9th International Symposium on Advanced Control of Chemical Processes

9th International Symposium on Advanced Control of Chemical Processes

9th
JuneInternational Symposium
7-10, 2015. Whistler, on Advanced
British Columbia,Control
Canadaof Chemical Processes
9th
9th International
JuneInternational Symposium
7-10, 2015. Whistler,
Symposium on
on Advanced
British Columbia,
Advanced Control
Canada
Control
Available of
of Chemical
online Chemical Processes
Processes
at www.sciencedirect.com
June 7-10, 2015. Whistler, British Columbia, Canada
June 7-10, 2015. Whistler, British Columbia, Canada
June 7-10, 2015. Whistler, British Columbia, Canada

ScienceDirect
IFAC-PapersOnLine 48-8 (2015) 451–456
Monitoring
Monitoring Safety
Safety of
of Process
Process Operations
Operations Using
Using Industrial
Industrial Workflows
Workflows
Monitoring
Monitoring Safety of Process Operations Using Industrial Workflows
Monitoring Safety
Safety of
of Process
Process
,1
Operations
Operations Using
Using Industrial
Industrial Workflows
Workflows
Sridhar Dasani*,
Sridhar Dasani*, Sirish
Dasani*, Sirish L.
Sirish L. Shah*,1,1,1,,, Tongwen
Shah*
L. Shah* Tongwen Chen**,
Chen**, Jay
Jay Funnell
Funnell and
and Robert
Robert W.
W. Pollard***
Pollard***
Sridhar
Sridhar Tongwen Chen**,
 Jay Funnell and Robert W. Pollard***
Sridhar Dasani*,
Dasani*, Sirish
Sirish L. Shah* ,, Tongwen
L. Shah* ,1
Tongwen Chen**,
Chen**,


Jay
Jay Funnell
Funnell and
and Robert
Robert W.
W. Pollard***
Pollard***
*Department of Chemical 
and
 Materials Engineering,
*Department
*Department of
of Chemical
Chemical and
and Materials
Materials Engineering,
Engineering,
**Department
*Department
*Department
**Department of
ofof
of Electrical
Chemical
Chemical
Electrical and
and
and
and Computer
Materials
Materials
Computer Engineering,
Engineering,
Engineering,
Engineering,
**Department
University of
**Department of
Alberta,
of Electrical
Electrical and
Edmonton,
and Computer
AB, Engineering,
Canada
Computer T6G 2G6
Engineering,
**Department
University
University of
of of
Alberta,
Alberta, Electrical and
Edmonton,
Edmonton, Computer
AB,
AB, Engineering,
Canada
Canada T6G
T6G 2G6
2G6
***Utilities
***Utilities Department,
University
University of
Department, University
Alberta,
ofUniversity of
Alberta, Edmonton,Alberta,
Edmonton,
of Edmonton,
AB,
AB,
Alberta, Canada
Edmonton, AB,
T6G
Canada AB, Canada,
2G6
T6GCanada,
2G6 T6G
T6G 2H1
2H1
***Utilities
***Utilities Department, University of Alberta, Edmonton, AB, Canada, T6G 2H1
***Utilities Department,
Department, University
University of of Alberta,
Alberta, Edmonton,
Edmonton, AB,
AB, Canada,
Canada, T6G
T6G 2H1
2H1
Abstract:
Abstract: An
An industrial
industrial workflow
workflow represents
represents aa sequence
sequence of
of tasks
tasks or
or actions
actions that
that describes
describes an
an operational
operational
Abstract:
procedure.
Abstract: An
In
An industrial
this paper,
industrial workflow
workflow
workflow represents
strategies
represents a sequence
are of
developed
aa sequence of tasks
tasks or
that
or actions
capture
actions that describes
operational
that describes an operational
knowledge
an by
operational
Abstract:
procedure.
procedure. An
In
In industrial
this
this paper,
paper,workflow
workflow
workflow represents
strategies
strategies sequence
are
are of
developed
developedtasks or
that
that actions
capture
capturethat describes
operational
operational an operational
knowledge
knowledge by
by
analyzing
procedure. event
In
procedure. event
analyzing In thislogs
this of
paper,
paper,
logs of how an
workflow
workflow
how an operator has
strategies
strategies
operator has executed
are a
are developed
executed procedure
developed that
aa procedure while
capture
that capture
while controlling
operational
operationalor
controlling or monitoring
knowledge
knowledge
monitoring by
byaaa
analyzing
process.
analyzing Inevent
this
event logs
work,
logs of
of how
our
how an
prime
an operator
focus
operator is has
on
has executed
infrequent
executed a procedure
process
procedure while
operations
while controlling
such as
controlling or
plant
or monitoring
start-up
monitoring and aa
analyzing
process.
process. In
Inevent
this
this logs ofour
work,
work, how
our an operator
prime
prime focus
focus is
is has
on
on executed aprocess
infrequent
infrequent procedure
process while controlling
operations
operations such
such as
as or monitoring
plant
plant start-up
start-up and
and
shutdown
process. procedures.
In this work,
process. Inprocedures.
shutdown this work,We We
ourpropose
prime
ourpropose a
prime focus workflow
focus is on conformance
infrequent
is on conformance
aa workflow method
process
infrequent process to continuously
operations
methodoperations
to such
such as
continuously monitor
as plant and compare
start-up
plant and
monitor start-up and
and
compare
shutdown
operator
shutdown procedures.
actions
procedures. We
We propose
with standard
standard
propose aa workflow
operating conformance
procedures
workflow (SOPs) method
conformance to
and identify
method continuously
identify
to procedural monitor
continuously and
violations
monitor and compare
that could
compare
shutdown
operator
operator procedures.
actions
actions with
with We propose
standard operatingworkflow conformance
procedures (SOPs) method
and to continuously
procedural monitor and
violations compare
that could
compromise
operator
compromise process
operator actions
actions withsafety
withsafety
process and operating
standard
standard
and efficiency.procedures
operating
operating
efficiency. An industrial
procedures
An
(SOPs)
industrial
procedures case and
(SOPs)
(SOPs)
case studyidentify
and
and
studyidentify
identify
is
procedural
is presented
presented
procedural
procedural
to
violations
to illustrate
illustrate
violations
violations
that
that could
applications
that couldof
could
applications of
compromise
workflow
compromise process
conformance
process safety
safety and
and efficiency.
monitoring to
efficiency. An
identify
An industrial case
operational
industrial case study is
problems
study is presented to
associated
presented to illustrate
with
illustrate applications
human of
factors,
applications of
compromise
workflow
workflow process safety and
conformance monitoring efficiency.
to An
identifyindustrial case
operational study is
problemspresented to
associatedillustrate
with applications
human of
factors,
workflow
process andconformance
process and
workflow instrumentation.
conformance
conformance
instrumentation.
monitoring
monitoring
monitoring
to
to
to
identify
identify
identify
operational
operational
operational
problems
problems
problems
associated
associated
associated
with
with
with
human
human
human
factors,
factors,
factors,
process
process and
and instrumentation.
instrumentation.
process
©
Keywords:and instrumentation.
2015, IFACStandard
(International Federation
operating of Automatic
procedures, Process Control)
safety,Hosting by Elsevier
Workflows, EventLtd. All rights
logs, reserved.
Workflow mining,
Keywords:
Keywords: Standard
Standard operating
operating procedures,
procedures, Process
Process safety,
safety, Workflows,
Workflows, Event
Event logs,
logs, Workflow
Workflow mining,
mining,
Conformance
Keywords:
Keywords: monitoring,
Standard
Standard Process
operating
operating monitoring.
procedures, Process safety, Workflows, Event logs, Workflow mining,
Conformance
Conformance
Conformance
monitoring,
monitoring, Processprocedures,
Process monitoring.Process safety, Workflows, Event logs, Workflow mining,
monitoring.
Conformance monitoring,
monitoring, Process
Process monitoring.
monitoring. 



