An inversion approach to the estimation of reaction rates in chemical reactors

The estimation of reaction rates is an important problem in mechanistic modeling, monitoring and control of chemical reactors. In contrast to standard estimation techniques where a model must be chosen for the reaction rates, we consider them in this work as unknown inputs. In this regard, the task is an ill-posed inverse problem, i.e. its solution is not unique and/or unstable with respect to data perturbations. Our estimation scheme is based on regularization which eliminates the difficulties arising with ill-posed problems. We investigate the problem in both the time and frequency domains. Guidelines for the design of the estimator structure and the selection of the regularization parameters are presented.