Using cluster analysis techniques to improve machine reliability

The requirement for continuous manufacturing on the shop floor is now a reality for many industries in order to minimise costs and remain competitive. Therefore, the need for systems capable of utilising a large range of sensor inputs for unsupervised control of complex production processes, whilst maintaining a high level of reliability has become increasingly acute. This paper considers one such system under development at the Department of Cybernetics at the University of Reading. The application which provides data and the test-bed for the methods developed is a shop floor production machine running at very high speeds. The main thrust of the work is in the use of sensor and control information from this production process in order to improve reliability and quality of maintenance on the machines. Cluster analysis techniques which have been developed at Reading are used to implicitly model the process and analyse the data