11th IFAC Symposium on Dynamics and Control of

11th IFAC
Process
11th Symposium
IFACSystems, on
on Dynamics
including
Symposium and
and Control
Biosystems
Dynamics Control of
of
11th IFAC
Process
June Symposium
Systems,
6-8, 2016. on Dynamics
including
NTNU, and
Biosystems
Trondheim, NorwayControlonline
Available of at www.sciencedirect.com
Process Systems,
Process Systems, including
including Biosystems
Biosystems
June 6-8, 2016. NTNU, Trondheim, Norway
June 6-8, 2016. NTNU, Trondheim, Norway
June 6-8, 2016. NTNU, Trondheim, Norway
ScienceDirect
IFAC-PapersOnLine 49-7 (2016) 526–532
Control loop performance monitoring – ABB’s experience over two decades
Control
Control loop performance monitoring –– ABB’s experience over two decades
Control loop
loop performance
performance monitoring
monitoring
1
– ABB’s
ABB’s
1
experience
experience 3over
over two
two decades
decades
Kevin D. Starr1, Heiko Petersen1 and Margret Bauer3
Kevin
Kevin D. Starr 1, Heiko Petersen1 and Margret Bauer3
Kevin D. Starr ,, Heiko
D. Starr Heiko Petersen
Petersen andand Margret
Margret Bauer
1 1 3
1 Bauer
11
ABB Inc., Westerville, CO 80305 USA (Tel: +1 614 818 6394; e-mail: kevin.starr@us.abb.com).
1ABB Inc.,
and Westerville,
Information CO 80305 USA (Tel: +1 of
614the818 6394; e-mail:Johannesburg
kevin.starr@us.abb.com).
2
2
ABB
ABB Inc.,
School of ElectricalInc., Westerville,
Westerville, CO 80305
80305 USA
Engineering,
CO USA (Tel:
(Tel: +1
University 614
+1 of
614 818 6394;
6394; e-mail:
Witwatersrand,
818 kevin.starr@us.abb.com).
e-mail:Johannesburg 2050, South Africa (email:
kevin.starr@us.abb.com).
2School of Electrical and Information Engineering, University the Witwatersrand, 2050, South Africa (email:
2School of Electrical and Information Engineering, University
School of Electrical and Information Engineering, University of the Witwatersrand, Johannesburg 2050,
of the Witwatersrand,
margret.bauer@wits.ac.za)
margret.bauer@wits.ac.za)
Johannesburg 2050, South
South Africa
Africa (email:
(email:
margret.bauer@wits.ac.za)
margret.bauer@wits.ac.za)
Abstract: When process control systems do not perform as they should, plant personnel will not obtain
Abstract:
Abstract:
positive When
results process
may control systems do not perform as they should,
entirely.plant personnel will not obtain
Abstract:
positive Whenand
When
results
process
process
and may
even give
control
control
even
up using
systems
systems
give up
do
using
some
do not
not
some
control
perform
perform as loops
as
control
they
they should,
loops should, To increase
plant
plant
entirely. To
personnel productivity
will
will not
personnelproductivity
increase
and
not obtain
obtain
and
positive
efficiency results
it
positive resultsmustand
be
and may even
ensured
may eventhat give
that the
give up using
control some
system
up usingsystem control
is used
some control loops entirely.
effectively. The
loops entirely. To increase
best
To way to
increase productivity
do this is
productivityto and
turn
and
efficiency
efficiency
the automaticit
it must
must be
be ensured
controlensured to the
that
on andthat tunecontrol
the control system is
it correctly. is used
used
This effectively.
effectively.
paper The
Thethe
describes best
best way
way to
to do
expertise in this
do this is
is to
control turn
to loop
turn
efficiency it must be ensured the control system is used effectively. The best way to do this is to turn
the
the automaticmonitoring
control ongathered
and to by
tune it correctly.
control This paper describes the expertise in control loop
the automatic
performance control
control on
automaticmonitoring
performance
and
and to
ongathered tune
to by it
it correctly.
tuneABB’s
ABB’scorrectly.
control
This
This paper
experts.
paper
experts.
describes
With
describes
With more
the
the expertise
more automated
automated
in
in control
production
expertise
production
loop
processes
control loop
processes
performance
and fewer
performance monitoring
experts, control
monitoring gathered
loop
gathered by
by ABB’s
performance
ABB’s control
control experts.
monitoring (LPM)
experts. With more
needs
With to
more automated
be able
automated to production
evaluate
production one processes
hundred
processes
and
and fewer
at a experts, control
of loop performance monitoring (LPM) needs
doneto be able to evaluate one hundred
and fewer
loops
loopsfewer
at a
experts,
experts,
time
control
time instead
control
instead of
loop
loop performance
loop-by-loop
performance
loop-by-loop
monitoring
analysis.
monitoring
analysis. This
(LPM)
This cannot
(LPM)be
cannot
needs
needs
be done
to be
be able
towith
with able to
to evaluate
training
training
or manual
evaluate
or
one
manual
hundred
one tools.
hundred
tools.
