8 ABB REVI EW Le v e r aging conne c tiv it y

—
L E V E R AG I N G CON N E C T I V I T Y

Intelligent Alarm
­Management directs
­operators to critical
events
Relying on data mining techniques to extract data and create
a sophisticated alarm model, ABB’s IAM is the ultimate alarm
management system for process industries.

—
Large processing plants such as petrochemical accurate sensors, unlimited data storage, and
Aldo Dagnino
ABB Advanced Analytics and power generation facilities rely on complex visualization display complexity necessitate a
Cary, NC, United States
control systems to keep processes, equipment more straight-forward control system that
aldo.dagnino@ and operations on track to produce successful focuses the operators’ attention quickly on
us.abb.com
outputs. Originally, panel boards with control pertinent information – to as few and important
Carsten Beuthel instruments connected to sensors yielded analog
S+ Operation
information on display to a well-trained expert
Mannheim, Germany
who decided what needed to be done. Emer- —
carsten.beuthel@
gency systems were activated to stop a process
de.abb.com By relying on the actual plant
in danger of surpassing safety, environmental or
Thomas-Christian
Skovholt
financially acceptable limits. operating conditions, the IAM
Digital Delivery
Oslo, Norway
permits operators to recognize
With the birth of distributed control systems
thomas.skovholt@ (DCS), operators could control processes without and address significant alarms.
no.abb.com
understanding the equipment. Configuring
Martin Hollender and deploying alarms with ease, engineers set
Marcel Dix
alarms throughout the system (typically, at 80 events/alarms as possible →01. But what if opera-
ABB Corporate Research
Ladenburg, Germany and 20 percent of equipment operating range). tors could predict critical events before they occur?
The consequence was a proliferation of alarms They could prevent the loss of production and
martin.hollender@
de.abb.com – every event set off an alarm. Suddenly process throughput; equipment breakdown and events that
marcel.dix@de.abb.com
plants had to deal with too many alarms due to threaten both environmental and human safety.
too much available data, yet insufficient ger- To accomplish this ultimate goal, ABB has
mane information. Limited visual display space focused on delivering its customers such a system:
required operators to follow instruction manuals ABB’s Intelligent Alarm Management (IAM).
in case of an event – a laboriously slow process.
Intelligent Alarm System
Even though the DCS is part of the Industrial Today, alarm management system design
Internet of Things (IIoT), the amount of big data, must incorporate ergonomics, instrumentation
level of automation, availability of inexpensive engineering, systems thinking, advanced analytics
 02| 2020 Inte ll ige nt al arm manage me nt 9

—
01 Control room oper-
and visualization to enhance the usability of respond to changes in mode of operation or
ators need to rapidly the control system. Seizing on their expertise in conditions. When a large compressor shuts down
recognize critical
events and take action
the IIoT ecosystem and DCS experience, ABB’s in a petrochemicals plant, many resulting alarms
to remedy any problems resultant IAM surpasses existing concepts to are noncritical and secondary; superfluous and
that arise; IAM relies on
actual plant operating
perfect analysis and usability. Not only is an
conditions and an array extra analysis level added to reduce nuisance
of analyses based on
data mining techniques
alarms (de-chattering), a further sophisticated —
three-layered second level of analysis has been
to identify data
patterns and support Relying on concepts like bas-
operators.
engineered. The IAM consists of a data layer that
extracts alarms and creates an alarm data model ketization, smart analysis was
according to 25 attributes, an analytics layer with
developed with historical data
newly developed algorithms to precisely and
repeatedly analyze alarm data, and the graphical from oil and gas process plants.
user interface (GUI) visualization layer based
on ABB’s user-design concept for viewing the
­analytical results →02. annoying, these alarms require the operator to
waste time searching and analyzing. ABB’s IAM
Alarms in control systems and their can remedy such problems. The dynamic system
management filters events and alarms based on the actual
Fundamentally, alarms alert the operator to current plant operating conditions, thereby
departures from normal operating conditions. clearing the path for operators to swiftly recog-
To be effective, operators must quickly and nize and address the significant alarms.
accurately assess those alarms that need
split-second attention and action. However, To accomplish this, the IAM conducts an array
conventional systems are static; the alarms don’t of five analyses: deep alarm de-chattering,

01
10 ABB REVI EW Le v e r aging conne c tiv it y

alarm sequence analysis, critical alarm analysis,

masking rules, and alarm flood analysis. All of
these methods rely on data mining techniques to
Industrial equipment identify data patterns.

IAM development
Developed with historical data from oil and gas
process plants among others, analysis relies on
concepts such as basketization, a process that
slices data into appropriate time intervals; and
Industrial control system
metrics, which enables results to be prioritized.

Basketization is used either to store repeating or

short duration alarms in the case of chattering
alarms, to divide alarm sequences into equal time
periods as with alarm sequence analysis, and to
identify alarm sequences that occur before or
after a specific event as in critical alarm analysis
→03. Alarm lifetime – the time between the return
to normal Rtnt and the activation time of an
alarm Actt – and the time gaps between alarms
Ingestion layer
(TG) – the time between the activation of one

—
Based on expertise and experi-
ence, ABB’s IAM surpasses exist-
IAM
ing concepts to perfect analysis
Data layer
and usability.

alarm and a second later occurring alarm

Alarm data Telemetry data (TGB-A = ActtB – ActtA) are determined by data
attributes →04; these are used for the de-chat-
tering analysis of alarm logs by applying two
Analytics layer
newly developed algorithms: one algorithm is
for the life of an alarm and its frequency of
repeatability; the other is for the repetitive
alarms: those in which the time gap between
activation and return to normal is short. The
analyst performs the de-chattering analysis first
and then can decide to remove these from the
Visualization layer
alarm log if appropriate.

