M. Fillon et al., AN EFFICIENT FORMULATION FOR BATCH REACTOR DATA RECONCILIATION, Industrial & engineering chemistry research, 35(7), 1996, pp. 2288-2298
An efficient formulation of the data reconciliation problem applied to
batch reactor experiments is proposed and discussed in comparison wit
h other works reported on this subject in the Literature. Under the as
sumption that all components in the reaction are known experimentally
and that their chemical compositions are defined, atomic balances are
used as constraints in order to correct raw data values. The formulati
on is based on an a posteriori simultaneous treatment of all measureme
nts over the whole time of experimentation. The main point of interest
in this formulation is the preservation of the measurement identity i
n each sample. Even if component holdups are systematically involved i
n mass balances by pairs to estimate a difference in time, here we cor
rect each component measurement and each sample independently. On this
basis, it is shown how classical statistical tests can be used to dis
criminate gross errors. A case study built from simulated true and noi
sy data gives an overview of the performances of the method. The resul
ts state that the overall time formulation is more powerful than the c
lassical one, which is essentially based on a sequential treatment by
periods of time. Evaluation of the performance for three adapted stati
stical tests leads us to conclude that the generalized likelihood rati
o (GLR) is not only the most appropriate but also the most efficient t
est.