1.
1. INTRODUCTION
INTRODUCTION
1.
1. INTRODUCTION
In today’s 1. INTRODUCTION
competitive INTRODUCTION
global market, some of the key
In
In today’s
today’s competitive
competitive global
global market,
market, some
some of
of the
the key
key
operational
In
In today’s
today’s
operational constraints
competitive
competitive
constraints confronting
global
global
confronting the
market,
market,
the process
some
some
process industry
of
of the
the
industry are
key
key
are
operational
shortage
operational of constraints
skilled
constraints confronting
workforce
confronting and the
the process
procedural
process industry are
violations
industry are
operational
shortage
shortage of
of constraints
skilled
skilled confronting
workforce
workforce and
and the process
procedural
procedural industry are
violations
violations
during
shortage
shortage
during normal,
of
of
normal, plant
skilled
skilled
plant start-up
workforce
workforce
start-up and
and shutdown
and
and
shutdown operations
procedural
procedural
operations which
violations
violations
which
during
result normal, plant start-up and shutdown operations which
duringin
during
result
result in
in
process
normal,
normal,
process
process
upsets,
plant
plant thereby
start-up
start-up
upsets,
upsets, thereby
thereby andnegatively
and shutdownimpacting
shutdown
negatively
negatively operationsprocess
operations
impacting
impacting
which
which
process
process
performance.
result
result in
in process
process
performance. Operational
upsets,
upsets,
Operational thereby
thereby inconsistencies
negatively
negatively
inconsistencies during
impacting
impacting
during grade
process
process
grade
performance.
transitions
performance. Operational
introduce variability
Operational inconsistencies
that affects
inconsistencies during
product
during grade
quality
grade
performance.
transitions
transitions Operational
introduce
introduce variability
variability inconsistencies
that
that affects
affects during
product
product grade
quality
quality
and financial
transitions
transitions
and financial performance
introduce
introduce of
variability
variability
performance of a
a company.
that
that
company. affects
affects product
product quality
quality
and
and financial
financial performance
performance of
of a company.
and
Basedfinancial
on 18 performance
plant studies ofinaa the
company.
company.
U.S, Canada, and Europe,
Based
Based on
on 18
18 plant
plant studies
studies in
in the
the U.S,
U.S, Canada,
Canada, and
and Europe,
Europe,
the
Based
Based
the Abnormal
on
on 18
18
Abnormal Situation
plant
plant studies
studies
Situation Management
in
in
Management the
the U.S,
U.S, (ASM)
Canada,
Canada,
(ASM) consortium
and
and
consortium has
Europe,
Europe,
has
the
the Abnormal
reported that
Abnormal Situation
40%
Situation ofManagement
abnormal
Management (ASM) consortium
situations
(ASM) in
consortium has
process
has
the Abnormal
reported that Situation
40% ofManagement
abnormal (ASM) consortium
situations in has
process Fig. 1.
Fig. 1. Sources
Sources of of abnormal
abnormal events
events in in process
process industry
industry
reported
operations
reported that
are
that 40%
directly
40% of
of abnormal
attributed
abnormal to situations
people
situationsand in
in process
workplace
process Fig. 1. Sources of abnormal events in process industry
reported
operations
operations that
are
are 40%
directly
directly of abnormal
attributed
attributed to
to situations
people
people and
and in process
workplace
workplace Fig.
Fig. 1.
1. Sources
Sources of
of abnormal
abnormal events
events in
in process
process industry
industry failure
settings
operations
operations
settings as
as shown
are
are
shown in Figure
directly
directly
in 11 (ASM
attributed
attributed
Figure (ASM to Consortium,
to people
people and 2007).
and workplace
Consortium, workplace
2007). In
In that hurt
that hurt process
process safety.
safety. CCPS
CCPS metrics
metrics for
for measuring
measuring failure
settings
the as
context
settings as shown
of
shown in
ASM
in Figure 1
Consortium
Figure (ASM
11 (ASM Consortium,
research,
Consortium,an 2007).
abnormal
2007). In
In that
to hurt
hurt process
follow
thatfollow procedures
process safety.
is CCPS
safety. definedmetrics
CCPS as “the
metrics for
“the measuring
measuring failure
percentage
for percentage of the
the
failure
settings
the
the as
context
context shown
of
of in
ASM
ASM Figure
Consortium
Consortium (ASM Consortium,
research,
research, an
an 2007).
abnormal
abnormalIn to
that
to hurt
follow procedures
process
procedures is
safety.
is defined
CCPS
