Data Based Classification of Roaster Bed Stability

摘要

An on-line application of a self-organizing map (SOM) has been developed for detecting and predicting instability of a fluidised bed furnace. In the application, which has been in use at the roster of Outokumpu Kokkola zinc plant for over one year, SOM is utilised for compressing multi-dimensional process measurement data and for visualising bed stability changes as a path on a two-dimensional map. Instability, which normally causes operational problems and lowered production, has been detected quite reliably. A rule-based system for proposing correcting actions is being developed as an extension to the SOM.