ILMA Compoundings – February 2019 - 18

observation of data and past experiences, like children
acquiring knowledge from examples."
AI isn't a pipe dream: It's here right now, and it can be
deployed in existing production lines provided that you
connect the right data for the problems you're looking to
solve and that setpoints can be tweaked to take advantage of
the generated actionable insights.

Pros to the Plant
Energy efficiency and sustainability are two of the most
important areas for lubricant manufacturers, and the
metrics pertaining to these areas need to be continuously
tracked. For many manufacturers, this is currently done
using spreadsheets, phone calls and manual processes, which
makes it difficult to accurately monitor and predict key cost
drivers for energy and sustainability.
"Performance-monitoring software can have a huge
impact here," Joshi said. "By leveraging systems that collect
data, performance-monitoring software can monitor and
predict the key business drivers automatically and continuously. This provides an unprecedented level of visibility into
these metrics."
AI automates the analysis of operational and historical
data collected from all assets within the plant. This increases
the speed and accuracy with which a plant operator can
gain actionable intelligence from the data. As a result, the
plant operator can keep ahead of issues that could lead to
unplanned downtime, excessive energy usage, inefficient
operations and decreased throughput.
AI allows plant operators to have the latest asset performance
data and insights at the ready, so they may ensure that
assets are always running at the highest efficiency.
"Poring over bloated spreadsheets not only increases the
time needed to complete a particular task, but it decreases
the accuracy as well," Joshi said. "After all, humans are not
machines. We tend to be error-prone when dealing with
computations involving a lot of numbers. Anytime a human
enters or manipulates data sets, the possibility for error
skyrockets."
AI allows plant operators to have the latest asset performance data and insights at the ready, so they may ensure
that assets are always running at the highest efficiency.
In its purest form, a typical continuous production process such as lubricant production begins with raw materials
and then applies a standardized recipe to fabricate a product
with a defined quality in a predetermined time. It's never
that easy in practice - the result is usually unstable and
inconsistent over time. Properties of the final material, such

FEBRUARY 2019

| COMPOUNDINGS | ILMA.ORG

as base aspect and mechanical and chemical properties,
always slightly differ, or worse.
Hugo says this variation of production quality often depends
on the heterogenous nature of the raw material, even in the
same batch, and in variables that cannot be controlled by the
operators, such as the exterior temperature, humidity, etc.
"Recipes are getting incredibly complex, while customers
expect faultless products. When rule-based systems and
classic engineering cannot compensate for these conditions
soon enough, AI is a tremendous helper to correct problems
before it's too late," he said. "Indeed, AI can predict in
real time the future of the production. It then enables the
operators to adapt the settings at the right time to change
this result to the best outcome."
AI doesn't try to resolve a problem in a solely "logical
way" like humans, but probabilistically, taking into consideration multiple variables and detecting correlations and new
improvement paths even in complex production processes.

The Data Collected
José Andrés García, lead data scientist with Wizata, notes
that a good AI and data science methodology can generate
concrete answers in a couple of months, with swift testing
and deployment.
"Apart from being a problem-solving expert detecting
hidden root causes, AI can predict, anticipate and adapt the
production in milliseconds," García said. "AI is also a fantastic helper to innovate more and at a faster pace in a highly
competitive environment, where high quality is expected
while pushing costs to a minimum."
For example, AI can also simulate how changes can affect
a plant's goals before operators try them on physical production lines and can help them to develop novel products.
According to Hugo, AI-based quality assurance could lead
to a 50 percent productivity increase and companies could
leverage profit margins 5 to 15 percent higher than their
competitors.
Every piece of machinery in a production line at a
lubricant plant generates data. Various performance data
collected include flow, pressure, temperature, energy
consumption and more. All of these data points can be used
to benefit the plant, from predicting maintenance issues to
determining the business value generated by each asset.
García shares that you cannot just feed all your raw data
to AI models and expect them to give you answers for
questions that you didn't even ask. Rather than trying to
capture all possible data by connecting every sensor to the
cloud and implementing AI everywhere at once, he advises
manufacturing clients to focus on concrete problems they
Continued on page 21

http://www.ILMA.ORG

ILMA Compoundings – February 2019

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