In
In
loops
this at
loopspaper,a
paper, time
at a time instead
ABB’s
instead of loop-by-loop
perspective on
of loop-by-looploop analysis. This
performance
analysis. cannot be
monitoring
This cannot done
as
be done with
well training
as
with as several
training or manual
novel
ornovel
manual tools.
tools. In
features is
In
this
this paper, ABB’s
ABB’s perspective
perspective on
on loop
loop performance
performance monitoring
monitoring as well several features is
described. Finally,
this paper,Finally,
described.
further
ABB’s further research
perspective
research
directions
on directions are monitoring as well as several novel features is
highlighted.
loop performance
are highlighted.
as well as several novel features is
described.
described. Finally,
Finally, further
further research
research directions
directions are
are highlighted.
highlighted.
© 2016, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd.
Keywords: Control loop performance monitoring, PID control, industry automation, process control, All rights reserved.
Keywords: Control
Keywords:industry,
chemical loop
Controlpaper performance
loopindustry. monitoring,
performance monitoring,
monitoring, PIDPID control,
PID control, industry
control, industry automation,
industry automation, process
automation, process control,
process control,
control,
Keywords: Control loop performance
chemical industry,
chemical industry, paper
industry, paper industry.
paper industry.
industry.
chemical

1. INTRODUCTION Once a poorly performing control loop has been identified it

Once
is necessaryaa poorly
poorlytoperforming
diagnose the control
problemloop
loop and has been identified it
1. INTRODUCTION
1. INTRODUCTION Once
Once a poorly performing
performing control
control loop has pinpoint
has been the root
been identified
identified it
it
Process automation 1.regulates INTRODUCTION the dynamics of an industrial is
is
cause.
is
necessary
necessary
Ideally,
necessary
tothe
to
to
diagnose
diagnose
root
diagnose
the
the
cause
the
problem
problem
should
problem be and
and pinpoint
pinpoint
definite
and and
pinpoint
the root
the root
translated
the root
Process automation regulates the dynamics of an
an industrial
industrial cause.
cause. Ideally, the
Ideally, the root cause should
should be definite
definite andand translated
translated
production
Process
Process
production
process to
automation
automation
process
ensure the
regulates
regulates
to ensure the
safedynamics
safe
and efficient
dynamics
and efficient of anproduction.
of
efficient industrial
production.
into
cause.
into Ideally, the root
a maintenance
aa maintenance root cause
notification
cause should
notification and
be
and suggested
be definiteactions.
suggested and translated
actions.
production
One key process
aspect of to ensure
automation safe
is and
process control production.
which in into maintenance
into a maintenance notification
notification and suggested
and suggested actions.
production
One
One
industry key
key
process
aspect
aspect
today isof
of
to ensure
automation
automation
realised by
safe
is
is
PID
and
process
process
efficient
control. control
control
The
production.
which
which
majority in
in Often the degradation of performance occurs actions.
gradually. The
One
industry key today
aspectisofrealised
automation by is process
PID control. control
The which of
majority in
of Often
impact
Often the
theof degradation
degradation
degradation of
ofisperformance
not felt
performance occurs
straight
occurs gradually.
away
gradually.and The
the
The
industry
regulatory
industry today
today is
control
is realised
is
realised by
by PID
installed
PID control.
and
control. The
tuned
The majority
during
majority of
the
of Often
impact theof degradation
degradation ofisperformance
not felt occurs away
straight gradually.and The
the
regulatory
regulatory control
phase,is
control is installed
is, shortlyand
installed aftertuned
and tuned during
during the impact
process of
will degradation
continue is not
operation, felt
if straight
at decreased away and
and the
efficiency.
commissioning
regulatory
commissioning control
phase, isthat
that installed
is, shortly and
after tuned
the during the
the construction
construction
of
the
of
impact
process
process
However,will
of
will degradation
continue
continue
if process
is not felt
operation,
operation,
equipment if
if
fails
straight
at
at decreased
thedecreased
awayefficiency.
resulting efficiency.
downtime
the
commissioning
the production
commissioning phase,
facility. that
phase, that is,
These shortly
are
is, shortly after
hundreds the ifconstruction
not thousands
after theifconstruction of
of process
However, will
if continue operation, if at decreased efficiency.
the
the
of production
production
PID control facility.
facility.
loops. These
These
Many are
are
plants hundreds
hundreds
purchase if not
not
an thousands
thousands
automation However,
usually causes
However, if process
if process
process
equipment
equipment
dramatic cost. Many
equipment
fails
fails the
fails the
the
resulting
production
resulting
downtime
resulting companies
downtime
downtime
the production facility. These are hundreds if not thousands usually
usually
today move causes
causes dramatic
dramatic cost.
cost. Many
Many production
production companies
companies
of PID
of PID control
of PID control
package and, once
control
loops.
loops. Many believe
Many plants purchase
Many plants
in place,
loops.
purchase
it can run
plants purchase
an indefinitely,
an automation
an automation
automation usually causes towards
dramatic predictive
cost. Manymaintenance to avoid
production companies
package and, once in in place,
place, believe it itsystems
can runrun doindefinitely, today
today
unexpected move shut-downs.
move towards predictive
towards predictive
In predictive maintenance
maintenancemaintenanceto avoid
to avoid
the
package
without
package and,
and, once
intervention.
once in But
place, believe
control
believe it can
can run indefinitely,
not run
indefinitely, today
unexpected move towards
shut-downs. predictive
Inexample maintenance
predictive maintenanceto avoid the
without
themselves.
without intervention.
