Cluster analysis for mineral processing applications

Most cluster analysis techniques are not well suited to the analysis of engineering data as they have been designed for more qualitative data that tend to produce compact clusters. A clustering technique is described which develops and uses local cluster shapes to build clusters that are suitable for use in engineering applications. The technique is finding application in the analysis of many types of data, particularly where multiple trends and outlying points occur in multiple dimensions. Applications of particular interest are the calibration of on-stream analytical equipment and the analysis of data on plant performance. In both cases the multidimensional nature of the data renders manual analysis very difficult