Just-in-time learning for cement free lime prediction with empirical mode decomposition and database monitoring index

Free lime (f-Cao) content is the key quality indicator in the control and optimization of cement calcination process. However, the cement process has many complex characteristics including large noise, time delay, nonlinearity, and time-varying system. To cope with the above characteristics, this study proposes just-in-time (JIT) learning with empirical mode decomposition (EMD) and database monitoring index (DMI) for free lime prediction. At first, EMD and Savitzky-Golay (S-G) filter can effectively deal with process large noise. Meanwhile, weighted time series analysis and mutual information are used to estimate large time delay. Then, least squares support vector machine (LSSVM) based on JIT and DMI is employed to solve the problems of process nonlinearity and time variation. The proposed strategy is validated on practical data from a cement plant.