They
intervention.are But control
made
But control
up of systems do
computers,
systems docontrollers,
not run
not run unexpected
condition
unexpected of shut-downs.
equipment,
shut-downs. forIn predictive
Inexample
predictive controlmaintenance
loops,
maintenance the
sensors
the
without
themselves. intervention.
They are But control
made up of systems docontrollers,
computers, not run condition
condition
and final of
of equipment,
equipment,
control elements,for
for example
is monitored control
controland loops,
loops, sensors
sensors
maintenance
themselves.
monitors,
themselves. They
firmware, are made
software
They are made and up
and
up of of computers,
hardware.
computers, If, controllers,
for example,
controllers, condition
and final of equipment, for example control loops, sensors
monitors,
engineerfirmware,
monitors,
an ignores asoftware
firmware, software and
fan, a controller hardware.
hardware.
can overheatIf,
If, for
for example,
example,
and result and
actions
and final
finalarecontrol
control
carried out
control
elements,
elements,
according
elements,
is
is monitored
is monitored
to the results
monitored
and
and
and
maintenance
maintenance
before the fault
maintenance
monitors, firmware, software and hardware. If, for example, actions
actions are
are carried out according to the results before the fault
an
an
in
an
engineer
engineer
an
engineerentire ignores
ignores
portion
ignoresof
a
a
of fan,
fan,
a
a fan,
a
a
plant controller
controller
being
a controller shutcan
can
can
overheat
overheat and
overheat and result
down. and result
result
occurs.
actions
occurs. are carried out according to the results before the
carried out according to the results before the fault
fault
in
in an
an entire
entire portion
portion of a plant
aa plant being
being shut
shut down.
down. occurs.
occurs.
in an entire portion of plant being shut
The performance of regulatory control usually deteriorates down. Modern plant control systems integrate tightly with industrial
The performance Modern
Modern
equipment plant
plant
such control
control systems
systems
as motors andintegrate
drives ortightly
integrate tightly
sensors.with
withAsindustrial
industrial
a result,
over
The
The the lifetime of
performance
performance of regulatory
of
of a plant because
regulatory
regulatory
control
control usually
controlequipment
usually deteriorates
ages and
usually deteriorates
deteriorates Modern
equipment plant
such control
as systems
motors and integrate
drives ortightly
sensors.withAs industrial
over
over
process the
the lifetime
lifetime of
of a
a plant
plant because
because equipment
equipment ages and
ages and equipment
a control system
equipment such as
such as motorsfar
includes
motors and
andmoredrives
drives or sensors.
elements
or sensors. As aaabefore.
than ever
As
result,
result,
result,
over thedynamics
lifetime change.
of a plant In the courseequipment
because of time, continuous
ages and aaIf control
any
control system
area
system includes
falls short
includes far
or
far more
fails,
more elements
a
elements site than
will
than ever before.
experience
ever before.
process
process
plants dynamics
dynamics
are not used change.
change.
as In the
In
intended thebecause
course the
course of time,
of time,
product continuous
continuous
changes aIf control system includes far more elements than ever before.
process dynamics usedchange. In thebecause
course the of time, continuous If any area
any
production area falls
falls short
short or fails,
or fails,
thataaahas sitea direct
site will experience
will experience
plants
plants
or the are
plants
are
are
not used
not
production
not used
as
as
as
intended
intended
volume
intended because
varies.
becauseThisthe
thenewproduct
product changes
changes
objective
product changesof If any areaor
production or
quality
falls
quality
variability
short or fails,
variability that sitea direct
has
impact on
will experience
impact on
or
or the
the
increased production
production
flexibility volume
volume
in varies.
varies.
production is This
This new
new
addressed objective
objective
in of
of
initiatives production
plant
production or
performance.quality
or qualityThis variability
This makes
variability that
it
that has
even a
has amoredirect
more impact
critical
impactthat
directcritical on
on
or the production
increased flexibility volume
in varies.is This
production new objective
addressed in of
initiatives plant
plant
automationperformance.
performance.
control This
loops makes
makesare it
it even
even
checked moreand critical that
that
calibrated
increased
such
increased as flexibility
Industrie 4.0
flexibility in
or
in production
smart
production is
is addressed
manufacturing.
addressed in
in initiatives
initiatives plant
automationperformance.
control This makes
loops are it even
checked and more critical that
calibrated
such
such as
as Industrie 4.0 or smart manufacturing. automation
regularly. control
control loops loops are are checked
checked and and calibrated
such as Industrie
Deteriorating Industrie 4.0
4.0 or
overall
smart
smart manufacturing.
orprocess manufacturing.
performance leads to the need
automation
regularly.
regularly.
calibrated
regularly.
