Page 309 - Sustainability in the Process Industry Integration and Optimization

286 Cha p te r T w e l v e

is itself based on (initially) rough assumptions, which are revised
during the course of several design loops. At the beginning there are
no specifications for the heat-transferring units, and neither are the
feed temperatures (and several other important factors) fully fixed.
At this stage, then, extremely precise data is not needed. It is important
to recognize that PI and the initial optimization are more about
screening and scoping than about detailed design. In seeking
potential energy savings, it is the general direction in which
optimization should proceed that is the initial concern. If the result
of this first step is that a 15 percent savings in energy is possible, then
this figure is sufficient to confirm the design approach and it doesn’t
much matter if the precise figure is actually 13 or 17 percent.
It is in the regions close to the Pinch that the data should be as
precise as possible (Linnhoff et al., 1982). Also, it is best to remain
inside the Composite Curves in the plot of temperature versus
enthalpy. At the start of data extraction, we might have only a vague
idea of where (and at what temperature) the Process Pinch will occur;
also, the Composite Curves are based solely on data extraction.
Therefore, data extraction must necessarily start from rough
assessments and then be corrected step by step.

12.1.3 How Can Considerable Changes in Specific Heat
Capacities Be Handled?

Further analysis of the flowsheet in Figure 12.1 reveals that phase
changes are very likely to occur when the temperature increases
from 10°C to 165°C. In general, C varies with temperature, but latent
p
heat is also a determining factor. Clearly, using a constant value for
C would not be sufficiently realistic.
p
A segmentation technique has been developed to deal with this
problem. This technique is used, for example, in the software tools
SPRINT (2009) and STAR (2009). The software tools treat the segments
as individual streams that are combined to form input and output
streams. Data extraction is affected by how many segments are used
and their boundary temperatures (see Figure 12.2). Increasing the
number of segments naturally increases the complexity, so this

T 1 T 2 T 3 T 4

? ? ? ?

ΔH ΔH ΔH ΔH

FIGURE 12.2 How should we linearize?