Merging of Advices from Multiple Advisory Systems - With Evaluation on Rolling Mill Data

The problem of evaluation of advisory system quality is studied. Specifically, 18 advisory strategies for operators of a cold rolling mill were designed using different modelling assumptions. Since some assumptions may be more appropriate in different working regimes, we also design a new advising strategy based on the on-line merging of advices. In order to measure actual suitability of the advisory systems, we define two measures: operator’s performance index and coincidence of the observed operator’s actions with the advices. A time-variant model of advisory system suitability is proposed. Merging of the advices is achieved using Bayesian theory of decision-making. Final assessment of the original advisory systems and the new system is performed on data recorded during 6 months of operation of a real rolling mill. This task is complicated by the fact that the operator did not follow any of the recommendations generated by the advisory systems. Validation was thus performed with respect to the proposed measures. It was found that merging of the advising strategies can significantly improve quality of advising. The approach is general enough to be used in many