defined as
metrics
as for
“the measuring
percentage of
failure
of the
situation
the
the is
context
context
situation is considered
of
of ASM
ASM
considered an
Consortium
Consortium
an event
event that requires
research,
research,
that an
an
requires operator
abnormal
abnormal
operator number
to follow
number
to followof
of safety
safety critical
procedures is
critical
procedures tasks
is tasks observed
defined
defined as
as “the
observed where
“the all steps
percentage
percentage
where all steps of
of
of
of the
the
situation
interventionis considered an event that requires aa operator number
relevant of safety
safe critical
working tasks
procedureobserved where
were where all
not followedsteps
followed of
to the
situation
situation
intervention
interventionis (Bullemer,
is consideredet
considered
(Bullemer,
(Bullemer, et
et
al.,
an
an
al.,
al.,
2011).
event According
event
2011).
2011).
that
that requires to
requires
According
According to
to
research
operator
aa operator
research
research
number
number
relevant
relevant
of safety
ofsafe
safety
safe
critical
working
working
tasks
tasks observed
criticalprocedure
procedureobserved
were
were where
not
not
all
all steps
steps to
followed
of
of the
to the
the
study carried
intervention
intervention
study carried out by
(Bullemer,
(Bullemer,
out by the
theet
et ASM
al.,
al.,
ASM Consortium,
2011).
2011). According
According
Consortium, the
the to
tomajority
a of
research
research
majority of total number
relevant
relevant
total numbersafe
safe of safety
working
working
of safety critical tasks
procedure
procedure
critical tasks observed”
were
were not
not
observed” (CCPS,
followed
followed
(CCPS, 2011).
to
to
2011).the
the
study
study carried
operating
carried out
procedure
out by
by the
the ASM
execution
ASM Consortium,
failures
Consortium, had the majority
occurred
the majority of
during
of total
total number
number of
of safety
safety critical
critical tasks
tasks observed”
observed” (CCPS,
(CCPS, 2011).
2011).
study carried
operating
operating out
procedure
procedure by the ASM
execution
execution Consortium,
failures
failures had
had the majority
occurred
occurred of
during
during total number
Monitoring and of safety critical
and identifying tasks
identifying procedural observed”
procedural deviations (CCPS,
deviations are 2011).
are critical
critical
abnormal
operating
operating
abnormal situations
procedure
procedure
situations (Bullemer,
execution
execution
(Bullemer, et al.,
failures
failures
et al., 2011).
had
had
2011). Factors
occurred
occurred
Factors such as
during
during
such as Monitoring
Monitoring and identifying procedural deviations are critical
abnormal
inadequate
abnormal situations
skills (Bullemer,
and experience et al., 2011).
levels of Factors
the such
operational as tasks to
Monitoring
Monitoring
tasks to ensure
and operational
identifying
and operational
ensure discipline,
procedural
identifying discipline, compliance
deviations
proceduralcompliance
deviations are and
are
and safety
critical
critical
safety
abnormal situations
inadequate
inadequate situations
skills
skills
(Bullemer,
(Bullemer,
and
and experience
experience
et
et al.,
al., 2011).
2011).
levels
levels of
of
Factors
Factors
the
the
such
such as
operational
operational as tasks
(Rains,to ensure
2010). operational
The workflowdiscipline,
frameworkcompliance
developedand safety
in this
crew, failures
inadequate to
skills make
and consistent
experience response
levels of and
the procedural
operational tasks
(Rains,to ensure
tasks to 2010).
ensure Theoperational
operational
workflowdiscipline,
discipline,
frameworkcompliance
compliance
developedand
and insafety
safety
this
inadequate
crew, skills
failures to and
make experience
consistent levels
response of the
and operational
procedural (Rains, 2010). The workflow framework developed in
crew, failures
violations during to make consistent response and procedural paper
(Rains, is instrumental
2010). The workflowin
workflow identifying instances
frameworkinstances
developedwhen in this
when thisaa
crew,
crew, failures
failures
violations
violations during
during to infrequent
to make
infrequent
infrequent
process
make consistent
consistent
process
process
operations
response
response
operations
operations
and
andsuch
such
such
as
as
as
start-
procedural
procedural
start-
start-
(Rains,
paper
paper 2010).
is
is
stipulated
The
instrumental
instrumental
operating
in
in
procedure
framework
identifying
identifying
is not
developed
instances
followed. The
in
when this
concept aa
up and
violationsshutdown
