Soft sensor for distillation column feeds

Abstract

Distillation columns are normally controlled at a pre-designed condition, which is the optimal for a specified feed condition. In reality, however, the feeds may vary in their flow rates, temperatures, or compositions, which results in the distillation process operating at non-optimal values. Therefore, it is ideal to optimally adjust the operations in conjunction with the real feed variation. The online hardware measurement of the feed status, however, is basically impossible for most columns. A method using the temperature measurements along the distillation column to detect the feed variations is proposed in this paper. Verification is conducted through simulations under different conditions. Both derivative-based and derivative-free optimization methods are discussed for the soft-sensor establishment.