Cost pressures have led many producers to reduce process
Deteriorating
to assess
Deteriorating the overall
performance
overall process
process of performance
control loopsleads
performance and to
leads the
answer
to the need
the
Deteriorating overall process performance leads to the need
need Cost pressures
engineering staffhave
that led
mightmany producers
perform these to reduce
services process
to
to assess
assess the
following
to assess
the
the performance
performance of
questions:
performance
of control
of control loops
control loops
loops and and answer
answer the
and answer the
the
Cost
Cost pressures
pressures
engineering staff
have
have
that
led
led
might
many
many producers
producers
perform these
to
to reducein-house.
reduce
services
process
process
in-house.
following questions:
following questions: engineering
In many advanced
engineering staff that might
staff that might perform
economies these
perform workers services
these services in-house.
in industrial
in-house.
following questions: In
In many facilities
many
production advancedareeconomies
advanced economies
getting older, workers
workersand many in industrial
in industrial
years of
¥ Are the controllers performing as expected? In many facilities
advancedareeconomies workers in industrial
¥¥ Are Are the
the controllers
controllers performing
performing as as expected?
expected? production
production
process facilities
expertise is are
being getting
getting
lost to older,
older,
retirement. andAsmany
and many
a years
years
result, of
of
plant
¥ Are the controllers performing as expected? production
process facilities
expertise is are getting
being lost to older, andAsmany
retirement. years
aa result, of
plant
¥ Are the actuators and sensors doing their job? process
personnel expertise
is is being
responsible lost
for a to retirement.
wider range As
of moreresult, plant
complex
¥¥ Are the actuators and sensors doing their job? process
personnel expertise
is is being
responsible lost
for to retirement.
a wider range As a
of more result, plant
complex
Are
Are the actuators and
andinsensors doing
doing their
their job? personnel is
¥¥ Are the
thereactuators
any faults sensors
the process? job? tasks,
personnel
tasks, is responsible
reaching
reaching responsible
from
for
for aa wider
from operations
operations wider
to
range
range of
to maintenance
maintenance of more
more
and
complex
and including
complex
including
¥¥ Are there any faults in the process? tasks,
control,
tasks, reaching
process
reaching from
from operations
engineering
operationsand to
to maintenance
software
maintenance and
solutions.
and including
including
¥¥ Are Are
Is any
there
there any
any faults
equipment faults in
in the
the process?
process?
malfunctioning?
control,
control,
control,
process
process
process
engineering
engineering
engineering
and
and software solutions.
and software
software solutions.
solutions. skill set,
¥¥ Is any equipment malfunctioning? Monitoring control loops requires a sophisticated
¥ Is Is any
any equipment
equipment malfunctioning?
malfunctioning? Monitoring
including
Monitoring
Monitoring
control
the
control
control
loops
fundamentals
loops requires
loops requires
requires
a
of control,sophisticated
aa sophisticatedknowledge
sophisticated
skill
skill set,
on
skill set,
set,
including the
including the fundamentals of control,
control, knowledge on
including the fundamentals of control, knowledge on
fundamentals of knowledge on
Copyright
2405-8963 ©© 2016,
2016 IFAC 526Hosting by Elsevier Ltd. All rights reserved.
IFAC (International Federation of Automatic Control)
Copyright
Peer review©
Copyright 2016
©under IFAC
2016 responsibility
IFAC 526
of International Federation of Automatic
526Control.
Copyright © 2016 IFAC
10.1016/j.ifacol.2016.07.396 526
IFAC DYCOPS-CAB, 2016
June 6-8, 2016. NTNU, Trondheim, NorwayKevin D. Starr et al. / IFAC-PapersOnLine 49-7 (2016) 526–532
electrical and process equipment as well as detailed process
understanding. Today and in the future there will be fewer
trained personnel. Thus, the diagnostic results of control loop
performance analysis have to be generated automatically and
must be easy to understand.
In this paper we will investigate loop performance
monitoring (LPM) from an industrial perspective and
describe ABB’s workflow and a solution called ServicePort.
The relationship to other performance measurement areas in
production such as maintenance, asset management and
efficiency studies are drawn in Sec. 2. Sec. 3 describes some
of the novelties in LPM that have been implemented in
ABB’s solution.
1.1 Loop performance monitoring literature
Systematic assessment and monitoring of the control loop
performance has attracted interest in the academic
community since the advent of digital control in process
plants. General problems of control loops were reported in
the 1990s (Bialkowski, 1993; Ender, 1993). Indices to assess
the performance of control loops were developed, famously
by Harris (1989) and addressing data signatures such as
oscillations (Hägglund, 1995; Thornhill and Hägglund, 1997)
as well as common hardware problems such as valve stiction
(Horch, 1999).
