Neural networks for the dimensional control of molded parts based on a reverse process model

This paper presents the application of neural networks in suggesting the ch ange of molding parameters for improving the dimensional quality of molded parts based on the concept of reverse process modeling. Instead of using th e molding condition parameters as input values and dimensional outcomes as output values, the reverse process model configures the dimensional outcome s as inputs and the molding condition parameters as outputs. With the mappi ng on input and output layers of neural networks based on this configuratio n, the trained neural networks learn the correlation between the dimensiona l outcome values and the corresponding molding parameters. This model, whic h serves to learn from sample data and induce the values for change of the operating molding conditions, has been implemented for the dimensional impr ovement of injection molding parts, the dimensions of which are primarily d etermined by the process parameters such as injection time and cooling temp erature. (C) 2001 Elsevier Science B.V. All rights reserved.