ILMA Compoundings – February 2019 - 21

Continued from page 18
want to tackle, before prioritizing these challenges based on
their impact and return on investment.
"It's only then that you should experiment on the specific
data that could help to solve the issues," he said. "To select
the right data and generate the right testable hypotheses to
build specific AI models, it's best to involve the right people
in innovation teams that link business and domain experts
with data experts, such as data scientists and data engineers."

Maintenance and Safety Issues
Because AI has the ability to analyze data rapidly and
accurately, it allows plant managers to be proactive about
maintenance.
"By comparing historical data to the current operational
data, AI can predict when a machine may behave anomalously," Joshi said. "This allows the operators to address the
issue before it becomes much more expensive."
Energy is one of the largest expenses for any manufacturing facility, and operators can also detect current issues with
performance, such as increased energy consumption.
"Increased usage is not only costly for the business, but it
can also be indicative of an issue or impending failure," Joshi
said. "This can be related to a machine, pump or a membrane. It can show that the asset is having to work harder to
accomplish the same throughput. This helps the operators to
figure out how to prevent excessive resource consumption."
García says AI optimizes the performance of production
lines by stabilizing the processes, and this results in machines
functioning longer and safely with less maintenance, while
using less water and energy.
Apart from streamlining production with the added effect
of lowering maintenance costs and ensuring more stable -
and, hence, safer - conditions, AI models can also help to
increase the reliability of production lines by managing the
machines and their setpoints based on the input so that they
function at an optimal level according to where they are in
their life cycle.
"Instead of fixed maintenance interventions, AI also
allows dynamic, preventive and data-driven interventions,
where failures are detected proactively before they happen
and result in extended damages, so that floor managers
can optimize resources and extend life cycles," García said.
"We're now going beyond predictive maintenance with prescriptive maintenance. Sophisticated AI can not only predict
when problems will happen, they recommend preventive
measures to avoid them to take place."
Efficient maintenance practices assisted by AI can greatly
help to ensure safe operations for machines and operators,
but AI can also specifically be tasked to maximize safety. For

example, there are AI models that predict temperatures of
blast furnace hours in the future and recommend actions,
allowing for operators to ensure that the machinery won't
function out of safe bounds, and to deploy corrective measures if there's a risk, well before it's too late to act.

Looking Ahead
Many lubricant plants are looking to utilize this technology
in the next five to 10 years and are starting to experiment
with automation and looking into how AI and robotics
can help them in the future. Martin Midstream Partners
in Metairie, Louisiana, for instance, has some automated
processes in action at its lubricant plant and is planning to
do more in the years ahead.
Overall, most experts in the industry believe AI will
continue to become an integral part of enterprise software
and robotics will leverage AI to become more efficient on
the production line.
"Over the next five to 10 years, we will see lubricant
plants increasingly leverage data to drive their decision-making process," Joshi said. "If information is provided in an
easy-to-digest format, then everyone would be willing to
consume it. The problem with today's performancemonitoring process at these plants is that it's manual and
laborious. AI will transform the way information is consumed by the operators. By surfacing relevant information at
the right time, it will help the operators become data-savvy
decision-makers."
Efficient maintenance practices assisted by AI can greatly
help to ensure safe operations for machines and operators,
but AI can also specifically be tasked to maximize safety.
Hugo anticipates that intelligent systems leveraging technologies such as deep learning will get smarter in the future
by digitizing, interpreting and learning from raw data such
as electromagnetic (e.g., images and videos), thermodynamic
(e.g., heat distribution), mechanical waves (e.g., sound) and
other signals measured by innovative sensors.
"This will enable AI to take into account more granular data to detect uncharted and unexplained patterns,
empowering employees to make better decisions," he said.
"Companies will take less time to clean and give context to
this data, letting the data science teams of clients experiment
more efficiently to innovate at a faster pace."
Loria is an award-winning journalist who has been writing for major
newspapers and magazines for close to 20 years, on topics as diverse as
sports, business and technology. You can view some of his recent writing
at keithloria.contently.com.

http://keithloria.contently.com

ILMA Compoundings – February 2019

Table of Contents for the Digital Edition of ILMA Compoundings – February 2019