]
Título da Revista
ISSN da Revista
Título de Volume
Editor
Resumo
There is an increasing trend for ceramic components to replace metal ones due to their excellent physical, chemical and mechanical properties. However, many of the characteristics that make ceramics so attractive also make them difficult to manufacture by traditional machining methods. The purpose of this study was to develop neural models based on acoustic emission and cutting power signals to estimate the roughness of advanced ceramics during the grinding process. Testing of alumina ceramic specimens was performed on a tangential surface grinder with a diamond wheel. The tests were performed using three cutting depths, 120 μm, 70 μm and 20 μm, a grinding wheel speed of 35 m/s and table speed of 2.3 m/s. Four neural models were studied: multilayer perceptron neural networks, radial basis function neural networks, general regression neural networks and adaptive neuro-fuzzy inference system. To better compare the performance of the neural models used in this study, an algorithm was developed to train all the possible combinations of inputs and parameters of each type of neural network. The results of the best models produced very low error values within the range of accuracy of the measuring instrument. Thus, it can be stated that these models achieved 100% accuracy in estimating workpiece roughness.
Descrição
Palavras-chave
Acoustic emission, Ceramic grinding, Neural network, Surface roughness
Como citar
International Journal of Machining and Machinability of Materials, v. 17, n. 5, p. 454-479, 2015.