during procedures
infrequent processpose serious
operations threat
such to
as plant
start- paper
paper is
stipulated instrumental
is operating
instrumental in
in identifying
procedure is not
identifying instances
followed.
instances The when
concept
when a
violations
up
up and
and during
shutdown
shutdown infrequent
procedures
procedures processpose
pose operations
serious
serious such
threat
threat as
to
to start-
plant
plant stipulated
of workflow
stipulated operating procedure
conformance
operating is not
monitoring
procedure is not followed.
is The
demonstrated
followed. The concept
from
concept
safety
up
up and
and
safety and profitability
shutdown
shutdown
and (Walker,
procedures
procedures
profitability (Walker, et
pose
pose
et al.,
al., 2011).
serious
serious
2011). threat
threat to
to plant
plant stipulated
of
of workflow
workflow operating procedure
conformance is not
monitoring followed.
is The
demonstrated concept
from
safety
safety and
and profitability
profitability (Walker,
(Walker, et
et al.,
al., 2011).
2011). an
of
an analysis conformance
of workflow
workflow
analysis of event
event logs
conformance
conformance
of
monitoring
logs to develop
monitoring
monitoring
to
is
develop
is demonstrated
burner rotation
is demonstrated
demonstrated
burner
from
rotation
from
from
safety
The and
Centerprofitability
for (Walker,
Chemical et
Processal., 2011).
Safety (CCPS) has an analysis
workflows of of event
boiler logs
operations toat develop
the Heatingburner
plant rotation
of the
The Center for Chemical Process Safety (CCPS) has an
an analysis
analysis
workflows of of
of event
event
boiler logs
logs
operations to
toat develop
develop
the Heatingburner
burner
plant rotation
rotation
of the
The Center
recommended
The Center for
for Chemical
metrics
Chemical for Process
measuring
Process Safety
Safety (CCPS)
process
(CCPS) has
safety
has workflows
University
workflows of
of
of boiler
Alberta.
boiler operations
Workflows
operations at
at the
that
the Heating
do not
Heating plant
conform
plant of
to
of the
set
the
The Center
recommended
recommended for Chemical
metrics
metrics for
for Process
measuring
measuring Safety (CCPS)
process
process has
safety
safety workflows
University
University of Alberta.
of
of boiler operations
Alberta. Workflows
Workflows at that
the do
that Heating
do not
not plant of
conform
conform to
to the
set
set
performance
recommended
recommended
performance of
of an organization
metrics
metrics
an for
for
organization (CCPS,
measuring
measuring
(CCPS, 2011).
process
process
2011). Procedural
safety
safety
Procedural procedures
University
University
procedures of are
of are further
Alberta.
Alberta. investigated
Workflows
Workflows
further that do
that do to
investigated to
not identify
conform
not conform
identify to and
toandset
set
performance
violations are of an organization
categorized under (CCPS,
unsafe 2011).
operating Procedural procedures
troubleshoot are further
operational investigated
problems. to identify and
performance
performance
violations
violations are
are
of
of an
an organization
organization
categorized
categorized under
under
(CCPS,
(CCPS,
unsafe
unsafe 2011).behaviors
2011).
operating
operating
Procedural
Procedural
behaviors
behaviors
procedures
troubleshoot
procedures
troubleshoot
are further
operational
are further investigated
problems.
investigated
operational problems.
to
to identify
identify and
and
violations
violations are
are categorized
categorized underunder unsafe
unsafe operating
operating behaviors
behaviors troubleshoot
troubleshoot operational
operational problems.
problems.
1
1 Corresponding author. Tel +1 780 492 5162; Fax +1 780 492 2281;
1 Corresponding author. Tel +1 780 492 5162; Fax +1 780 492 2281;
Corresponding
1Email:
1 author. Tel +1
sirish.shah@ualberta.ca 780 492 5162; Fax +1 780 492 2281;
Corresponding
Email: author. Tel +1
sirish.shah@ualberta.ca 780 492 5162; Fax +1 780 492 2281;
Email: sirish.shah@ualberta.ca 780 492 5162; Fax +1 780 492 2281;
Corresponding author. Tel +1
Email: sirish.shah@ualberta.ca
Email: sirish.shah@ualberta.ca
2405-8963 ©
Copyright © 2015, IFAC (International Federation of Automatic Control)
2015 IFAC 451Hosting by Elsevier Ltd. All rights reserved.
Copyright
Peer review© 2015 responsibility
IFAC 451Control.
Copyright ©under
2015 IFAC of International Federation of Automatic
451
Copyright
Copyright © 2015 IFAC
© 2015 IFAC
10.1016/j.ifacol.2015.09.009 451
451
IFAC ADCHEM 2015
452
June 7-10, 2015. Whistler, BC, Canada Sridhar Dasani et al. / IFAC-PapersOnLine 48-8 (2015) 451–456

