An experimental investigation of modeling, control, and optimization techniques for batch distillation

Problems in the modeling, optimization, and control of a binary batch distillation were studied with the goal of providing appropriate approaches to industrial batch distillation optimization. Due to a paucity of experimental results in the literature, a batch distillation column was tested to validate models and optimization techniques. This column was a ten sieve-tray, six inch diameter pilot scale apparatus with a reboiler charge capacity of approximately 80 liters. The column was extensively instrumented, with temperature, flow, pressure, and level sensors. A Honeywell TDC 3000 Distributed Control System (DCS), interfaced to a VAXStation 3200 computer running Setpoint, Inc. SETCON software was used for closed loop control and data acquisition. Accurate continuous analyzers of novel design were installed to measure the overhead and reboiler compositions.; A number of batch distillation experiments were performed using an ethanol-water mixture. The initial charges for these experiments were all approximately 75 kilograms comprised of 20-25 percent ethanol by weight. Ideal stage and tray efficiency models were analyzed and compared to experimental results. Model parameters relating to tray holdup and to mass transfer efficiency were estimated for the time-varying system. Ideal stage models (e.g., BATCHFRAC) were shown to be inadequate for the accurate modeling and optimization of the distillation column over the entire length of the batch. Model parameters and run conditions (e.g., tray holdup and distillate flow rate) regressed from data using such models were inconsistent with measurements and estimates made from physical arguments. A model incorporating stage efficiencies and tray hydraulics was required to adequately model and fit column behavior. Using these results, a new and efficient model was developed that adequately describes both long and short term behavior of the distillation column.; The dynamic response of a batch still varies over the duration of a batch, due to changes in flow rates and composition. The variability of the still was analyzed, and a control system strategy was proposed that eliminates interaction and noise problems that often plague industrial batch stills. Gain scheduling was implemented in simulation studies and shown to be effective in obtaining a constant composition trajectory, with control performance superior to conventional (PID, ratio) control.; Arguments using sensitivity analysis show that for optimization of batch column operation, a model of higher fidelity is required than that needed for conventional continuous control analysis. Optimization results were shown to be sensitive to tray efficiency parameters, and insensitive to the tray holdup parameter. On-line optimization was not justified for the experimental column.