Nonlinear adaptive temperature control of multi-product, semi-batch polymerization reactors

Nonlinear Multiple Model Predictive Control of Solution Polymerization of Methyl Methacrylate

“…In their study, they compared their neural network control with a classic PID controller. Clarke-Pringle and MacGregor [8] …”

A Model-Based Control Scheme for Chemical Batch Reactors

“…In order to avoid numerical differentiation of the temperature measurements, an observer can be used to estimate both the heat released by the reaction and the heat-transfer coefficient (see, e.g., [9], [10]). In [9], a nonlinear adaptive control strategy is adopted, based on an extended Kalmann filter to achieve on-line estimation of the time varying parameters involved in the control law; however, convergence and robustness of the overall scheme are not theoretically proven. In [10] the estimation law suffers from singularities; moreover, the dynamics of the mass balance in the reactor is not taken into account, since the heat released by the reaction is estimated as an unknown constant parameter.…”

“…Therefore, adaptive and robust control strategies have to be considered. To this purpose, a few adaptive control strategies have been proposed in the literature: in [8] a nonlinear controller, designed via a differential geometric approach, is augmented with an indirect parameters estimation algorithm, in [9] an extended Kalmann filter is adopted and, more recently, in [10] a GMC control scheme is designed, where the estimation of some unknown quantities, -namely, the heat released by the reaction and the heat-transfer coefficient-are estimated by adopting the nonlinear adaptive observer proposed in [11].…”

Adaptive Feedforward Control with Parameter Estimation for the Chylla-Haase Polymerization Reactor

“…Furthermore, the values of σ 2 ( q 1 ) and σ 2 ( T out j ) are smaller during the monomer feed interval in order to accurately estimate the significant decrease of the heat transfer coefficient U(t) via q 1 (t) in (14). 4 The covariance matrix of the measurement noise is set to diag(σ 2 (y 1 ), σ 2 (y 2 )), whereby noise is added to the measurements y = [T, T T with the standard deviation σ(y 1 ) = σ(y 2 ) = 0.02 K. The sampling time of the EKF is set to ∆t EKF = 4 s. Fig. 4 shows the estimated profiles of m M (t), m P (t), Q rea (t), and U (t) for polymer A compared to the nominal values (--).…”

“…An optimal cooling jacket temperature is calculated in [7] by minimizing a cost function, which keeps the reactor temperature constant in the nominal case. A nonlinear adaptive controller is designed in [4] to adjust the cooling jacket temperature as the setpoint for an underlying PI controller of the cooling jacket. A further contribution [1] deals with the design of a neural network controller to maintain the reactor temperature at its setpoint.…”