Application of principal component related variable residual analysis to Imperial Smelting furnace Process monitoring

The principal component related variable residual (PVR) statistic is introduced to industry process monitoring and fault diagnosis in pb-zn smelting process instead of the traditional square prediction error SPE statistic, which confines to Imperial Smelting Process(ISP) as the research background. The PVR statistical is not only able to provide more particular information about the process conditions, but also to overcome the conservation of SPE statistical test. As a result, the reasons to provoke the change of the Hotelling T
²
test between the normal working condition change and the real process fault can be unambiguously identified. Then the practice application study in the production process of ISP has shown the effectiveness of the metallurgical production analysis and monitoring in using of the PVR statistic.