Research On Data Rectification And Its Application In Engineering

Data rectification technique is one of the key elements of CMS (Computer Integrated Manufacture System) in process industry. By using the redundancy in measurements and process models, it can eliminate measurements with gross error (MGEs), reduce random measurement errors and estimate unmeasured variables. In this thesis, the problem encountered in industrial applications of data rectification was investigated in the view of engineering. The main works are as follows:1. Literatures on data reconciliation, gross error detection and sensor network redundancy analysis are surveyed. Advances in the theory and application of data rectification are reviewed.2. The principles of gross error detection, data reconciliation in static linear systems and Quasi-steady-state linear data reconciliation are introduced.3. An approach for estimation of undeletable MGEs is proposed based on temporal redundancy information. The predicted values of MGEs can be used in the data reconciliation. The pollution of gross errors is minimized and the reconciliation precision can be ensured. Simulation results demonstrate the efficiency of the proposed method.4. Mass balance model for data rectification is changed when scheduling is performed. A new method is thus proposed by the using of Bayesian network. In this way, schemes are identified in time when the scheduling events happen. Accordingly, mass balance equations for data reconciliation are modified adaptively and the feasible of data rectification is enhanced.5. Yield models of unit operation are introduced in data reconciliation. The pollution of gross errors is minimized and the reconciliation precision can be ensured. Average yield is defined for scheduling event in order to extend the applicability of yield models. Simulation results demonstrate the efficiency of the proposed method.The thesis is concluded with a summary and series of vistas of future data rectification technique.