2. EVENT LOGS AND WORKFLOW MINING
2.1 Event Logs in Process Industry
With the availability of state-of-the-art information systems
used in the process industry, the majority of the operator
interactions with the plant floor (process and equipment) such
as set-point changes, controller tuning steps, control loop
diagnostic steps, and instrument troubleshooting tasks are
archived in the event log database as shown in Figure 2.
of capturing the best operating practices by analyzing text-
based event logs associated with the execution of a process.
Workflow mining techniques are widespread in the areas of
business process management (Turner, et al., 2008). In this
paper, we intend to use these techniques in the context of
operations related to the process industry. For extracting
workflows from archived event logs, we use a demo version
of Disco, an automated process discovery software developed
by Fluxicon, a spin-off of process mining research at
Eindhoven University of Technology (TU/e) (Dongen, et al.,
2005).

3. INDUSTRIAL CASE STUDY: BOILER OPERATIONS

AT A HEATING PLANT

Operator interactions 3.1 Description of the Process

with control systems
The Heating plant at the University of Alberta supplies steam
Application Invocation required for campus space heating and research facilities. The
(Historians, Asset plant is a cogeneration facility that has a steam production
Management MES –
ERP – Alarm & Event) capacity of 650 tonnes per hour and the steam load varies
Logs) with season. The plant has two boilers which produce steam
at 2,760 kPa and three boilers producing steam at 6,200 kPa.
We focus on the start-up procedure of Boiler #4 which is a
6,200 kPa natural-gas fired boiler which runs continuously
Shift – Handover log, during the winter terms. A schematic of a natural-gas fired
Field Maintenance Task Process
drum-type boiler is shown in Figure 3.
A drum-type boiler has two drums; namely a steam drum and
a mud drum. These two drums are connected through a series
Event Log of water-tubes called risers and downcomers. Steam is
Event Pre-processing,
Alarm Filtering &
generated when these water-tubes are exposed to heat within
Log Data-Mining the gas combustion furnace where a mixture of natural gas
Database Industrial and air is ignited at a gas burner. Gas burners ensure proper
Workflow Mining Workflows
mixing and combustion of natural gas and air in the boiler
furnace. Steam and water rise up in the tubes that are closer
Fig. 2. Block Diagram for developing industrial workflows
to the furnace (risers) and water drops in the tubes which are
Application invocations such as operators viewing process farther away from the furnace (downcomers) creating a water
trends, historians, process alarms, requests for system circulation phenomenon (Heselton, 2005). Steam
diagnostic information and maintenance tasks are also accumulated at the steam drum (top drum) flows through a
archived in the event log database. Hence, an event log series of tubes which are exposed to the furnace where any
database is a rich source of process and operational entrained water droplets in the steam are completely
knowledge. These archived text-based event messages are vaporized (resulting in dry steam). This part of the boiler
collected in the form of Excel spreadsheets and are pre- section where the temperature of the steam is raised is called
processed, filtered and data-mined to extract industrial a superheater.
workflows using workflow mining tools such as Disco
As shown in Figure 3, most industrial type boilers are
software (Günther & Aalst, 2007). In our industrial case
equipped with multiple burners. Figure 4 shows the physical
study, we extract workflows by analyzing event logs
orientation of the six burners associated with Boiler #4 in the
associated with the start-up procedure of Boiler #4 at the
Heating plant. The warm-up rate of a boiler during a start-up
Heating plant on November 09, 2012. Start-up of Boiler #4 is
is specified by the boiler manufacturer. Rapid warming up of
a complex task and consists of 18 steps which span over 21
a boiler during start-up results in temperature swings creating
pages of a paper-based SOP. To simplify our analysis, we
thermal stress on thick metallic tubes associated with boiler
focus on developing workflows for a critical sub-process
drums, superheaters, and economizers. Especially,
called burner rotation sequence. Section 4 provides a
superheater headers are subjected to slow warm up to prevent
detailed explanation of event log pre-processing and filtering
hand-hole welds from cracking. Hence, during a start-up,
steps associated with the burner rotation.
boilers are warmed up gradually by rotating burners. Based
on the design specifications of Boiler #4, the rate of warm-up
2.2 Workflow Mining Tools from a cold setting is limited to 38 deg C rise per hour (1500
kPa per hour) while rotating the burners.
Workflow mining or process mining refers to the technique

452
 IFAC ADCHEM 2015
June 7-10, 2015. Whistler, BC, Canada Sridhar Dasani et al. / IFAC-PapersOnLine 48-8 (2015) 451–456 453

Steam Boiler #4. Notice that each burner is ignited for 20 minutes
Drum
before switching to another burner. The subsequent burner is
Superheated
Economizer
ignited before shutting down the current burner. This is done
Steam
to avoid a purge of the burner section gases, which
automatically occur in case of a flame failure. The number of
Superheater
burners ignited and rotated during a warm-up period depends
on whether Boiler #4 is brought to operating conditions from
Boiler a cold-setting or from a hot standby. During a boiler start-up,
Distance between Superheater and Burners

Feedwater operators at the Heating plant always follow the burner

rotation sequence as shown in Figure 5 to prevent uneven
Downcomers
heating of superheater headers. As shown in Figure 4, it is
Risers
inferred that the burners which are located in the bottom (#1,
Mud
Drum
#2, and #3) are ignited first during the burner rotation
sequence to ensure uniform heating of the boiler components.

4. PRE-PROCESSING EVENT LOG DATA FOR BURNER

ROTATION SEQUENCE
Boiler #4 is controlled and monitored by a PROVOX control
system that is interfaced to an OSIsoft PI Server from where
Gas event data is retrieved in the form of Excel files for analysis
Burners
and report generation. We analyze alarm and event messages
Fig. 3. Schematic of a drum-type natural-gas fired boiler which are associated with the start-up of Boiler #4 on
November 09, 2012. The campus heating load is generally at
its peak at this time of the year and hence this boiler was
started up from a hot standby condition. The boiler start-up
Burner 4 Burner 5 Burner 6 procedure began at 02:00 PM and was brought to operating
conditions by 05:00 PM on the same day. During this period
of three hours, a total of 735 alarms and events were archived
Burner 1 Burner 2 Burner 3 in the PROVOX system.

