Optimal Sorting of Raw Materials, Based on the Predicted End-Product Quality

Abstract

Raw material variation is one of the most important factors causing unstable end-product quality. A methodology for sorting raw materials into homogenous groups with constant and optimized processing within each group is presented. The sorting criterion is based on the squared distance between the predicted response and its target value. The raw materials are split into homogenous categories by a partitioning algorithm related to the fuzzy-c-means algorithm. The method has been tested for raw material properties in one and two dimensions and with different degrees of fuzziness. The method shows good flexibility and can also be used with a penalty function penalizing unfavorable process settings.

Keywords:

• Sorting
• Robust process optimization
• Raw material quality
• End-product quality
• Fuzzy clustering
• Penalty function

Acknowledgment

We would like to thank E. M. Færgestad for kindly providing the data used in the example.