Page 83 - Design of Simple and Robust Process Plants

68 Chapter 4 Process Synthesis and Design Optimization
The onion model is complemented with an arrow crossing all layers to emphasize
the need to interactively design between the different layers, while applying optimi-
zation techniques and simplification efforts through idea generation as tools to
evolve to the optimum process design. The modified onion model is named the
interactive onion model to emphasize the interaction between the layers in compar-
ison to the conventional onion model, Figure 3.3. Process interaction optimization
and simplification between the layers will form a ªred lineº through this chapter.
The onion approach seems contradictory to those who like to solve all the options
as one large structural optimization problem, in a MINLP (mixed integer non linear
program) environment. It is the author's opinion that this will not happen during
the approaching decades because:
1. The number of design options that may be implemented is greatly under-
assessed; in other words, there is an incomplete set of design options.
Options often evolve during a design ± they are not available beforehand.
Some examples of process simplification illustrating the impact on the devel-
opment of a distillation sequence are shown in Chapter 5 (Section 5.3.2).
There is no doubt that in the long term, process knowledge and design alter-
natives will be extended and captured using a structural method.
2. The problem structure often leads to non-convex optimization problems,
which in turn usually leads to locally optimal solutions (or not).
3. Engineers always wish to keep track of flowsheet development in order to
understand and evaluate the choices. In other words, even if the process
design can be captured and solved as one problem, engineers will evaluate
the steps in the program in a layered manner so that they understand the
logic and bases behind the crucial decisions.
4. The mathematical techniques available are not (yet) sufficient to cope with
these large-scale MINLP problems
Based on the above arguments, the hierarchical approach is preferred where design
decisions made at each level of hierarchy bound the optimal solutions, which was
also concluded by Douglas and Stephanopoulos 1995.
The strong point is that a synthesis methodology, computational tools and search
algorithms exist to evaluate and optimize the search for an optimal design. The
weak point in the whole synthesis process as by to day is the lack of a structured set
of design alternatives. The design of a process still heavily depend on the creativity
and experience field of the process designers to develop alternative options. The
development of heuristic rules for the pre-selection of alternative separation units
made progress but the disadvantage will remain its inherent restricted validity and
applicability. The development of a separation system is next to the reactor section
the most challenging tasks. It is here where Douglas and Stephanopoulos 1995
made progress by introduction of the general separation system superstructure, see
Section 4.2.2 on separation.