Page 308 - Sustainability in the Process Industry Integration and Optimization

T y p i c a l P i t f a l l s a n d How t o Avo i d T h e m 285

150°
160°
ΔH 3
190°
120° 40°
Reactor 80° 45° 45° 10°
ΔH 2 ΔH 1 Feed
140° Unit 140°
FIGURE 12.1 Partial fl owsheet for example process.

heated further by two heat exchangers and then enters a reactor. The
reactor requires the feed stream to be at 165°C. The question is: How
many streams should be extracted?
1. Three streams: 10–45°C, 45–80°C, and 80–165°C
2. Two streams: 10–45°C and 45–165°C
3. One stream: 10–165°C
Choosing option 1 yields a design exactly like the original one;
there will again be three heat exchangers with the same heat transfer
duties as before. This case accounts for claims of “no process
improvement” by some PI critics. Option 2 presents more degrees of
freedom: the first heat exchanger would be the same, but the other
could be modified. Extracting two streams would be appropriate for
cases when the processing unit requires a feed temperature of close
to 80°C.
Option 3 provides the most degrees of freedom and scope for
improvement, but with this design the processing unit feed could be
at any temperature between the 10°C supply and the reactor target of
165°C. If the processing unit is, say, a filter (as assumed by Smith,
2005), then there would probably be some restriction on the filter
supply temperature to ensure proper operation of the filter. If the
processing unit is storage (as assumed by Linnhoff et al., 1982), then
the supply temperature might be restricted to a different range
(depending, e.g., on whether liquid or gas is stored). This simple
example demonstrates that choosing the right stream from which to
extract data cannot be a fully automatic process. Making this choice
requires assessments related to specific processing units and
performance requirements for the plant.
12.1.2 How Precise Must the Data Be at Each Step?
There are frequently questions about required data precision, and a
common excuse for not applying PI analysis is that the data on a
plant is not sufficiently precise. However, the process of applying PI