Summary

Practical data reconciliation is not easy to handle due to less than optimal redundancy of flow measurement, inherent inaccuracy of measurements, unexpected process events, and other factors.

The mass balance constraint enforced by classical data reconciliation methods is often not sufficient, but while component and enthalpy balance constraints can improve the accuracy of reconciled results, it is impractical to build and maintain component and enthalpy balances for all streams and processes in the entire plant.

Gross error detection and elimination is key to accurate and manageable data reconciliation but if the reconciliation requires nonlinear optimizations, gross error detection becomes very difficult.

Our approach to data reconciliation with additional constraints may be a breakthrough to solving these difficulties in practice with the following two strong points:

• We can selectively and easily add balance models and constraints to cover the component, enthalpy and other kinds of constraints.

• We can use the same gross error detection and elimination methods as used in the linear data reconciliation problem of mass balance.