Beruflich Dokumente
Kultur Dokumente
ontrol system metries can be highly effective in managing a . Metrics answer suppositions such as, if
C process control system. They help ensure overall system health
and integrity, focus available technical support resources on
the highest-priority areas and serve to pUt all stakeholders on the
our regulatory controls are sound, most
same page with regard to control system performance issues. control valves will be in automatic; if our
The metrics approach can be applied to control systems of any
vintage. New control systems normally don't achieve peak integrity safety systems are intact, few functions
or functionality for several years. Metries can accelerate the matura- will be in bypass; and if our alarm
tion process and go on to help sustain peak performance over the
control system life. For older control systems, metrics can be used management has been effective, our
to assess their integrity and manage risk area improvement.
Control system metrics monitor fundamental control system alarm rates will be within operable limits.
health, like measuring vital signs of a person. When they are within
healthy limits, the system can be expected to continue to perform It's all about control system health, not control loop perfor-
reliably, but when they fall to risky levels, further investigation and mance. Control loop performance is only one of several metrics,
treatment are needed to bring risk back to acceptable levels. Con- and not an especially critical one at that, relative to system faults,
trol system metrics serve this role for control system health. safety functions, alarm rates, ete. But the common preoccupa-
Metries answer suppositions such as, if our regulatory controls tion with loop performance, especially when coupled with the
are sound, most control valves will be in automatic; if our safety first mistake about people, has killed many attempts to deploy
systems are intact, few functions will be in bypass; and if our alarm online metrics before ever getting beyond this first one. Simple
management has been effective, our alarm rates will be within and objective techniques to address this metric are included in
operable limits. the discussion below.
Extensive engineering efforts go into these areas, but sanity It's about saving time and resources, not consuming them.
checks such as these are often overlooked, setting the stage for Avoid deploying metrics that contribute little to control system
unwanted incidents or performance headaches to reveal the gaps. improvement while saddling personnel with application sup-
Metrics reveal the gaps proactively by gauging success in funda- port and manual data collection and reporting duties. Instead,
mental ways, regardless of the engineering approach (proven or deploy fully automated metrics utilizing existing DCS/historian
novel) employed. Once in place, metrics guard against long-term capabilities. This is inherently robust and effective. In addition,
performance degradation, which is another concern affecting automated and historized me tries provide their own benchmarks
many control system aspects. and trends provide essential insight for addressing shortfalls.
Control system metrics, a.k.a. key performance indicators Go gradually. Avoid attempts to engineer all the metries before
(KPIs), have been popular in recent years, bUt many initiatives all the lessons have been learned. Start with one or two metrics
have stalled due to some common mistakes. This article helps to address the biggest concerns. Build on success with additional
to avoid the mistakes, identifies guiding principles and provides metrics while keeping the original ones in place to sustain long-
several useful example metrics. term control system health. Overall, the strategy is to create a
minimal set of metrics, with each one representing some funda-
Principles and mistakes. Control system metrics are about mental aspect of control system integrity or functionality. The
the control system, not the people. Many efforts have derailed key metrics are:
due to concerns aboUt metrics reflecting on individual job per-
formance. In practice, nearly all metrics have shared responsibility Control valves in automatic: Not to be confused with
between engineering, operations and maintenance. And nearly control loop performance, this metric measures basic control valve
all "below-target" results pose business challenges, not individual asset utilization-are they under automatic control or are they in
ones. By selecting fundamental metrics and tackling risk areas manual and, therefore, failing to earn a real-time return on their
systematically, it stays about the control system, not the people. investment? Measuring higher levels of control, i.e., the trouble-
some comrolloop performance issue, has steeply diminishing applications that utilize flow data, whether directly (such as advanced
returns, while this simple metric answers 90% of the question. comro!) or indirectly via the historian (such as LPs, simulations,
This metric is traditionally implememed based on comroller efficiency monitoring and design or trouble-shooting activities).
mode (in this case, of only controllers directly attached to valves), The concept is to implement mass balance equations around
but is better implemented based on the actual output value-ifit is each vessel, plant and utility system. Although most facilities will
changing, the valve is in automatic; ifit is not changing, the valve is claim to have a firm handle on their mass balances, an online, drill-
in manual. This approach is more generic, captures valves that spend down, robust application with good closure remains an industry
time saturated and side-steps the sticky issue of "normal mode." rarity. This metric provides a positive path to this capability.