Fig. 4. Arrangement of burners in Boiler #4 at the Heating 4.1 Filtering Event Logs Corresponding to Boiler #4
Plant, University of Alberta
The raw event log spreadsheet was complex and contained
3.2 Recommended Practice for Burner Rotation Sequence about 225 Excel columns representing various data attributes
related to Boiler #4 operations. Some of the data attributes
START
such as system nodes, system configuration alerts, device
address, alarm grouping, and alarm priorities were less
FIRE Burner #1 for 20 minutes relevant to workflow mining tasks. To develop workflows for
burner rotation sequence, pre-processing is carried out on the
LIGHT Burner #2 raw event log spreadsheets by filtering events that are
relevant to burner rotation sequence shown in Table 1.
SHUT OFF Burner #1

FIRE Burner #2 for 20 minutes

Table 1. A section of filtered event messages corresponding
to the burner rotation sequence
LIGHT Burner #3
Event Event Point Point
Area SpStr PvStr
Date Time Tag Desc
SHUT OFF Burner #2 B4
02:51:17 BB- Off On
11/09/2012 BOILER-4 Burn1
PM 4050 Line Line
FIRE Burner #3 for 20 minutes Sta
B4
02:51:20 BB- On On
LIGHT Burner #1
11/09/2012 BOILER-4 Burn1
PM 4050 Line Line
Sta
SHUT OFF Burner #3
B4
02:51:22 BB- On On
11/09/2012 BOILER-4 Burn1
PM 4050 Line Line
END Sta
B4
02:51:41 BB- On Off
11/09/2012 BOILER-4 Burn1
PM 4050 Line Line
Fig. 5. Recommended burner ignition sequence for Boiler #4 Sta
B4
02:52:09 BB- On Off
Figure 5 shows a recommended practice obtained from the 11/09/2012 BOILER-4 Burn1
PM 4050 Line Line
Sta
SOP for the burners ignition process during the start-up of

453
IFAC ADCHEM 2015
454
June 7-10, 2015. Whistler, BC, Canada Sridhar Dasani et al. / IFAC-PapersOnLine 48-8 (2015) 451–456

Event log pre-processing also ensures that the event log file is Table 3. A section of custom event messages created based
compatible with Disco software, which can then be used to on the current status of the burners
generate the burner rotation workflows automatically. From
Time Point Point
Table 1, columns “Event Date” and “Event Time” represent Stamp Tag Desc
Custom_Event_Message
the actual date of occurrence and time stamps of the events
BB- B4 Burn1
respectively. Column “Area” denotes the unit or the piece of 02:51:17 BB-4050-B4 Burn1 Sta(On Line)
4050 Sta
equipment under consideration which is Boiler #4 in our BB- B4 Burn1
case. Columns “PointTag” and “PointDesc” describe the tag 02:51:20 BB-4050-B4 Burn1 Sta(On Line)
4050 Sta
name of an instrument and its description respectively. BB- B4 Burn1
02:51:22 BB-4050-B4 Burn1 Sta(On Line)
Column “PvStr” represents the current status of an instrument 4050 Sta
where as “SpStr” denotes the set-point status of an instrument BB- B4 Burn1
02:51:41 BB-4050-B4 Burn1 Sta(Off Line)
as set by the control room operator. 4050 Sta
BB- B4 Burn1
02:52:09 BB-4050-B4 Burn1 Sta(Off Line)
4050 Sta
4.2 Creating Custom Event Messages

Although the PROVOX system archives all events associated Similarly, event messages archived under “PvStr” column
with boiler operation, any single data attribute (each column take two states namely; ‘On Line’ or ‘Off Line’ which
in Table 1) alone does not convey meaningful information of represents the current status of burners during boiler warm-up
the entire process by itself. It is, therefore, necessary to create operation. Each Custom_Event_Message string shown in
unique text messages from the event logs, which will enable Table 3 indicates whether a burner is currently ignited or shut
workflow mining tools to capture all critical tasks associated off. There are 23 custom event messages associated with
with the burner rotation sequence. Text character strings of burner rotation during the start-up of Boiler #4 on November
each and every event associated with columns “PointTag”, 09, 2012. Data pre-processing allows Disco to automatically
“PointDesc” and “SpStr” or “PvStr” are concatenated to extract workflows based on time stamps and custom event
create unique Custom_Event_Message text-strings as shown messages. Each custom event message string appears as a
in Tables 2 and 3. block in the workflows which are shown in Figures 7 and 8.
Event messages archived under “SpStr” column take two
states namely; ‘On Line’ or ‘Off Line’ which represents the 5. WORKFLOWS FOR BURNER ROTATION SEQUENCE
set-point status of burners as set by the control room operator
during boiler warm-up operation. Hence, each 5.1 Based on Set-point Status Set by the Operator
Custom_Event_Message string in Table 2 represents an
operator action while rotating the burners. The workflow shown in Figure 7 is extracted from custom
event messages which are created based on set-point status of
Table 2. A section of custom event messages created based the burners (“SpStr”) as described in Table 2. A control room
on operator set-point status of the burners operator sets these states in the PROVOX control system
Time Point Point
when a burner is ignited or shut off during boiler warm-up. A
Custom_Event_Message burner rotation workflow developed based on the set-point
Stamp Tag Desc
BB- B4 Burn1 status of the burners (Figure 7) captures operator actions (set-
02:51:17 BB-4050-B4 Burn1 Sta(Off Line) point changes in the operator console) during the burner
4050 Sta
BB- B4 Burn1 rotation sequence. According to the workflow shown in
02:51:20 BB-4050-B4 Burn1 Sta(On Line)
4050 Sta Figure 7, Burner #1 is taken online (ignited) first and then
02:51:22
BB- B4 Burn1
BB-4050-B4 Burn1 Sta(On Line)
switched to Burner #4 while shutting off (offline) Burner #1.
4050 Sta After igniting Burner #4, ignition is switched to Burner #3
BB- B4 Burn1 while Burner #4 is shut off. Finally Burner #2 is taken online
02:51:41 BB-4050-B4 Burn1 Sta(On Line)
4050 Sta
(ignited). Warm-up of Boiler #4 continues with Burner #2
BB- B4 Burn1 and Burner #3 ignited while the boiler reaches operating
02:52:09 BB-4050-B4 Burn1 Sta(On Line)
4050 Sta
steam conditions. The frequency of occurrence of an operator
action is shown with a number within each block. Similarly, a
Each custom event message string can be deciphered into directed arrow connecting two blocks indicates the path of
three components for our analysis as shown in Figure 6. actions or events during the burner rotation sequence. A
number on the directed arrow indicates the number of times a
BB-4050-B4 Burn1 Sta(Off Line) particular transition occurred between two blocks in the
direction of the arrow.
Tag Tag Set-point/
Name Description Current Status 5.2 Based on Current Status of the Burners

