Process measurements are often corrupted with varying degrees of noise. Measurement noise undermines the performance of process monitoring and control systems. To reduce the impact of measurement noise, exponentially-weighted moving average and moving average filters are commonly used. These filters have good performance for processes under steady state or with slow dynamics. For processes with significant dynamics, more sophisticated filters, such as model-based filters, have to be used. The Kalman filter is a well known model-based filter that has been widely used in the aerospace industry. This paper discusses another model-based filter. the dynamic data reconciliation (DDR) filter. Both the Kalman and the DDR filters adhere to the same basic principle of using information from both measurements and models to provide a more reliable representation of the Current state of the process. However, the DDR filter can more easily incorporated in a wide variety of model structures and is easier to understand and implement. Simulation results for a binary distillation column with four controlled variables showed that the DDR filters had equivalent performance to the Kalman filter in dealing with both white and autocorrelated noise. (c) 2005 Elsevier Ltd. All rights reserved.