The problem of estimating unmeasured state and disturbance variables for nonlinear process control applications is considered. We assume that the process model is affine with respect to the unmeasured disturbances and any unmeasured state variables. The disturbances are considered as additional state variables, and a nonlinear observer is designed for the augmented state-space system. The observability of the augmented system is analyzed, and a constructive procedure for calculating the nonlinear observer gains is proposed. Full-order and reduced-order observers are constructed for both the full-state and partial-state feedback cases. Stability results for the nonlinear observer are presented. An output feedback controller is obtained by combining the nonlinear observer with an input-output linearizing controller. The proposed estimation strategy is compared to linear estimation techniques using a fluidized bed reactor model. (C) 1998 Elsevier Science Ltd. All rights reserved.