A Systematic Method for Optimum Sensor Selection in Inferential Control Systems

This paper considers the optimal selection of sensor locations in square inferential control systems, where the number of measurements employed is equal to the number of actuators. A mixed-integer linear programming (MILP) problem formulation is proposed based on the assumption that the control objectives can be related to the variability of certain process variables. The method is applied to three case studies including a methanol−water column, a deisobutanizer column, and a benzene, toluene, o-xylene column. The selection of tray temperatures for inferential composition control is the problem considered in these case studies. To investigate the validity of the proposed method, closed-loop simulations are carried out. The method is shown to provide a satisfactory answer to the optimum sensor selection problem.