In multirate chemical processes, soft sensor is commonly used for fast-rate quality estimations. As the kernel of soft sensor, modeling technique has drawn much attention from researchers. But most of the modeling methods have disadvantages of uncertain estimation errors and limited modeling capability. To these issues, soft sensor calibration is considered as an efficient alternative, which however is often ignored by researchers and the current application level is relatively low. A novel soft sensor calibration method is proposed in this paper. Under multirate sampling conditions, data fusion technology based on Kalman filter is introduced into soft sensor maintenance to integrate the soft sensor model estimations with process measurements. The performance of the algorithm is evaluated through simulations and a laboratory scale experiment. The factors that may influence the performance are also discussed in detail. The results demonstrate that the multirate Kalman filter approach is able to improve the accuracy and reliability of soft estimations, when the essential dynamics is included in the Kalman filtering model and the filter parameters are properly tuned. (C) 2010 Elsevier Ltd. All rights reserved.