Abstract: In industrial process, data acquisition is very important for process control and optimization online. However, the measurement data are inevitably affected by the errors. And what's more, measurement data will be inaccurate for gross error by instrument fault, pipe leak and other abnormal situations. Data reconciliation is one of the available technologies to ensure the reliability and accuracy of data. Kalman filter has been widely used in traditional dynamic data reconciliation. However, when gross error existed in the measurements, the reliability of reconciliation could not be guaranteed by only using Kalman filter. In order to improve the accuracy and reliability of reconciled values, especially when gross errors exist in measurement data, a novel gross error detection method based on dynamic Bayesian model is proposed to detect the gross error in time based on the traditional Kalman filter. Firstly, measurement data are filtered by traditional Kalman filter, and the filtered measurement data in the current sampling time are standardized, then probability distribution model is built by extended dynamic Bayesian network. The filtered measurement data are used in the probability distribution model to calculate the conditional probability of the measurement with gross error and normal condition. Then gross error is detected by the result of conditional probability calculation. According to the result of detection, the measurement variance matrix and Kalman filter gain and related parameters can be modified correspondingly. Thus, the accuracy of reconciled value can be improved. And the simulation of water and benzene temperature change in jacketed type exchanger is used as an example. Comparing the novel detection method based on dynamic Bayesian with the traditional Kalman filter method, the simulation results show that the proposed method can efficiently detect gross error and improve the reconciled value accuracy through adjusting the parameters timely when gross error exists in the measurement data of dynamic system.