Over the following decades, data-driven and model-based
algorithms were developed and several reviews have been
published that provide a comprehensive list of methods in
control loop performance monitoring for single and
multivariate control loops (Qin, 1998; Harris et al., 1999;
Jelali, 2006). Overall, a significant amount of research effort
has gone into the development of methods, partly using data
from industrial problems of poorly performing loops and
partly by using simulated data from continuous stirred tank
reactor or the Tennessee Eastman Process which allows the
simulation of a complete chemical process (Downs and
Vogel, 1993).
1.2 Big Data and Data Analytics
In the process industry, capturing, storing and maintaining
data is primarily associated with cost. The market research
organisation Gartner has coined the term ‘dark data’1 when
discussing operational technology (OT) in comparison to
information technology (IT). Dark data is defined as ‘the
information asset organizations collect, process and store
during regular business activities, but generally fail to use for
other purposes’. Process data in particular has many potential
applications but is inherently underused.
Data analytics retrieve information and turn data into value.
Loop performance monitoring is one of very few areas where
process data is analysed and turned into information that
improves productivity and process efficiency. The results of
LPM give an indication about the state of the process control
1
http://www.gartner.com/it-glossary/dark-data
527
527
system. The key is that no extra sensors need to be installed
and no extra cost in terms of data capture occurs. Instead, the
existing dark data is turned into value.
In the age of Industrie 4.0 and big data it is important to
ascertain the data ownership. Increasingly, data is seen as an
asset rather than a liability. In the automotive industry, car
manufacturer compete with the car owner, insurance
companies and app developers for the data generated when
driving the car, in particular for the geospatial positioning. In
the process industry at present, the data belongs to the
production company where the data originates. The
importance of data ownership, privacy and security increases
dramatically.
2 INDUSTRIAL CONTROL LOOP PERFORMANCE
MONITORING
ABB has been active in the area of loop performance
monitoring for at least two decades. The efforts that were
previously distributed across the ABB organization
worldwide were consolidated in an effort to take the best out
of all solutions and combine it into one offering that can be
easily installed by process control experts and non-experts. In
2013, the ABB product ServicePort was launched. Within the
first 6 months 70 installations were sold.
Here, we will explain the concept of the solution and
highlight the novelties. In total, six patents were granted to
protect the intellectual property (Starr, 2003; Tran et al.,
2007, Trans and Mast, 2009; Starr et al., 2013, Murphy et al.,
2014; Mast and Starr, 2014.). Further nine patents are
pending.
In recent years, the notion that LPM is a software product has
given way to the concept of supplying LPM as a service
(Bauer et al., 2016). The difference is that service has a large
consulting aspect providing advice on risk assessment and
priority settings. It requires a change in mind-set.
Fig. 1: ABB Senior Service Engineer uses ServicePort’s
built-in retractable user interface to analyse Key Performance
Indicators (KPI) on customer equipment and processes.
IFAC DYCOPS-CAB, 2016
June 6-8, 2016. NTNU, Trondheim, Norway
528 Kevin D. Starr et al. / IFAC-PapersOnLine 49-7 (2016) 526–532
Table 1. Delivery schedule of Loop Performance Fingerprint.
Day 1 Day 2
Project ¥ Set up data collection software
introduction ¥ Begin collecting security data
meeting ¥ Interview key plant personnel
¥ Check data and make configurations accordingly
LPM is often treated as a software solution only. ServicePort
contains a hardware component that is located on site for
proximity to the process and security reasons, see Fig.1. The
service engineer can access the data on site via a retractable
user interface. In a first step, ServicePort collects data from
the process control system. Once configured, the data
gathering occurs automatic and secure. In a second step the
data is analysed on an ongoing basis and stored on the same
hardware. The data as well as analytic results can be accessed
both on site and remotely by either plant personnel or ABB
technical support. A remote access platform (RAP) provides
a detailed audit trail of anyone connecting to ServicePort by
logging each remote session.
Once the data has been captured, advanced service ‘channels’
address equipment, process and industry issues. Here,
‘channel’ refers to a different aspect of data analytics. The
most popular channels are control loop performance
monitoring and control loop tuning. Other channels include
data analytics for cyber security, low voltage drives and the
automation system. Data analytics for the automation system
identifies software, hardware, system and network
performance irregularities. For example, there may be a
module bus failure limiting the communication of the control
system.
In the remainder of this paper we will focus on the
functionality of the LPM. However, similar procedures are
applied to other channels. Also, the number of channels
continues to grow as more data analytic functionalities are
added to the offering. In the following, we discuss the
procedure, views and workflow of ServicePort for LPM. We
then highlight what has been taken over from previous ABB
research effort and what is new within ABB and in the LPM
community.
