Page 100 - Design of Simple and Robust Process Plants

4.1 Process Synthesis 85
Simple and surveyable methods are preferred to discriminate between alterna-
tives which is the main effort in synthesis step 1. For reactors, discrimination can be
made between CSTR(s)/PFR(s) or a combination of these, including a preliminary
energy configuration based on variable margin. This is comparable with the attain-
able region approach as discussed by Biegler et al. (1997). The difference is that in
the proposed approach the alternative solutions are compared on a variable operat-
ing margin.
Separation selection can be carried out through heuristic rules, or on cost compar-
ison, while for sequencing of distillation columns, minimization of the marginal
vapor flow provide good results (Biegler et al., 1997).
It was noted previously that most practical process synthesis studies have limited
alternatives due to constraints and gradual technology developments. In many
cases, each alternative might be evaluated individually and compared with other
alternatives to avoid solving a MINLP problem. For a larger set of alternatives,
MINLP is the way to proceed.
Integration studies in this layer emphasize the integration of:
. process streams, by taking advantage of less purified streams (including
streams as hydrogen which might form part of an overall network); and
. utility streams as energy, process water, cooling, refrigeration.
The constraints for integration need to be defined, although at this stage they are
preliminary. Integration studies are used to calculate the target consumption for the
flowsheet under study, while options are exercised to achieve lower targets by
manipulation of the process conditions. The outcome of the optimization layer 1 is
a limited set of process alternatives for further evaluation.

4.1.5.2 The second layer
The second layer during synthesis step 2 is also divided into reaction separation and
integration. The intermediate prices are updated, while the energy prices are kept at
the site energy values.
In this layer each alternative is optimized at the maximum achievable NPV. This
means that the cost estimation section and the optimization section for DFC and
NPV calculation has been connected to the simulator. The cost estimation functions
need to be adopted in the working areas to create continuous behavior, and also to
avoid discontinuities. (Most cost estimation databanks have discrete cost data
against equipment size.)
Mixed integer problems are to be avoided, so a continuous optimization problem
is created, and a stage-to-stage calculation method as quite common used for rectifi-
cation which apply integers (number of stages) isn't prefered. Modeling is preferably
carried out with short-cut models as applied for distillation by Fenske±Underwood
equations. For extraction, Kremser equations are used to describe the unit operation.
Separations such as distillation, extraction, stripping, and absorption can also be de-
scribed in mass transfer functions to avoid the equilibrium model description.
For the reactor system the selected CSTR/PFR combination is branched into alter-
native designs where the process simplification has its major input, see Section