Chemical Engineering June 2021 - 34

PAT is also key to enabling the
adoption of continuous processing
strategies, which can further streamline
production and processing operations,
reducing running costs, cycle
times, energy and resource utilization.
In particular, a study examining the
fractional cost differences between
operating continuous flow and batch
reactors in chemical manufacturing
plants concluded that operating
costs are between 15 and 40% lower
when continuous processes are used
[4]. More precisely, a fourth of costs
for catalysts were required in hydrogenation
stages while 35% could be
saved in expenses associated with
energy usage during nitration [4].
PAT & 'smart' are interconnected
To provide successful, data-driven
insight into processing activities
with PAT, a highly interconnected
infrastructure needs to be created
that can transfer, store, process,
analyze and subsequently visualize
the large volumes of information
continuously generated. To implement
an effective
industrial
internet
of things (IIoT) system, manufacturers
need to move away from
conventional, centralized highperformance
computing strategies
that focus on supercomputers or
clusters of collaborative computers.
Instead, they should favor
high-throughput computing, which
relies on distributed data sharing
and content delivery applications.
As a result of this paradigm shift,
it is necessary to develop suitable
data fusion strategies that can facilitate
the integration of readings
from different instruments to create
optimized quality predictions and a
comprehensive process overview.
PAT can address
these needs
using the latest innovations in sensor
and network technologies, big data
analytics and data mining. In effect, it
can support the creation of a highly
interconnected, comprehensive and
intelligent system that enables industry
4.0 applications. This ultimately
facilitates next-level realtime process
control and quality assurance.
While big-data technologies are
extremely important, it is also essential
to create a suitable infrastructure
that can support
them. One of the
most promising
enabling technologies
is cloud
computing, as it
offers near unlimited
space for
storage and data
processing power
on demand, as
well as decentralized
data availability
(Figure 2).
Hence, it is a key
tool for chemical
manufacturers interested
in setting up big-data-driven
process intelligence strategies aimed
at enhancing their operations and
end-product quality.
Optimal
FIGURE 2. The cloud can enable the creation of an enterprise-wide PAT
system, which is able to acquire and integrate key knowledge from multiple
manufacturing facilities, regardless of their geographical location
Next-generation PAT
The next generation of PAT should
support data transfer to the cloud.
The cloud allows companies to truly
leverage the power of big data by
providing a flexible stack of massive
computing, storage and software
services in a scalable manner and at
a low cost. It offers this without the
need to build the required physical
high-capacity computing infrastructures
that need substantial CAPEX
and OPEX as well as increase the
manufacturing footprint.
A practical example can be found
in (Fourier-transform) near-infrared
(NIR) process spectrometers, which
are able to generate multiple spectra
per second. The volume of data
produced by a single analyzer, let
alone a network containing multiple
analyzers, can quickly exceed the
gigabyte range. Therefore, existing
on-premises servers and data centers
in chemical plants may not have
the ability to store and analyze such
large datasets from multiple devices
and analytical instruments.
While edge computing is generally
better suited to support timecritical
process control, the cloud is
well-equipped to address the needs
of advanced process control during
predictive model-building phases.
In particular, cloud computing can
support the creation of an extremely
advanced PAT knowledge-manCHEMICAL
ENGINEERING WWW.CHEMENGONLINE.COM JUNE 2021
agement system that can perform
complex functions and computing
algorithms, which require more
computing power than can be possibly
provided by edge-computing
devices. Time-critical model execution
can then be delegated to realtime
control systems, such as programmable
logic controllers (PLCs)
and distributed control systems
(DCSs), which are acting on qualitydriven
setpoint changes from the
cloud-based PAT system. These will
then work together to automatically
adjust CPPs, thus optimizing quality.
Consequently, a PAT solution
that can pump data to the cloud
can help chemical manufacturers to
set up advanced, automated operations
that increase production flexibility
and responsiveness.
An additional benefit of utilizing a
framework that supports cloud connectivity
is that companies can set
up an enterprise-wide PAT system,
which is able to acquire and integrate
key knowledge from multiple
manufacturing facilities, regardless
of their geographical location. By
doing so, chemical manufacturers
can gather larger volumes of data,
creating ever more accurate and
precise predictive models. Furthermore,
they can standardize processes
across multiple plants, creating
shared best practices. Such a
system can also support multi-site
collaborations and coordination.
Ultimately, chemical manufacturers
can optimize all aspects of their
operations while increasing consistency
across different factories.
33
http://WWW.CHEMENGONLINE.COM
Chemical Engineering June 2021

Table of Contents for the Digital Edition of Chemical Engineering June 2021