Fig. 6. Interpreting a custom event message string Figure 8 shows a burner rotation workflow extracted from
custom event messages which are created based on the
current status of the burners (PvStr) as described in Table 3.

454
 IFAC ADCHEM 2015
June 7-10, 2015. Whistler, BC, Canada Sridhar Dasani et al. / IFAC-PapersOnLine 48-8 (2015) 451–456 455

conformance events that arise during an abnormal plant

Burner Rotation Sequence Starts
behavior.

1 Burner #1: SP = Offline Burner Rotation Sequence Starts

BB-4050-B4 Burn1 Sta(Off Line)
(2)
1 Burner #1: PV State = Online
1 1 BB-4050-B4 Burn1 Sta(On Line)
(3) 2
BB-4050-B4 Burn1 Sta(On Line) 1
3
(4) 1 Burner #1: PV State = Offline
Burner #1: SP = Online BB-4050-B4 Burn1 Sta(Off Line)
Burner #4: SP = Offline (3) 2
BB-4040-B4 Burn4 Sta(Off Line) 1 Burner #4: PV State = Online
(2) BB-4040-B4 Burn4 Sta(On Line)
1 (3) 2
Burner #4: SP = Online 1 1 Burner #4: PV State = Offline
BB-4040-B4 Burn4 Sta(On Line)
3 BB-4040-B4 Burn4 Sta(Off Line)
(4) 1
1 (3)
Burner #3: SP = Offline 1 1
BB-4030-B4 Burn3 Sta(Off Line) BB-4020-B4 Burn2 Sta(Off Line) 1
1
(2) (1)
1 Burner #3: SP = Online 1 Burner #3:
Burner #2: PV State = Offline
BB-4030-B4 Burn3 Sta(On Line) PV State = Offline
4
(5) BB-4030-B4 Burn3 Sta(Off Line)
1
Burner #2: SP = Offline 1 (3)
BB-4020-B4 Burn2 Sta(Off Line) Burner #3: PV State = Online 2 1
(2) BB-4030-B4 Burn3 Sta(On Line)
2
Burner #2: SP = Online 1 (4)
BB-4020-B4 Burn2 Sta(On Line) Burner #2: PV State = Online 1
(2) 1
BB-4020-B4 Burn2 Sta(On Line)
1 2
(3)
1
Burner Rotation Sequence Ends
Burner Rotation Sequence Ends
Fig. 7. Burner rotation workflow developed based on
operator actions Fig. 8. Burner rotation workflow developed based on the
We can infer from this workflow that Burner #1 is ignited current status of the burners
first and then switched to Burner #4 while shutting off Burner When a workflow is extracted from event logs, these
#1. After igniting Burner #4, ignition is switched to Burner abnormal alarms and non-conformance events appear as
#3 while Burner #4 is shut off. Boiler #4 is brought to message blocks along with the anticipated events. Hence,
operating conditions by continuing ignition with Burner #2 process deviations are identified visually on-the-fly by
and Burner #3. A burner rotation workflow developed based comparing current workflows with stipulated operating
on the current status of the burners (Figure 8) captures on- procedures or best operating practices.
going events or burner status during a boiler warm-up
procedure. Such a workflow captures the status of the process Conformance monitoring helps to continuously monitor
(burners in our case) and provides an insight into the health operator actions and on-going events by comparing them
and diagnostic information about the instrumentation as well. with an existing workflow model or SOP or a ‘best-practice’
operational guideline. Hence, conformance monitoring helps
6. WORKFLOW CONFORMANCE MONITORING to identify operators who deviate from an established
procedure. Deviations observed in the workflow activities are
Very often a process deviates from its expected behavior for further investigated to understand and solve operational
several reasons such as; changing operating conditions or problems associated with process, equipment and human
constraints, operators failing to follow an established factors. Based on the workflows (Figures 7 and 8) developed
procedure, operational inconsistencies during start-up, for burner rotation sequence of Boiler #4, we arrived at 4 key
shutdown and transitions, process and equipment failure to findings which are summarized in the following subsections.
name a few. As event log database continuously archives These outcomes were further validated by discussions with
events and alarms that are generated in the plant, process the shift operational crew who had brought Boiler #4 online
deviations are recognized in terms of alarms and non- from a hot standby on November 09, 2012.

