Hydrocarbon Processing - June 2022 - 81

Biofuels, Alternative/Renewable Fuels
This basic approach is the mechanism
used to generate information that
informs decision-making for actions in
areas such as asset health, efficiency and
predictive emissions.
What can a data lake facilitate? A
data lake is not an end, but rather a means
for analytics tools to deliver the desired
benefits. A larger implementation, with
a data lake and applications, unifies data
while connecting systems and people.
Together, these drive operational performance
and workplace efficiency through
automated workflows, advanced analytics
and enhanced decision support. This
is broad conceptually, so we need to focus
on three applications:
* Feedstock variability
* Energy management
* Asset health.
First, renewable diesel production can
begin with a wide variety of feedstocks,
each of which must be treated differently
to convert them into a suitable intermediate.
While the variety is greater than with
conventional refining processes, different
fats or biomass can be separated into categories
based on their characteristics. A data
lake allows operators to compile a detailed
catalog of feedstocks, including process adjustments
needed to accommodate each of
these feedstocks. Throughout time, analysis
of historical process data from daily
operations builds predictive models that
simplify the steps that operators must take
as they facilitate a feedstock change, no
matter how drastic a shift this change represents.
Once the transition is completed,
the predictive model can optimize the process
for the duration of that feedstock run.
Second, renewable diesel-like its
conventional counterpart-requires energy
inputs during manufacturing. The
same fired heaters, boilers and the like
are used, and these processes are under
the same environmental regulatory constraints.
Optimizing energy consumption
is critically important, calling for detailed
analysis of all energy consumers to vary
their efficiency. This necessary level of detail
depends on extensive instrumentation
and data analysis for every installation.
One beneficial side effect of having all this
data is the ability to use predictive emissions
techniques. When sufficient statistical
information is available via a data lake,
regulatory agencies will often allow a plant
to base its emissions monitoring on modeling
rather than on actual measurements
via analyzers-which is far less expensive
than maintaining a fleet of analyzers.
Third, the following will examine operating
asset health and a specific centrifugal
pump installation (FIG. 3). This
hypothetical installation has been outfitted
with several WirelessHART condition
monitoring instruments, including those
for measuring bearing temperature, vibration,
motor current draw and flow.
Using open communication protocols,
the data lake takes operations data from
these instruments and then centralizes
and contextualizes it, even if the different
sources represent different data types.
Real-time data is delivered to the maintenance
and reliability teams via interactive
dashboards so they can see what is happening
at any moment. This capability
alone would be a major advancement for
many plants, but it is only a start. As historical
data accumulates, it becomes possible
to perform operational analytics.
In our scenario involving this pump,
there are two main analytical methodologies-principles-driven
and data-driven
analytics-and both are available. Each
approach has its capabilities and should
be matched to the application. In some
cases, both can help.
Principles-driven analytics techniques
use laws and principles that have been
codified for many years. Engineers understand
how a pump works (e.g., pump
head, speed, pressure differential), and,
for example, how the current draw vs. liquid
moved provides basic efficiency.
However, the concept can be extended
to examine other variables (FIG. 4), such
as vibration or changes in suction and discharge
pressure, both of which can point to
abnormal situations indicative of developing
problems. With enough data, it is possible
to characterize the deterioration by
using failure mode and effects analysis. For
example, it will be possible to know that
when vibration reaches a specific threshold,
the installation is nearing failure. This
will be based on the specific unit, but will
also draw on experiences with other units
like it within the plant and perhaps within
the larger company.
Simultaneously, data-driven analytics
create statistical models to fit the data. For a
pump, analytics can create a graph plotting
probability of failure on demand (PFD)
against operating hours for a given motor/
pump configuration. The larger the number
of comparable installations, the more
reliable the analysis will be. If the company
FIG. 3. The centrifugal pump in this example
is outfitted with instruments to monitor
critical condition indicators.
FIG. 4. Many equipment condition monitoring techniques use principles-driven analytics to
determine when problems are developing.
Hydrocarbon Processing | JUNE 2022 81
Hydrocarbon Processing - June 2022

Table of Contents for the Digital Edition of Hydrocarbon Processing - June 2022