All DCSs since 1980 have built-in pressure and temperature
Mass balance closure: This metric amounts to "poor man's compensation functions for flow. These can be used to improve
data reconciliation," plus it works. It serves to improve flow meter- from an initial target closure of 3% to an optimum range of less
ing performance across a facility and contributes to the integrity of than 1%, an achievement most facilities could be proud of, espe-
cially in a robust online format.
TABLE 1. Summary of example metrics
Where flow measurements are missing,
Metric Target Objectives they can sometimes be estimated (for exam-
Control valves Target: >75% Indicates the condition and utilization of control valves and other final- ple, by a heat exchanger energy balance or a
in automatic control elements. valve sizing analysis). Or, the missing flow
Optimum: >90% A high number indicates reliable performance, good asset utilization can be calculated to close the mass balance.
from control and operation standpoints and high return on investment. In these cases, the quality is set to "fair" or
A low number indicates unreliable or unsuitable control valves or poor "poor" to bring attention to the missing
regulatory control design. measurement, which may be needed for
Target: 0
more rigorous offline data reconciliation or
Safety functions Indicates reliability and availability of safety functions.
in bypass loss-accounting calculations.
Safety functions in bypass: This metric
A high number indicates safety system degradation, usually due to
field instrument issues or conflicts between operating needs and doesn't assure that safety systems are techni-
safety system design. cally correct or built according to best practice,
A low number indicates successful safety system deployment and
but it assures that the expected safety functions
confidence that expected safety functions are available. are available. Safety functions have bypasses to
facilitate testing and repair, but often (though
Mass balance Target: <3% Indicates overall integrity of flow meters and mass balance data,
closure both current and historical.
usually incorrectly) bypasses are also used dur-
ing nonroutine operations, such as startup,
Optimum: <1 % A low number indicates reliable flow instruments and reliable flow data
for all data users.
shutdown or upsets. Or they may be kept in
bypass for long periods due to design or field
A high number indicates poor flow monitoring and introduces error in
instrument problems. This metric sheds light
related applications.
on such practices and helps address them.
System faults Target: 0 Indicates control system hardware and communications interface
It's a growing industry practice to com-
health from a systems standpoint.
pletely eliminate the use of bypasses except
A low number indicates good control system reliability with low risk
as imended (for testing or repair). This is
of unexpected failures.
being accomplished with a combination of
A high number indicates control system reliability problems with risk
more sophisticated safety function logic,
to availability.
startup permissives that (at least) minimize
Disabled alarms Site specific Indicates integrity of alarm design, alarm management process, and of
effective bypass time and more stringem
expectations for successful alarm-driven operation under normal and
administrative controls on the use of bypass
abnormal conditions.
...
switches. This metric provides necessary
Alarm rate Consult industry On-target numbers indicate good alarm integrity and realistic
information to manage these issues.
guidelines expectation of successful alarm-driven operation.
System faults: This metric, like safety
Peak rate Off-target numbers indicate alarm design and management
bypasses and mass balance closure, tends to
shortcomings. with risk to alarm-driven operation during normal or
be a historical problem area that responds
abnormal conditions.
well to visibility. Control system status dis-
Control loop Target: >75% Indicates degree of success with advanced regulatory control and
plays should be "all green", but in many
performance (if included) multivariable or other advanced control.
control systems, old and new, users live with
Optimum: >90% High numbers indicate successful advanced control programs.
standing system alarms. Each one represents
low numbers indicate difficulty in deploying and maintaining a risk, desensitizes users to the possibility
advanced controls. of compound faults and undermines the
Smart field Target: Based Indicates percent of smart transmitters and valves. obvious maintenance mission of keeping the
on annual goals system error free. This metric can serve as
A high number indicates progress toward capture of productivity and the starting point for hardware and system
reliability gains associated with a smart field. maintenance personnel on a daily basis.
A low number indicates control system intelligence ends at the rack Alarm-driven operation: These metrics
room with lost opportunity regarding improvements associated with are not alarm management, but they indicate if
a smart field. your alarm management program is working.
After integrity of safety systems and regulatory controls, alarm man- "MV utilization.'" At this point, few MPC practitioners still attach
agement represents the biggest opportunity for operational improve- much credibility to "service factor." HP
ment (if done correctly) and the biggest risk to plant availability and
preventable incidents (if done incorrectly), because about half of LITERATURE CITED
1 Kern, A. G., "Online moniroring of multivariable control utilization and
console operation is alarm-driven (as opposed to procedure-driven).
bener. t5," Hydrocarbon Processing, Ocrober 200 5.
If the alarm system is unhealthy, operation is compromised.
Alarm management has grown to include several schools of
thought, but the number of disabled alarms and the alarm rate r ~