MULTIVARIABLE CLUSTER-ANALYSIS FOR HIGH-SPEED INDUSTRIAL-MACHINERY

The overall operation and internal complexity of a particular producti on machinery can be depicted in terms of clusters of multidimensional points which describe the process states, the value in each point dime nsion representing a measured variable from the machinery. The paper d escribes a new cluster analysis technique for use with manufacturing p rocesses, to illustrate how machine behaviour can be categorised and h ow regions of good and poor machine behaviour can be identified. The c luster algorithm presented is the novel mean-tracking algorithm, capab le of locating N-dimensional clusters in a large data space in which a considerable amount of noise is present, Implementation of the algori thm on a real-world high-speed machinery application is described, wit h clusters being formed from machinery data to indicate machinery erro r regions and error-free regions. This analysis is seen to provide a p romising step ahead in the held of multivariable control of manufactur ing systems.