enhance the SMEs’ decision making processes by providing this
information and identifying data patterns and trends that are
related to particular failure modes. This information is often
not obvious, even to the experienced SME.
One way to begin the data mining process is to look at
equipment history data to gain a better understanding of
potential failures, degradation mechanisms and the impact
of current maintenance and reliability (M&R) strategies.
Past performance is the best indicator of future failure
and the equipment work history is the key to providing
the information needed to improve M&R efficiency and
effectiveness. When properly mined, this data can provide a
wealth of information about past performance, opportunity
identification and strategy improvement.
Data mining for APM
Well-constructed and well-maintained pipelines provide
a safer, more efficient and more cost effective means to
transport crude oil. This is, in part, because pipeline owners/
operators have a broad array of solutions from which to
choose to help ensure pipeline integrity. Asset performance
management (APM) solutions in particular can improve the
performance of pipeline assets with innovative technology to
reduce risk and improve profitability.
APM data mining draws data from several primary sources
including, but not limited to, work history events, production
loss events, asset specifications, functional locations and
inspection events. Work history, production loss and
inspection data are all capturing events, making it absolutely
essential that codes are used to enforce consistency of data
collection.
The largest number of records is usually found in work
history. Work history is made of maintenance work orders
which typically contain detailed information about the
functional location, equipment number, actions taken, cost,
who did the work, failure mode, failure cause, part replaced,
etc. The work order may be due to a failure, preventive
maintenance procedure, etc.
Ideally, all of the data to be mined would be contained
in a single enterprise system, but this is not always the case.
When the data is not contained in a single database, data
gathering becomes an important step in the data mining
process. These separate data sets must be set up with vectors
to link key data elements together.
Joining event records together with the asset hierarchy
(category–class–type) and the functional hierarchy (site–unit
–system–function) allows event data to be presented and
understood in a broader context.
There can also be challenges getting the data into a
condition which is appropriate for use in reliability analytics
and other statistical methods. Data coming from the field can
have numerous problems which must be corrected. In some
cases, the data may need to be cleaned. These corrections can
be handled by a data conditioning module.
Practical data mining results
Data mining is a powerful tool to improve APM. In practical
terms, the results of data mining activities can provide the
user with information to support asset improvement activities.
Typical actionable information include identifying key assets
where limited resources can be applied to create maximum
value, quantifying the value of unreliability by functional
location and/or equipment classification, and providing
information about asset strategy performance. Additionally,
organisations must identify the sources of maintenance work
and measure the effectiveness of preventive maintenance
programmes.
Preventive and predictive maintenance are key parts of
the overall maintenance strategy. The best practice is to
link this data to the enterprise level APM system for routine
data mining, analysis and strategy development. When
coupled with product and process data, a picture of pipeline
equipment performance and failure degradation mechanisms
can be seen for different products and operating windows.
With ageing pipelines, maintenance is the primary method
to restore performance and prolong the equipment’s life. This
is why data mining is a powerful tool to improve APM.
Advanced APM technology and data mining
Data is no longer just numbers and words. Visual technology
is also a critical component for data mining and advancing
technology allows operators to see the failures with
3D imaging and colour maps. For example, visual APM
solutions enable asset performance or maintenance
engineers to visualise or contextualise key APM data so
that they can better analyse and draw conclusions on the
heath of pipeline assets and identify any current risks to
operations.
Visual technology allows engineers to immediately
compare the health of two neighbouring pumps, rather than
identifying them separately without visual capabilities. Not
only does visual technology enable quick identification of
location and proximity, the colour coding features gives
instant access to health and risk assessment data so engineers
can move forward with any necessary repairs to the pumps.
As an industry example, AVEVA and Meridium jointly
developed visualAPM, an APM solution designed to improve
use of data to develop maintenance and asset management
strategies. The solution can display the key mechanical
integrity criteria of a selected area of plant pipework using
the configurable colour-coding system, which is based on data
for the pipe’s thickness, wall loss, corrosion and replacement
date. These metrics are calculated using predictive algorithms
from data obtained during plant inspections. Using visualAPM,
the asset performance engineer can identify the proximity of
areas of concern on the 3D model (for example, understanding
the mechanical integrity of a particular section of pipe by
highlighting critical areas) and then make appropriate decisions
for corrective action.
The future of data mining
The future of data mining will be critically important as
technology and methods for asset management continue to
develop. In particular, as pipelines continue to age and require
maintenance, asset management and data mining will be
necessary to predict failures and ensure reliability.
34
World Pipelines
/
SEPTEMBER 2014
