Reconciliation
Usually more data is measured than necessary. Reconciliation makes redundant data consistent with the mathematical model. Moreover, unmeasured parameters of the model are estimated on the maximum likelihood principle. Reconciled data are generally more accurate than the measured ones. In general, reconciliation is a method for optimum estimation of model's parameters. Moreover, reconciliation represents basis for other activities related to validation of plant data - especially to elimination of gross measuring errors and to optimisation of the overall measurement process.

Propagation of measurement errors
Results of data processing are usually of different accuracy. In practice we can meet in some cases with errors in hundreds of per cents of real values. Information about accuracy of results (confidence intervals) should accompany all values important in further decisions.

Data analysis
Confront your data with the model. A bad fit can be caused either by gross measurement error, or by a model error. Both discrepancies can devalue your results.

Detection and identification of gross measurement errors
Reconciliation provides powerful tools for detection of gross errors presence and also for finding sources of gross errors (gross errors identification).

Model errors
Sometimes are discrepancies between data and model caused by model's inadequacy (neglecting process dynamics, unmeasured leaks, too simplified models of unit operations, etc.). Reconciliation represents an efficient method of model building, to incorporate more complex features in the model.

Measurement optimisation
Even advanced methods of data processing can't substitute for a bad measurement plan. Reconciliation provides methods for either optimising existing instrumentation system, or designing a new one.

Case studies
The most important fields of reconciliation will be covered: mass, component and energy balancing reconciliation of non-linear models