Abstract: A soft sensor model based on the simplified nonlinear dynamics equation with device factor is used to predict the 4-CBA concentration,which is an important quality index in PX oxidation processes. The parameters of the model are estimated using the unscented Kalman filtering method. Compared with the traditional nonlinear least squares method,the proposed method does not require linearization on calculation of the Jacobian matrix. Industrial data are applied to test three soft sensors for 4-CBA concentration prediction. The results show that the proposed method is effective.