Abstract
It is usually difficult to estimate process parameters used in digital simulation because online data from chemical plants include errors. To improve the reliability of data, this paper proposes a data reconciliation method based on direct use of the measured variables, called SEBDARM, instead of the conventional method with balanced variables. The new method is applied to a pilot plant and its validity is discussed by use of experimental data and simulation. As a result, the following conclusions were obtained.
1) On-line data can be adjusted satisfactorily by SEBDARM.
2) It is easy to investigate the necessity of calibration of each sensor because there is one-to-one correspondence between measured variables and reconciled variables.
3) The reconciled data are not influenced by the numbers of reduced balance equations, so we can attempt to observe variables corresponding to some purpose by using the minimum number of balanced equations.
This method can be utilized to observe the change of parameters and the state of the operation.
