IEN / OCTOBER '16
35
www.ien.com
Today's Designer
may be irrelevant for other solutions. As a result, organizations still have a significant oppor-
tunity for competitive differentiation by understanding and deploying pragmatic data-analyt-
ics solutions that can predict machine failures ahead of time.
As data analytics for IoT continues to mature, forward-thinking, pragmatic companies that
deploy analytics in a way that solves real-world problems do so in a sequence of steps, each
of which builds on the previous step. This maturity process starts with basic connectivity
and combines insights from master mechanics that have intimate knowledge of how the
machines operate.
The critical insight from master mechanics informs IoT solution designers about what to
measure to answer unknown questions (e.g., "How do we automatically determine when an
elevator is 60% likely to fail within 90 days?"). Only after those intuition-driven hypotheses are
developed can data analytics be applied to real machine data to prove or disprove the hy-
potheses over time. The
combination of human
input and smart data au-
tomation is the only way
to meaningfully predict
machine failure.
As data is collected
and analyzed from ma-
chines, leading orga-
nizations can not only
reduce system down-
time, but also reduce
the amount of human
input required in or-
der to make a mainte-
nance decision. Maturity
Stages for a diagram
of how different types
of analytics approach-
es relate and build on
one another.
Best
Practices
Here is a set of best
practices that leading
manufacturers of smart
connected equipment
use to predict machine
failure, reduce opera-
tional expenses for their
customers, and increase
overall product and pro-
cess quality:
1. Rely on human
intuition from
expert mechan-
ics. Don't discard
human intuition;
in fact, rely on it.
Leading compa-
nies assign their
top mechanics or
machine design-
ers to predictive
maintenance
projects that
understand the
machine's internal
operation and
understand which
components often
fail and why. It is
actually the com-
bination of these
master mechanics
with smart data
automation that
produces world-
class predictive
maintenance
results.
2. Take a crawl,
walk, run ap -
proach. Don't
succumb to the
false truth that
finding predictive
analytics insights
can be auto-
mated. Starting
slow, collect data,
prove/disprove
hypotheses over time, and then zero in on the
right mix of sensors and report frequencies as a
learning process that evolves over time.
3. Embrace predictive analytics as a journey. There is
a causality dilemma when it comes to knowing which
comes first: the data or the analytics. In reality, the two are
inextricably co-dependent. Embracing predictive analytics
as a journey that starts with basic connectivity is a key recipe for success.
Mark Benson is the CTO at Exosite, an Internet of Things (IoT) platform that makes it easy
to build connected products, solutions, and businesses. For more information, visit www.
exosite.com.