ILMA Compoundings - June 2021 - 22

must be automated efficiently. Then, AI can be introduced.
" Computers require information to flow in a way that is precise, "
said Williams. " You can't just say, 'I want AI to fix that
problem over there.' The system has to be organized from a
physical and data-streaming point of view. "
When the problem is identified, it's time to decide on an
algorithm to employ. The algorithm, said Williams, is " the
strategy that the AI will use to reach a conclusion. You don't
need to understand exactly how the algorithm is working,
but you need to appreciate the difference. "
Available algorithm systems are:
* Expert: Like the Olyslager Market Analyzer, this AI
collects mountains of domain knowledge to reach a
conclusion.
* Neuro: Like the human brain, neuro systems draw
conclusions by learning and deducing.
* Staccatic: Typically used for predictive maintenance,
these systems draw conclusions using probability and
statistics.
* Deep learning: These systems draw conclusions from
long chains of cause and effect, such as the computer
that loses its first chess game but consistently wins by
the 900th match.
* Machine learning: The system collects data on its own
operations and adjusts accordingly.
Once you've narrowed down the algorithm needed, seek
out best practices in enterprises that are similar and have
" embraced automation first and AI second, " said Williams.
" Take it away from the magical and see the tangible. "
At this point, a classic cost-benefit analysis comes into
play. " Every lubricants manufacturer has lots of processes, "
McCusker said. " Receiving dock, quality department,
supply chain - in each of those areas where there's an
opportunity to introduce more automation, you need to
study what the return on investment is going to be. What's
the value proposition? "
For lessons on capitalizing on AI, even small and mid-size
companies can take cues from such giants as Amazon. Costs
in warehouse management, distribution and manufacturing
can come down just by, in the style of Amazon, positioning
the most frequently used products for pickup and staging
through automated forklifts and processes. " Once you have
all of this data, you can optimize everything, " McCusker
said, adding that AI prices will come down to viable levels as
technology evolves.
Affordable out-of-the-box solutions are currently available,
said Salas. " You can leverage existing tools and use
them against the data you're already collecting, " he said.
22
JUNE 2021 | COMPOUNDINGS | ILMA.ORG
Avoiding information overload is not a simple matter,
he added. Data storage is cheap, but too much data can
create " signal noise " that clouds the search for answers to
inquiries. " As you've got these large datasets, it's important
to start with asking a question and see if the data can inform
it, " Salas said. " Over time, as you amass data, you can see
elements that are missing or that you never use, and you can
reshape the dataset. "
The potential for data and AI to streamline production
and guide strategic planning in other lubricants-dependent
fields, such as marine and mining, is materializing, said
Oosting. " The positive thing from AI is the ability to possess
and analyze massive amounts of data in a short time, " he
said. " We are looking for more uses for our data, and for AI
we can use with that data. "
McCormick is a Pennsylvania-based freelance writer.
" A lot of decisions are still made in the
industry based on estimations and even
emotions. They do a lot of research, but
the last part they're missing is bringing
the pieces together and having the real
insights, the real figures. "
Harald Oosting
President and CEO, Olyslager
http://www.ILMA.ORG
ILMA Compoundings - June 2021

Table of Contents for the Digital Edition of ILMA Compoundings - June 2021
