Distillation Siso control system design and analysis using internal composition measurements

Distillation control continues to be an area of intense research activity within the control world. While many researchers focus on developing more complex control algorithms, most industrial distillation columns continue to use single-input, single-output (SISO) control loops. Traditionally, industry has used internal temperature sensors and SISO loops to inferentially control the column end-product qualities. However, recent advances in analyzer development are making possible fast, accurate, on-line composition measurements from within distillation columns. If these technological advances are to be exploited, developing a clearer understanding of the possible roles for these measurements in the control of distillation columns is crucial. The current tools, methodologies, and understanding for selecting and locating SISO control sensors need to be extended. This work addresses the potential use, design implications, and benefits provided by internal composition measurements in the control of distillation columns. Singular Value Decomposition (SVD), Closed-Loop Disturbance Vectors (CLDVs), and other analytical tools are used to design control schemes for the distillation columns in this work. The current SVD sensor selection procedure is extended to consider the possibility of mixed sensors, i.e., using both temperature and composition sensors. Also, a sensor screening technique that estimates the column closed-loop profiles during disturbance rejection is introduced. These tools can help identify (1) the strengths and weaknesses of a control strategy, (2) the most appropriate sensor types and locations for SISO distillation control, and (3) the anticipated propagation of variation within the column during closed-loop disturbance rejection. Lastly, this work reports an experimental application of an internal composition analyzer that uses Raman spectroscopy. This implementation served as the impetus for the aforementioned sensor selection and location studies.