Process monitoring and modeling using the self-organizing map

The Self-Organizing Map (SOM) is a powerful neural network method for analy sis and visualization of high-dimensional data. It maps nonlinear statistic al dependencies between high-dimensional measurement data into simple geome tric relationships on a usually two-dimensional grid. The mapping roughly p reserves the most important topological and metric relationships of the ori ginal data elements and, thus, inherently clusters the data. The need for v isualization and clustering occurs, for instance, in the analysis of variou s engineering problems. In this paper, the SOM has been applied in monitori ng and modeling of complex industrial processes. Case studies, including pu lp process, steel production, and paper industry are described.