455
IFAC ADCHEM 2015
456
June 7-10, 2015. Whistler, BC, Canada Sridhar Dasani et al. / IFAC-PapersOnLine 48-8 (2015) 451–456
6.1 Finding #1 – Mismatch of Burner Tag Names
The tag name of Burner #1 should be BB-4010. However,
based on burner rotation workflows shown in Figures 7 and
8, it is visually evident that the tag name of Burner #1 is
incorrectly identified as BB-4050. Actually, the tag name
BB-4050 corresponds to Burner #5 which was not ignited
during the start-up of Boiler #4 on November 09, 2012. This
discrepancy in burner tag names has been reported to plant
personnel for further investigation. Mismatched tag names
result in archiving faulty information about an instrument in
process trends or historians and can cause confusion when
responding to alarms in an emergency situation.
6.2 Finding #2- Procedural Deviations
According to the SOP for burner rotation sequence shown in
Figure 5, when Boiler #4 is warmed up from a cold setting,
burners are rotated such that Burner #1 is ignited first for 20
minutes; then Burner #2 is ignited for 20 minutes while
Burner #1 is shut off, and then Burner #3 is ignited for 20
minutes. However, the ignition sequence captured in
workflows which are extracted from the actual events
(Figures 7 and 8) deviate from this recommended practice.
Based on the information from shift supervisor’s log stored
on November 09, 2012, it was found that Boiler #4 is brought
online from a hot standby, and hence any of the two bottom
burners can be ignited without rotating the burners. This fact
clears up the reason behind the observed procedural
deviation. However, it is apparent the SOP should be revised
to reflect this practice.
6.3 Finding #3- Issues with Sensitivity of Flame Scanner
Burner #4 is physically situated on top of Burner #1 in Boiler
#4 as shown in Figure 4. Burner #4 appears in burner rotation
workflows captured from the actual events. However, the
shift supervisor’s log indicates that Burner #4 was not ignited
during start-up on November 09, 2012. On further
investigation, it was found that the sensitivity of flame
scanner is too high such that it is picking a signal from
Burner #1 and is misinterpreted as a flame at Burner #4.
6.4 Finding #4- Operator Actions Versus On-going Events
By comparing the workflows shown in Figures 7 and 8, it is
possible to monitor whether the field instruments are actually
responding in a way they are set by the operator in the control
room. Non-conformance message blocks observed in
workflows developed from operator set-points and on-going
events point towards a possible requirement for preventive
maintenance of the field devices. In our case study, the
sequences of burner ignition in both workflows match.
6.5 Auditing of Operating Procedures
Gruhn et al. (2006) stated that in order for operating
procedures to be accounted for as a protection layer for
process safety, procedures are required to be documented;
456
plant personnel are trained to follow them and their
implementation must be audited. This statement emphasizes
the importance of conformance monitoring of operator
actions with respect to process safety. Conformance
monitoring provides documented information (audit) on
operator actions while executing a procedure in order to meet
regulatory compliance. By comparing Figures 5 and 7, one
can check if the sequence of operations for burner rotation
were followed correctly as per the SOP guidelines.
7. CONCLUSIONS
One of the key visual aspects of the proposed workflow
framework is that event messages corresponding to operator
actions or on-going events appear as text-based message
blocks or activities in the workflow diagram. Hence,
procedural violations, through audits, are identified in real
time or on-the-fly to ensure process safety and procedural
compliance. The benefits of workflow conformance in
identifying operational problems are demonstrated using an
industrial case study. Further, workflows extracted from the
event logs of an expert operator capture valuable operational
knowledge which can then be used to train a new generation
of operational crew. Hence, the proposed workflow
framework can significantly benefit the process industry by
providing tools to improve operator skills, achieve superior
operational visibility and enforce operational discipline.
REFERENCES
ASM Consortium, 2007. Abnormal Situation Management -
Effective Automation to Improve Operator Performance,
College Station.
Bullemer, P. T., Kiff, L. & Tharanathan, A., 2011. Common
procedural execution failure modes during abnormal
situations. Journal of Loss Prevention in the Process
Industries, November , 24(6), p. 814–818.
CCPS, 2011. Process Safety Leading and Lagging Metrics,
New York.
Dongen, B. F. v. et al., 2005. The ProM framework: A new
era in process mining tool support. Application and
Theory of Petri Nets 2005, Issue 3536, pp. 444-454.
Gruhn, P. & Cheddie, H. L., 2006. Safety Instrumented
Systems: Design, Analysis and Justification. 2nd ed.
Research Triangle Park, NC: ISA.
Günther, C. . W. & Aalst, W. M. P. V. D., 2007. Fuzzy
Mining – Adaptive Process Simplification Based on
Multi-perspective Metrics. Business Process
Management - Lecture Notes in Computer Science,
Volume 4714, pp. 328-343.
Heselton, K. E., 2005. Boiler Operator’s Handbook. New
York: The Fairmont Press, Inc.
Rains, B. D., 2010. Operational Discipline: Does Your
Organization Do the Job Right Every Time?,
Wilmington.
Turner, C. J., Tiwari, A. & Majeed, B., 2008. A review of
business process mining: state-of-the-art and future
trends. Business Process Management Journal, 14(1),
pp. 5-22.
Walker, P. D., Cammy, N. E., Ellis, B. J. & Seibert, K. D.,
2011. Operations Skills for 21st Century, Des Plaines.