2.1 Loop Performance ‘Fingerprints’
An initial analysis referred to as ‘fingerprint’ is proposed as
an entry analysis for process plants which have previously
not assessed their control loop performance. The loop
performance fingerprint is used to benchmark current control
loop performance, identify issues that reduce control
performance and outline and recommend an improvement
plan. Control loop data analysis is delivered as a consulting
service by trained ABB engineers and identifies troublesome
loops through data collection, model identification, feedback
tuning, feed-forward tuning and controller simulation. A
typical schedule for a fingerprint consulting service is shown
in Tab. 1. The first day is used to introduce the concept and
methodology of LPM to the plant personnel. During Day 2
528
Day 3 Day 4-5
¥ Complete process ¥ Prepare summary
data collections of findings and
¥ Data analysis begins finalize report
¥ Exit meeting ¥ Present findings
the data collection is prepared and key plant staff such as
operators, process engineers and automation experts are
interviewed. During the third day the data collection is
completed and the data analysis is conducted. In the
following two days a report is generated and the findings are
presented and discussed with the plant personnel.
2.2 Loop performance workflow
After the initial fingerprint assessment the performance
should be maintained to ensure that the results are consistent
and do not degrade over time. This implementation includes
regular reports and updates that are aligned with plant
personnel meetings. ServicePort provides three different
support methods for the continuous monitoring of the control
loop performance.
View – The view option allows the engineer to view the raw
data used for data analysis. The data can be selected by
equipment or process. Raw data is informative to the
experienced user who can judge from the nature of the data
which root causes are likely. The expert user can conduct this
analysis probably more reliably than any algorithm. Key is
that the routines need to be automated since the expert user
does not have time to inspect each variable individually.
Scan – Data analysis results are presented in the scan view.
All analysis results are expressed as key performance
indicators (KPIs). The summary of these KPIs is ranked by
severity so that the user can begin addressing issues in order
of priority. All methods are configured automatically and the
user is not required to set any parameters.
Track – KPIs can be managed so that rules defined for each
KPI determine when an alert should be sent out. The user sets
his or her own thresholds for the KPIs but default thresholds
are suggested. If the threshold is exceeded, alerts are sent out
by email or SMS.
2.3 Loop performance assessment results
Presently, there are about 600,000 PID loops assessed daily
with ServicePort. ABB conducted an analysis of control
systems used by customers in multiple industries that
revealed the following finding, see Fig. 2. In the investigated
plants over the 600,000 loops, the control performance
distribution is 30% manual operation, 30% increasing
variation, 25% improving production and 15% output out of
actuation range. This means that up to 75% of a typical
plant’s automation investment is not providing benefits.
Instead of solving process problems, the systems are not
IFAC DYCOPS-CAB, 2016
June 6-8, 2016. NTNU, Trondheim, NorwayKevin D. Starr et al. / IFAC-PapersOnLine 49-7 (2016) 526–532
Figure 2. Loop performance assessment results of data
analysed by ABB’s ServicePort.
operating as they were designed, due to the lack of
comprehensive tuning. This indicates that many control
system owners are under-utilizing their automation
investments.
3 NOVEL LOOP PERFORMANCE CONCEPTS
Loop performance monitoring is, after more than two
decades, an established discipline. Before looking into what
is new in ABB’s ServicePort implementation we define the
aspects that are not new.
Loop performance monitoring methods: As detailed in the
literature review in Sec. 1.1, there are several established
indices to assess the performance of a controller and detect
unwanted disturbances such as oscillations and nonlinearity
in the time trend. Methods as reported in the literature (see
for example the review by Jelali, 2013) are implemented.
These range from simple algorithms such as detecting
whether a loop is in manual to more complex computations
detecting intermittent process disturbances.
Rating of control loop performance quality: The performance
of control loop is not a binary value of good or bad but
instead it is a grey area. For the ease of interpretation the
loops are usually grouped into categories such as good, fair,
poor and excellent. Paulonis and Cox (2003), for example,
define a ‘fair’ performing control loop in detail: “Loops in
this category are not performing up to potential and should be
improved. Control is probably being maintained in a broad
sense, but deviation setpoint is likely to be degrading process
performance. These loops should be investigated further.
Often, the problem is not difficult to find and improvement
Figure 3. Categories of LPM reported problems by origin in the feedback loop.
529
529
can be obtained without significant effort. Occasionally, level
loops with a flow smoothing objective which are performing
adequately end up in this category. (Score range 20-40).”
There are several novel aspects to ABB’s ServicePort which
are subject of several patents. Two of these aspects are
explained in the following section and are embedded in the
workflow as described in Sec. 2.2.
3.1 Loop performance monitoring as KPIs
Initial implementations of LPM calculated loop indices and
presented them to the LPM software user. This was found to
be a major hindrance to the roll-out of the solution because
the user where not always trained to interpret the results
correctly. Even if a performance index was scaled to be
between 0% and 100% the user was uncertain whether 60%
was a sufficiently good result or whether the loop required
attention.
