Page 63 - Sustainability in the Process Industry Integration and Optimization

40 Cha p te r T h r ee

feature reasonable costs. However, this is seldom the global
minimum for the total annualized cost (i.e., the sum of
annual operating costs and annualized investment costs).

The logic behind evaluating a system in terms of the upper bound
on its performance is this: if the best possible system performance is
still insufficient to satisfy the specified requirements, then no further
time and effort should be spent on designing that system. The next
section gives more details on the use of targeting to managing design
problem complexity.
3.10.5 Applying Process Insight
Mathematical tools are absolutely necessary for optimizing the
design and operation of industrial processes. However, meaningful
and applicable results are obtained only when process insight is used
to guide the model building, the optimizing, and the interpretation
of results. Each optimization problem has its own particular features.
For example:
• The processing of fruits for canning imposes different water
requirements and practices from those for the processing of
poultry.
• When designing HENs, the various underlying process
operations that need cooling or heating should be properly
examined for data extraction. In some cases, process knowledge
may aid in the lumping of different heating/cooling needs,
thereby simplifying the flowsheet. Process knowledge is also
employed when partitioning properly into segments a process
stream whose heat capacity flow rate varies widely.

Every specific requirement discovered in the iterative process of
model improvement should be thoroughly documented and
implemented in the model—for instance, in the form of constraints
or simplifying assumptions.
A good practice is to perform targeting for the desired application:
a heat recovery problem, water management, a separator network, or
reactor network design. Targeting provides information on potential
performance. Targeting procedures have been developed and well
tested for a number of applications and domains (see Chapter 2 for a
review).
The benefits of using variants of Pinch Analysis or other targeting
are twofold. First, the designer can estimate the best possible
performance of the system by using simple models and calculations,
even before using rigorous design procedures, saving valuable time.
The obtained targets can be used in preliminary sensitivity studies
to determine which operations and units should be included in the
design and which should be left out. This approach can greatly