Nonlinear Models Provide Better Control of Annealed Brass Strip Microstructure

Abstract Research conducted at a brass-strip plant in Zutphen, Netherlands shows how nonlinear modeling can be used to predict the grain size and hardness of annealed strip based on composition and process variables. Production data gathered over the course of a year were used to train a feed-forward neural network that achieved a correlation coefficient of more than 86% when presented with new input data. The standard deviation between predicted and measured grain size was about 4.2 µm.