A shift in mind-set occurred when it became clear that loop
performance indices should be expressed in terms of the
process and not in terms of control. All reported LPM results
should have a diagnostic capability so that personnel who are
not LPM experts can understand the result. A first step is that
the indices are grouped into three categories according to the
underlying problem: Control, process and signal processing,
see Fig. 3. Arguably, the most important category relates to
the process since the issues flagged in this category can
trigger a maintenance action.
A loop performance monitoring result should be expressed as
a key performance indicator (KPIs). KPIs are understandable,
measurable, and can be monitored against a target. They are
assessed periodically and measures to improve the KPIs are
suggested. It is important to note that indices tend to be
practical loop performance indices. An example of a KPI is
the number of loops in manual, the final control element out
or range or compression of the measured variable.
Fig. 4 shows a screenshot of the KPI dashboard in ABB’s
ServicePort. The top graph shows an accumulated view of all
KPIs in the three categories – control, process and signal
processing – over the past five days from Thursday the 9th to
Monday the 12th. The data range can be selected in the
dialog box below. The three categories are further detailed in
the three plots underneath in Fig. 4. This shows that the most
common causes in each category are loops in manual (17%),
final control element (FCE) out of range (36%), and step out
or quantisation (17%) respectively. Clicking on each bar
provides a drill down analysis and will show the individual
loops with the issues.
IFAC DYCOPS-CAB, 2016
June 6-8, 2016. NTNU, Trondheim, NorwayKevin D. Starr et al. / IFAC-PapersOnLine 49-7 (2016) 526–532
530

Figure 4. Screenshot of ABB’s ServicePort loop performance monitoring channel.

Figure 5. Contour map plotting final control element out of range problem over time.

530
IFAC DYCOPS-CAB, 2016
June 6-8, 2016. NTNU, Trondheim, NorwayKevin D. Starr et al. / IFAC-PapersOnLine 49-7 (2016) 526–532
3.2 Contour map
Identifying control loop problems is equally important as
tracking the progress of an error. In this new analysis report,
KPIs are therefore plotted over time in a contour map. An
example of such a contour map is shown in Fig. 5. The x-axis
is the time axis divided into eight hour intervals or operator
shifts. The On the y-axis are all control loops in the plant
listed and represented by a horizontal bar. There are roughly
270 in the example. The shading of the bar indicates whether
there is a fault present in the loop – dark areas indicate that
the severity is high while white areas indicate that not fault is
present.
The contour map not only allows to view the development of
faults in each loop but also the identification of clusters of
loops. Also, there are periods of plant upset where all control
loops show problems. These manifest themselves as vertical
lines that pervade most of the process. The visual inspection
of the contour plot gives an immediate picture of the plant
health over time. Further developments will look into
automatic detection of clusters.
4 RESEARCH DIRECTIONS
In our view, there are two main areas that warrant the
investment of more research effort. The first relates to the
indices on their own and the second to the various aspects of
plant performance of which loop performance assessment is
only one.
First, many indices to monitor the performance of control
loops have been developed. When implementing these
indices in an industrial solution these indices have to fulfil
certain criteria:
¥ Suitable as KPIs: Performance indices have to be
meaningful so that a non-expert LPM user can interpret
them. Indices should reflect the performance of the
process rather than of the control strategy.
¥ Reliability: Indices should give a 100% reliable result that
does not warrant discussion. Particularly damaging to the
credibility are false alarms. In general, a loop
performance index should rather have false negatives than
false positives.
¥ There should be no ‘tuning’ required to the index, that is,
no parameterization. Parameters need to be proposed by
guidelines. In reality, if there are parameters to set, the
index will not be used.
¥ Standard categories (good, fair, poor) should enable the
comparison of different indices, for example see NE 152
published by NAMUR (2014).
¥ An un-biased comparison of indices assessing the same
issues using standardised sets of industrial data would be
helpful.
¥ To our knowledge there is no reported index measuring
quanitization of the measured variable which is due to
senor inaccuracies.
531
531
The second area of future research concerns the integration of
different aspects relating to plant performance, in particular
the following
¥ Integration between loop performance monitoring and
loop tuning. Often, the two areas are implemented
unrelatedly but synergies can be achieved if the two
concepts are combined.
¥ Automated procedure for plant-wide disturbance analysis.
There have been several attempts to track the root cause
that has travelled along the product stream or utilities in a
process. However, root cause analysis is still a manual
task in comparison to LPM analysis.
¥ Inter-connection between different aspects of plant
performance monitoring, or ‘inter channel causality’. An
advantage of software solutions such as ABB’s
ServicePort is that information from different solutions is
stored in one location. We often think in silos when it
comes to asset health. However a drive can impact a
control can impact an alarm limit. A full hard drive can
stall a controller, and result in failed KPIs. Being able to
apply cause and effect to channel KPIs would be very
helpful with regards to troubleshooting.
5 CONCLUSIONS
This paper looked at loop performance monitoring from an
industrial implementation and solution viewpoint. Today,
loop performance monitoring is seen more holistically and
combined with the monitoring of physical equipment or
assets. The terminology ‘loop health’ and ‘asset health’ is
used to express the tight coupling between control loop
monitoring and plant maintenance.
