Two time delay compensation techniques, the Smith predictor and the analytical predictor, are used for bottom composition control of a pilot scale methanol-water distillation column. The closed-loop performance of the two predictor schemes is compared to that for a proportional-integral controller in experimental and simulation studies. The predictors resulted in improved control for both set point and feed flow disturbances.
SCOPE
Significant time delays can occur in processes due to the presence of recycle loops, distance-velocity lags in fluid flow, and the dead time inherent in many composition analyses. The detrimental effects of time delays on closedloop stability and control system performance are wideIy recognized. Thus, there is considerable motivation for the development of time delay compensation techniques that provide improved control of systems with significant time delays.
This investigation provides an experimental evaluation of two time delay compensation techniques, the Smith predictor (Smith, 1957(Smith, , 1959) ) and the analytical predictor (Moore, 1969; Moore et al., 1970). Both techniques employ a simple dynamic model to predict future outputs based on current information. These time delay compensation techniques and a conventional proportional-integral (PI) con-trolIer are used to control the bottom composition of a pilot scale, methanol-water, distillation column. The same techniques have been evaluated for top composition control in a related study (Meyer et al., 1977).
This investigation and the related study (Meyer et al., 1977) provide the first experimental applications of time delay Compensation techniques to distillation column control that have been reported in the open literature.
CONCLUSIONS AND SIGNIFICANCE
In general, the analytical and Smith predictor control schemes performed better than a PI controller in controlling the bottom composition of the pilot scale column. The simulation study demonstrated that both time delay compensation techniques can provide significant improvements in regulatory and servo control. Use of the analytical predictor resulted in shorter settling times and lower integral which, in turn, outperformed the PI controller. The ex-