connected device or equipment asset. The digital model can provide
real-time status information, historical operating characteristics,
and benchmark equipment performance against other similar and/
or best-performing equipment at the facility. It can also be used to
support business process improvements such as prioritizing repair
work, establishing a prescriptive repair methodology, and supporting
inventory management and work scheduling.
Using the digital model established for each equipment asset, and
the specific rules established to prescribe normal operating parameters,
any marked deviations in operating conditions (such as for temperature,
pressure, and power) can trigger an automatic alert to technical and
operating teams. This early warning signal enables teams to foresee
the increased likelihood of potential equipment failure, or emergent
operating issue, and help to identify prospective corrective and/or pre-
ventative measures. However, this is not the same as predictive failure.
While it is true that many operating parameters are defined to
minimize the amount of time equipment operates under conditions
that might lead to an increased risk of failure, the simple fact that a
piece of equipment might drift outside of the range for a short period
of time does not mean it will fail. The analytics behind predictive failure go much deeper, to consider the unique operating conditions for
every single piece of connected equipment over its entire lifecycle.
Machine learning constantly analyzes all of the historical and current information to find patterns and trends buried deep in the data.
For example, a pump might fail because of a unique combination of
high operating temperature, pressure, fluid viscosity, flow rate, head and
shaft power … all for a duration of seven hours. There are many factors
- and combinations of factors - that could lead to a mechanical failure
that it is simply beyond our human capability to connect all of the dots.
Key benefits of an Io T solution
Simply put, machine learning technology can find the patterns and
trends in massive volumes of data that humans cannot. The ability to
predict equipment failure with sufficient advance warning to prevent
forced downtime and allow for a well-planned remediation effort can
drive down operating costs and maximize production. Further, the ability
to aggregate and analyze information from dispersed production sites
can help to refine these predictive insights across an entire portfolio of
Real-time monitoring, via a customized dashboard, can be used to moni-
tor activity and “visualize” the predictive capabilities of the data solution
(such as expected time to failure for each piece of critical equipment or
component). Dashboards allow teams to view relevant asset telemetry
and compare current operating conditions with established utilization
metrics and benchmarks. The dashboard can be integrated with existing
on-site data management and control systems and, if connected to the
cloud, can be viewed from any location around the globe with internet
access. The dashboard also provides the ability to compare and optimize
performance, efficiency, and safety metrics across multiple operations.
The ability to incorporate human knowledge into the Io T business
logic minimizes the uncertainty and learning curve that can occur in
the absence of a central repositor y of exper t technical knowledge and
operating history. Combined with the powerful data-driven predictive
insights related to abnormal operating conditions and prospective
equipment failure, facility operators will have better intelligence
from which to make informed decisions to help minimize unplanned
downtime and maintain production.
When a piece of equipment stops working, every hour of downtime
spent on repairs increases service costs and lowers revenue through
non-production. Rather than simply generating a time-stamped error
or fault code, the Io T solution can provide an important source of
data-driven insight to guide technicians to a probable root cause and
remediation solution faster. The machine learning capability of the
software can also identify diagnostic steps taken, the problem that was
identified, the fix that was applied and the outcome.
A key challenge with any data-driven technology solution is the mas-
sive quantity of data that must be captured and acted upon in real time
prior to being forwarded to a database for processing, analysis, and
storage. In a typical deployment, large data sets are for warded to cloud
databases where the machine learning software works to develop more
accurate digital device models. In this circumstance, the analytics work
is done in the cloud and pushed down to the site for local processing.
However, in many circumstances, a balance is required between the
volume of data needed to drive the accuracy of a “deep data” analytics
solution and the bandwidth required to support it. This capability is
relevant where there is limited connectivity and/or bandwidth con-
straints. Advanced Io T solutions use a process called “edge analytics”
whereby the complex set of rules that embody the reason and business
logic are used to detect specific operating conditions and orchestrate
required actions right at the data source (the edge). •
Prasantha Jayakody is a Sr. Product Manager for Bsquare where he focuses on lowering barriers to adoption for Io T via turn-key solutions and cloud-based delivery.
Starting with a sound Io T strategy, companies can achieve maximum
value upon completion of all five steps.