Adaptive and high gain nonlinear observers are used for state and parameter estimation in an industrial polymerization process. The solution homopolymerization of acrylic acid is considered. The process is monitored online by near infrared spectroscopy giving the residual amount of monomer. Using this measurement, a continuous adaptive observer is first constructed to estimate the concentration of radicals. A second high gain continuous-discrete observer uses off-line measurements of the polymer molecular weight to estimate the termination rate coefficient that might be time varying due to the gel effect. The transfer to solvent rate coefficient is estimated under steady-state conditions and is assumed to be constant during the reaction. The identified model is validated by varying the process parameters: the concentrations of solvent, monomer, radicals and the residence time in the reactor with interest of their impact on the polymer molecular weight. (C) 2007 Elsevier B.V. All rights reserved.