Page 102 - Design of Simple and Robust Process Plants

4.1 Process Synthesis 87
set free. All optimizations are carried out on economic performance as NPV. As
prices will vary over the lifetime of the project, sensitivity studies will be executed
after the majority of the design is fixed in order to evaluate the impact of these cir-
cumstances on operation and any significant design parameter.

4.1.5.4 The fourth layer
In the fourth layer, which is the successive activity during the optimization synthesis
step, optimization is performed for the whole process. The intermediate prices and
energy prices are updated. The reaction and finishing sections are combined, and
the overall process is optimized at this point. The DOFs are limited to all operational
DOFs and some major process design parameters such as reactor dimensions and
major recycle equipment dimensions. Most design parameters will be fixed, for
example the number of theoretical stages for separation columns and the overall
heat/mass transfer coefficients will normally be fixed. The intermediate conditions
between the subsections need to be verified with those used during layer 3. A large
deviation might mean that layer 3 has to be repeated. The process sections are com-
bined for the ultimate process optimization. In this case, the combined sections are
optimized as one overall process with the operational DOFs and the major design
parameters free. It will be clear that robustness of the model is a primary require-
ment, particularly if more recycle loops have to be closed. The optimization is con-
cluded by varying the prices to judge the impact on operational as well as major
design parameters. The model might be used as a basis for an operational optimiza-
tion model.

Summary
. The methodology for optimization is built up on a layered approach. This
gives the engineer valuable feedback on the model building and process
design evolution. As MINLP optimizations are not yet robust in commercial
software, the concept also follows the evaluation of individual optimized alter-
natives, although both methods are applicable.
. The methodology is based on four optimization layers. Each higher layer will
gradually increase the size of the models (more detailed), and increase the
optimization accuracy, but reduce the number of alternative options while
descending the decision tree and select the number of DOFs as relevant for
the problem.
. The first layer to be executed during synthesis step 1 is supportive to:
± selection of the reactor PFR versus CSTRs or combination;
± selection of the separation techniques; and
± pre-selection of the separation sequence.

The optimizations are based on the evaluation of alternatives within a process sec-
tion with comparisons of operational costs for reactor type selection and distillation
sequence selection. Only in specific cases might capital cost be introduced.