Title :
Artificial neural networks for data modelling of a plastic injection moulding process
Author :
Costa, N. ; Ribeiro, B.
Author_Institution :
Centro de Inf. e Sistemas, Coimbra Univ., Portugal
Abstract :
Injection moulding is the most common manufacturing method for the production of high-volume plastic parts. Although theoretically simple, the existence of strong nonlinearities and the unpredictability that is inherent in the raw material used transforms the method into a quite complex process. The development of a process model, based on the analysis of the mathematical relationships between the variables, although possible, would be extremely difficult. Artificial neural networks, due to their capabilities and simplicity, are an attractive approach. Several training algorithms are presented in this paper, from which resilient backpropagation has been used to train a neural network that is capable of modelling the process behaviour. Because of the complexity involved, due to the large number of available variables, principal component analysis has been applied to reduce the dimensionality of the model to be learned by the neural networks
Keywords :
backpropagation; data models; moulding; neural nets; plastics industry; principal component analysis; production engineering computing; artificial neural networks; complexity; data modelling; high-volume plastic parts; manufacturing method; model dimensionality reduction; nonlinearities; plastic injection moulding process; principal component analysis; process behaviour; process model; resilient backpropagation; training algorithms; unpredictability; Artificial neural networks; Backpropagation algorithms; Injection molding; Manufacturing; Mathematical model; Neural networks; Plastics; Principal component analysis; Production; Raw materials;
Conference_Titel :
Neural Information Processing, 1999. Proceedings. ICONIP '99. 6th International Conference on
Conference_Location :
Perth, WA
Print_ISBN :
0-7803-5871-6
DOI :
10.1109/ICONIP.1999.844686