Tidsskrift: Journal of Chemometrics, vol. 21, p. 440–450, 2007
Open Access: none
This paper deals with the optimisation of production processes in a situation where the raw material varies. The objective is to minimise production costs for a given raw material quality, with restrictions on the end-product quality. Three important issues are addressed. First, we discuss the collection of data for response surface modelling. In full-scale production processes it is very difficult to perform large experiments. We show that merging smaller designed experiments can be a good alternative. Secondly, we illustrate the use of an approach for combining data obtained using an experimental design with spectroscopic raw material measurements in a joint regression model. Finally we discuss some issues related robustness and sensitivity in optimisation. We also show how this optimisation can be used as a tool when purchasing raw materials.