Process Analysis Techniques for Cement Kilns

Some useful techniques for the process analysis of cement kilns, related experience, and theoretical considerations are discussed in this paper. In the practical application of process analysis techniques-i.e., autocorrelation, spectral density analysis, crosscorrelation, scatter diagram, and estimation of impulse response by the use of the least-square method-the most important problem is how to treat the data prior to data analysis in order to minimize the effect of noise. The data gathered directly from a kiln contains, in many cases, two kinds of noises-trend and high-frequency noise-and the most useful information lies in the medium-frequency range. In our process analysis system, the high-frequency noise within the range 0.01-60 Hz is eliminated by the use of non-linear analog filters specially developed during the data acquisitions. Remaining high-frequency noise and trend are eliminated by digital filters, i.e., high-pass and low-pass filters, respectively. If these filters and process analysis techniques are applied properly, remarkably clear relationships are obtained among the kiln data, notably: clinker quality, burning-zone temperature, kiln drive torque, secondary air temperature, etc. In the technical theory it is proved that the application of filters does not alter the estimated process characteristics when the same filter is applied to both input and output data of the process to be analyzed. Concerning the estimation of the impulse response, very important considerations are the selection of the data sampling interval, the length of response to be estimated, and feedback loops contained in the process itself.