According
to Löfsten,1
“the costs of maintenance are estimated to be between 15% and 40% of production
costs.” Consequently,
efforts to reduce such costs without compromising risk to equipment, personnel
safety and environmental integrity are welcomed.
The
function of maintenance should be to enable the level of product output
required by the business at the minimum maintenance cost, while honoring the
endemic risks. Typical approaches balance corrective maintenance with preventive
maintenance. Corrective maintenance restores equipment function after it fails
or is deemed to fail and covers strategies of planned run-to-failure and the
unexpected failures with unplanned downtime. The most expensive maintenance
is emergency maintenance after a forced shutdown when all resources
are utilized in an urgent situation to fix the problem and restore production. Running
equipment to the point of failure can cost 3–10 times the cost of regular
maintenance.2 Consequently, preventive maintenance attempts to
quantify potential failures and implement corrective action steps to ensure
equipment does not reach a failure mode, thus avoiding such emergency costs.
Strategies
for preventative maintenance. Strategies for preventive maintenance have
been developed based on interpretations of the P-F curve: a simple diagram
highlighting the transition from normal running equipment, through degradation,
and functional failure up to catastrophic failure.
The
P-F curve was first introduced in 1978 by United Airlines engineers Stanley
Nowlan and Howard Heap. They performed a study for the U.S. Department of
Defense that showed that repairing or replacing assets based on the condition
was more effective than doing so based on age.3
The
practice of preventive maintenance has been immersed in developing strategies
to plan maintenance that reduces failure occurrences. The generally held
consensus is that the interval for planned inspections should be at least half
of the calculated P-F time—perhaps a distortion of the Nyquist
frequency (or folding frequency), the highest frequency that can be
restored from a recorded time sample dataset4 to fully represent a
situation, which suggests that sampling at half the P-F failure frequency is
adequate to interpret failure frequencies. All of this is based on the assumed
accuracy of the P-F time. This leads to an extremely labor intensive and costly
strategy, along with variability and inconclusive efforts to discover the
degradation that will lead to failure. In real situations, equipment that
exhibits higher risk and consequences of failure would be inspected and
serviced at a higher frequency, which elevates costs.
Detecting
impending failures with advances in condition-based monitoring. Fortunately,
advances in condition-based monitoring, especially predictive analytics in asset
performance management (APM) implementations, can greatly improve detection of
impending failures (the P-F interval). This results in early action to prevent
or correct equipment abnormalities and greatly reduces those costs. APM-based
predictive maintenance uses advanced monitoring techniques with machine-learning
algorithms to provide recognition of patterns that indicate normal operations,
abnormal operations and the distinct patterns representing extremely early
detection of impending failures. In effect, such techniques move the p-point
much earlier, providing earlier warnings of degradation and failure in weeks
and months, not just days. The main principle is a precise prediction of likelihood
of failure affording an appropriate time window to determine the proper time
for safe, planned maintenance intervention.
This
is a sound strategy for critical equipment, but more unrealized benefits can
emerge from considering a newer maintenance strategy involving automated
monitoring. At the core of the technology is a monitoring engine that performs intense,
detailed inspection of dozens—and up to hundreds—of variables on and around a
piece of equipment. Such an engine is much more capable of recognizing more
detailed patterns over many more dimensions in the multi-variate and temporal digital
shadows than humans can. In addition, a monitoring engine can perform this duty
every few minutes for earlier warnings of imminent and potential failures in
months and weeks, rather than days.
In
contrast, a planned maintenance inspection with a once-a-week or once-a-month schedule
cannot compete with the inspection intensity or periodicity, and it cannot
provide the extreme early warnings of equipment malfunction. Additionally, the
capability and cost of such inspections on important but lower criticality
equipment can be dramatic. One maintenance manager at a large processing plant
declared 52% of his maintenance budget was absorbed by inspections on pumps,
but not the highly critical equipment. A large South American oil company was
able to reduce its annual maintenance budget
by 25% by replacing several manual inspections with lower cost and more
thorough automated services.
Determining
the right time for a shutdown. Upon a prediction of time-to-failure, APM-based
initiatives enable facilities to determine the optimal time to initiate the
shutdown and advise the correct order of activities to minimize profitability
losses. The shutdown can be planned according to the availability of spares and
staffing resources. Knowing a failure is imminent enables the orchestration of an
improved operations schedule leading up to and during the outage, which can
facilitate honoring customer delivery commitments.
APM
tools can also determine the actual or real criticality of a piece of
equipment, not just as a standalone but as an integral part of the system as a whole,
calculating its real value to the bottom-line production. Consequently, these
tools can determine the absolute risk of poor equipment performance and the
cost of different decisions on availability, redundancy, equipment sizing and
process intermediate storage. This capability contributes significantly to plans
made in the maintenance department and their bottom-line performance.
In
all of this, maintenance does not stand alone, but as a functional department
contributing to the wider financial and sustainability goals of a business
enterprise. Reducing the estimated 15%–40% of production costs is a lofty goal.
The first step is granting planned maintenance the early detection of
malfunction to avoid unplanned shutdowns, plan orderly safe and shorter
shutdowns where necessary, and minimize production losses. In addition, it soon
becomes clear that using an APM monitoring engine in place of ongoing
inspection services is cheaper and more thorough. This reduces manual
maintenance costs and has a significant effect on overall plant profitability. HP
LITERATURE
CITED
MIKE BROOKS is Global Director, APM Solutions at AspenTech. Previously, he was the Chief Operating Officer at Mtell, which pioneered machine-learning for managing the health of industrial equipment. He has also served as a venture executive with Chevron Technology Ventures and held senior roles at five startups. Brooks began his career as an engineer at Esso and Chevron.