Experimental validation of a full-horizon interval observer applied to hybridoma cell cultures

For process optimisation purposes, it is of interest to monitor the concentration of monoclonal antibody produced by cultures of hybridoma cells. However, the on-line hardware measurement of this concentration is difficult and the design of a software sensor is the most appropriate alternative. As the process model is uncertain, a robust estimation strategy is needed, and in this study, an interval observer based on the principle of full-horizon estimation is developed and evaluated in realistic experimental conditions. The observer is based on a data-driven model and involves the solution of a nonlinear programming problem at each sampling time. The results are satisfactory and appear as a practical solution to the soft sensor development.