A methodology for on-line process identification is developed for use the chemical process industry.The measurements of process output that arrive at different sampling rates are incorporated into the identification strategy based on nonlinear programming techniques.Measurement delays associated with some of process output and constraints on identified variables are also implicitly included in the formulation．A probabilistic interpretation shows that maximum likelihood estimation is identical with taht by minimizing the least squares objective function .The evaluation of the proposed multirate state and parameter identificator through simulations and experimental application on a wall-cooled fixed-bed reactorgives atisfactory performance and illustrates the practicality of this approach.