Neural modeling and control of a 13C isotope separation process

The paper presents a solution for the ¹³C isotope concentration control inside and at the output of a separation column, solution based on the Internal Model Control strategy. The ¹³C isotope results from a chemical exchange process carbon dioxide - carbamate, which is a distributed parameter process. In order to model the mentioned process, an original form of the approximating analytical solution which describes the process work in transitory regime is determined. The evolution of the approximating solution depends both on time and on the position from the column height. The reference model of the fixed part of the control structure is implemented using neural networks, representing an original solution due to the fact that a neural model is determined for a distributed parameter process. The controller is, also, implemented using neural networks, its main parameter being adapted in relation to the transducer position change in the separation column. The advantages of using the proposed concentration control strategy consist of: the possibility of controlling the value of the ¹³C isotope concentration in any point from the separation column height; the improvement of the system performance regarding the settling time; the possibility to reject the effect of the disturbances.