Page 404 - Design of Simple and Robust Process Plants

9.6 Project Methodology for Operation Optimization 391
. The model has been developed internally.
. The model is described in the literature.
. The model is available commercially.
. The model needs to be developed.
The models might be available in empirical or fundamental forma. The crucial ques-
tions is whether it has the required accuracy to contribute to the optimization. It
was argued previously that the contribution of optimization to operational cost sav-
ings is in the order of a few percent. Therefore, the reactor model needs to meet (in
most cases)a high accuracy. The development of a fundamental model is, in gen-
eral, a research activity which is not discussed at this point in time. However, the
development of an empirical model and the validation of reactor models will de-
tailed in the following text.
The development of an empirical steady-state reactor model might be based on
input/output analysis of the process. The technique to develop these emphasizes
the following elements:
. Development of an experimental design program, including all operational
degrees of freedom.
. Accurate measurements of all relevant process conditions and feed and prod-
uct compositions.
. Process stability during testing.
. Selection of the proper parameters to update the model for ageing over the
life cycle of the reactor (catalyst).
. Modeling is preferably done as a regression model where the equations still
have a sort of physical representation, including parameters and which obey
mass and energy conservation laws.

As was proposed at the feasibility step, the specifics of the reactor and the contribu-
tion to the economics, set the effort and the accuracy requirements. The develop-
ment of such a model depends heavily on process stability, the operational experi-
mental range, and the measurement accuracy. Particular measurements and analyti-
cal techniques often need to be installed and developed before the model
development can effectively start.
The validation of a reactor model (fundamental as well as empirical)is seen as
a separate effort from the overall process model validation. This is particularly so
as specific measuring and analytical techniques are involved which include meas-
urement of impurities and product properties (as in polymers)that can have a
major impact on the process. Fundamental models developed based on laboratory
data also need to be validated in the plant to confirm their validity at plant condi-
tions.
The technique is based on minimization of the square of the deviation between
measurements and prediction. The degrees of freedom for the developer to cope
with mismatch are limited, and include: