Object-oriented methodology has characteristics such as data ion and inheritance and is widely used for analysis and modeling of complex systems and problems. Therefore, object-oriented viewpoint was applied on data reconciliation problem and complex data reconciliation problems were divided to simpler ones. Data reconciliation as an object has attributes such as measurements and reconciled values and can be a basis of A data reconciliation software tool, namely DCON, is also developed based on developed In order to show the performance of artificial neural networks for nonlinear dynamic data reconciliation, a new method, namely NetDDR, similar to the NARMA-L2 system identification method based on neural networks is developed. The network of this method is trained using true and noisy values obtained by process simulation. Then the trained network is used for dynamic data reconciliation. NetDDR method was tested by two illustrative examples of simulation of a two-component distillation column and a chemical reactor. In each time step Gaussian white noises are added to the true values obtained by simulation in order to simulate real measurements. Then noisy data are sent to DCON software for performing data reconciliation. In performed cases, reconciled values properly fit true values and the variance of errors of reconciled values becomes very small compared to the variance of measurement errors. Extended Kalman Filtering (EKF) method is also used for dynamic data reconciliation and comparison of its results with results of NetDDR method. Comparison shows that NetDDR method produces better results than EKF method while it doesn’t need any information about the state variables and Jacobian matrices of state and measurement variables. This method is also faster that EKF method and it can also be used for on-line applications as EKF method. NetDDR method has many tuning parameters as EKF method while tuning of its parameters is much easier than those of EKF method.
