Improvements on multivariable control strategies tested on the Petlyuk distillation column

In this work, a multivariable control methodology, named minimum square deviation (MSD), is applied to the challenging Petlyuk distillation column. This process is able to separate a ternary mixture in the same shell reducing the capital investment cost respect to that needed if a classic layout of two integrated columns is used. By this way, the energy consumption can be reduced up to 30%. However, some problems have been reported, caused by the intrinsic coupling among the control loops implemented on this column. The MSD method selects a control structure systematically by minimizing the sum of square deviations (SSD) and net load evaluation (NLE) indexes. From SSD procedure a diagonal controller structure is achieved. The NLE stage is a valuable tool for deciding which is the best control configuration among full, diagonal or sparse. It depends on the weighting factors which quantify the relative importance between servo and regulator problems. On the other side, a well recognized multivariable control strategy, named self optimizing control (SOC), was previously implemented on the same Petlyuk distillation column. Different controlled variables were selected by applying singular value and null space methods which gave different decentralized control structures. Thus, the Petlyuk distillation column represents a proper example to highlight MSD achievements. Then, the first contribution of this work is to quantify the skills of both methods, MSD and SOC, through the use of the same complex case. The second contribution is to propose some improvements on SOC technique. The third contribution is to obtain new efficient control alternatives for the Petlyuk distillation column. Then, to give support to the final conclusions, a complete set of simulations and quantitative comparisons is given for several control structures and scenarios.

► New multivariable methodology for interacting chemical plants.
► Computer aided tools for supporting the calculations.
► Multivariable tuning method for testing the final control structure.
► Improvements for self optimizing control method.
► Comparisons with other strategies for Petlyuk column.