NONLINEAR ESTIMATION METHODS FOR TUBULAR REACTORS

The problem of nonlinear recursive state and parameter estimation is studied for tubular reactors. A realistic model of a packed bed reactor is developed, and the method of orthogonal collocation is used to convert the set of partial differential equations to a larger set of ordinary differential equations. A discrete time state space model is developed by linearization and discretization in time.;Since an accurate knowledge of parameters and states is essential for reactor control and optimization the estimation problems are studied. The linear state space model is used to determine the optimal placement for the measurements in the reactor, and to design a nonlinear constant gain filter for state and disturbance estimation. States and a disturbance are estimated simultaneously by modeling the disturbance as a random walk process, augmenting the reactor model with the model of the disturbance, and solving the resulting nonlinear estimation problem. Modeling the disturbance as a random walk process allows the use of a constant gain filter, and it overcomes filter divergence problems encountered in previous works.;A pseudo static model is developed in order to design nonlinear constant gain filters to estimate slowly changing reactor parameters. By modeling the parameters as a random walk process convergence of the estimator is enhanced. Simultaneous estimation of catalyst activity and heat transfer coefficient proved to be difficult with temperature measurements only. Inclusion of a concentration measurement considerably improves the estimator performance.;The problem of the sensitivity of a pseudo static estimator to suboptimal filter gains is studied, and an adaptation scheme is employed to determine the optimal whitening gain for a nonlinear pseudo static estimator.