When identifying high and low priorities of maintenance it is
useful to draw on the topics ‘circle of concern’ and ‘circle of
influence’ to gain ownership of actions (Covey, 2000). The
circle of concern is comprises a wider range of concerns a
plant manager has – from managing people to ensuring
safety, quality and maximum profit. The circle of influence,
on the other hand, comprises only the concerns that can be
acted on. For example, a plant manager can decide to produce
more of a certain product but will not be able to change the
price of crude oil. Focusing on the circle of influence allows
prioritizing the issues in the plant that are of concern but also
in the region of influence. Thus, the circle of influence can be
expanded and the plant performance can be improved
holistically.
As automation equipment grows, the need for a person to
utilize technology to cast a larger net is also needed. In the
past the manual efforts of physical inspection was the norm
for performing asset health evaluations. Today, the
expansion of automation makes the traditional efforts much
less effective. Today, rather than evaluate one loop at a time,
we need the ability to evaluate one hundred loops in the time
it took to do one. This cannot be done with training or
manual tools. Technology has to help us with that. This is
why ABB has invested in ServicePort. It provides a platform
for automation users to be more efficient, identify incidents
IFAC DYCOPS-CAB, 2016
June 6-8, 2016. NTNU, Trondheim, NorwayKevin D. Starr et al. / IFAC-PapersOnLine 49-7 (2016) 526–532
532

in record time, and provide the building blocks to establish Thornhill, N. F., & Hägglund, T. (1997). Detection and
predictive notification. In so doing, a user can move from a diagnosis of oscillation in control loops. Control
reactive service where waiting for operations to identify a Engineering Practice, 5(10), 1343-1354.
failure occurs to a proactive solution where operations never Tran, P. Q., Starr, K. D., and Mast, T. A., 2007. U.S. Patent
see an automation failure. No. 7,300,548. Washington, DC: U.S. Patent and
Trademark Office.
REFERENCES Tran, P. Q., and Mast, T. A., 2009. U.S. Patent No.
7,584,013. Washington, DC: U.S. Patent and Trademark
Bauer, M., Horch, A., Xie, L., Jelali, M. and Thornhill, N.F.,
Office.
(2016). The current state of control loop performance
NAMUR Recommendation NE152, 2014. Controller
monitoring – A survey of application in industry. To
Performance Management: Monitoring and Optimisation
appear in Journal of Process Control. of Regulatory Control in Production Plants. Leverkusen,
Bialkowski, W. L. (1993). Dreams versus reality: a view Germany. www.namur.net.
from both sides of the gap: manufacturing excellence
with come only through engineering excellence. Pulp &
Paper Canada, 94(11), 19-27.
Covey, S.R., 2000. The 7 Habits of Highly Effective People:
Powerful lessons in Personal Change. Simon &
Schuster, New York.
Downs, J. J., & Vogel, E. F. (1993). A plant-wide industrial
process control problem. Computers & Chemical
Engineering, 17(3), 245-255.
Ender, D. B. (1993). Process control performance: Not as
good as you think. Control Engineering, 40(10), 180-
190.
Hägglund, T. (1995). A control-loop performance monitor.
Control Engineering Practice, 3(11), 1543-1551.
Harris, T. J., Seppala, C. T., & Desborough, L. D. (1999). A
review of performance monitoring and assessment
techniques for univariate and multivariate control
systems. Journal of Process Control, 9(1), 1-17.
Harris, T. J. (1989). Assessment of control loop performance.
The Canadian Journal of Chemical Engineering, 67(5),
856-861.
Horch, A. (1999). A simple method for detection of stiction
in control valves. Control Engineering Practice, 7(10),
1221-1231.
Jelali, M. (2006). An overview of control performance
assessment technology and industrial applications.
Control Engineering Practice, 14(5), 441-466.
Jelali, M. (2013). Control performance management in
industrial automation. Advances in Industrial Control,
DOI, 10, 978-1.
Murphy, T. F., Starr, K., and Mast, T. A., 2014. U.S. Patent
No. 8,630,728. Washington, DC: U.S. Patent and
Trademark Office.
Mast, T. A., and Starr, K. D., 2014. U.S. Patent No.
8,702,908. Washington, DC: U.S. Patent and Trademark
Office.
Paulonis, M. A., & Cox, J. W., 2003. A practical approach
for large-scale controller performance assessment,
diagnosis, and improvement. Journal of Process Control,
13(2), 155-168.
Qin, S. J. (1998). Control performance monitoring—a review
and assessment. Computers & Chemical Engineering,
23(2), 173-186.
Starr, K. D., 2003. U.S. Patent No. 6,597,958. Washington,
DC: U.S. Patent and Trademark Office.
Starr, K. D., Mast, T. A., and Garverick, R. T., 2013. U.S.
Patent No. 8,489,209. Washington, DC: U.S. Patent and
Trademark Office.

532