The sequence model analysis

After de-chattering the alarm log, the IAM
uses the activations times of historical alarm
data to identify and characterize alarms that
occur frequently together. Using two measures
of significance: support and togetherness,
alarm sequences can be ranked according to
importance [1]. The algorithm creates four equal
time baskets of a specific duration: each basket
contains the alarms in the log – ordered accord-
ing to activation time. The sequence mining
02 Domain experts
 02| 2020 Inte ll ige nt al arm manage me nt 11

—
02 Industrial control
algorithm identifies closed sequences of alarms operations: shut-down of equipment for preven-
systems monitor com- with support – the proportion of observations tive maintenance or cleaning purposes, etc. If no
ponents that contain
sensors. Sensors relay
in the data set that contain the specifically new alarms ensue, then these alarm floods might
information to the defined item set [1]. Togetherness is then used to be “normal” and require no further action from
ICS which filters some
nuisance alarms. ABB’s
rank the sets of closed sequences according to an operator.
IAM adds a higher level accepted procedures [1].
of analysis with three
more layers: the data
Visualizing alarms to boost usability
layer to extract and cre- Critical alarm analysis Following the extensive testing of the analytical
ate an alarm data model
from current sensor
Every operator and manager in a process plant models with data from the oil and gas industry, a
data and historical not only tries to solve critical events immediately, major contributor to the development of alarm
data, an analytics layer
that uses algorithms to
they try to avoid tripping events at all costs. The systems, ABB considered a variety of display
analyze alarm data and IAM facilitates this process by identifying those visualizations concepts. After all, the operator
a visualization layer for
viewing.
alarms that occur frequently, before or after a should know what needs to be done and how much
critical alarm (often listed for tracking purposes). time is available to do it. Targeted visualizations
—
03 Basketization (time
Relying on alarm activation times in the historical
baskets) is important alarm log, and forward and backward analysis
for alarm sequence and
critical alarm analyses.
over a specific timespan (usually 60 minutes), —
the model identifies sequences of alarms around
Different types of
baskets are needed for IAM uses an array of five analyti-
each type of analysis.
critical alarms. Each event is associated with a
level of severity that determines event criticality. cal models to provide operators
03a An example of a
critical alarm basket with better functionality.
created around a More analysis, better functionality
specific critical alarm, C;
this is used to identify
In addition to de-chattering, sequence, and
alarm sequences that critical alarm analytic models, ABB employed
occur before and after reduce the operator load, enable better event
masking rules analysis and alarm flood analysis
this specific event. Here, predictions and improve alarm system configura-
four backward and to provide operators with better functionality.
forward baskets were tions. [2]. Along the way, some well-known chal-
The masking rules method is used to determine
created for the critical lenges surfaced: limited space and too much
alarm C of length t. if the lifespan of an alarm is contained within
data on the screen; and color confusion. By
that of another alarm; this suggests whether the
03b An example of time developing concepts simultaneously to represent
slice basketization cre- contained alarm is redundant or not. As the name
ated with ABB’s newly alarm analysis, allow for data filtering and group-
states, alarm flood analysis identifies waves of
developed sequence ing of visualizations: conventional list view →05a,
mining algorithm is alarms that may occur following certain plant
shown. The algorithm sunburst diagrams →05b, and parallel
creates three equal time
baskets and each basket
contains the alarms
that are recorded in the
alarm log: These are
ordered according to
t t t t
time stamps (the time
when each alarm was
activated).
C C C C time

t t t t

C C C C time

03a

Basket 1 Basket 2 Basket 3

9:03, A_act, 9:07, B_act, 9:09, A_rtn, 9:23, A_act, 9:27, B_act 9:44, D_act, 9:50, Oper_Action_M
9:12, B_rtn, 9:13, C_act, 9:18, C_rtn 9:35, A_rtn, 9:38, B_rtn 9:56, D_rtn

A_act A_rtn C_act C_rtn A_act A_rtn D_act D_rtn

B_act B_rtn B_act B_rtn Oper_Action_M

9:00 9:20 9:40 10:00

03b
12 ABB REVI EW Le v e r aging conne c tiv it y

Global module
Attribute attribute
No (fixed) Details with example

1 Timestamp Timestamp when the event occurs. Can have various assist with monitoring alarm and safety system
formats but most popular is:
MM/DD/YYYY hh:mm:ss.msec AM/PM
performance; transforming data into actionable
eg 7/19/2014 6:09:27.527 PM insights to enable informed decision making.
2 Active Timestamp linking the various alarm status such as RTN,
timestamp inactive to its source when it is active
The use of advanced alarm analysis takes the
3 Priority Priority of the alarm typically assigned by an ordinal “guesswork” out of alarm rationalization. The
value. Lower number reflects higher severe situation eg, 1.
This attribute serves to identify if an alarm is critical. deep-learning gained from analysis can be
4 Condition An attribute indicating the associated condition of the incorporated back into the engineering tools and
alarm such as H, HH, L, LL etc. This attribute in conjunc- used to optimize or maintain the automation and
tion with “Priority” helps to rank the severity of the alarm
and also to identify the name of the alarm safety systems. ABB’s holistic process completes
5 Device Plant device associated with alarm
the alarm lifecycle management support.
6 Process area Process area in the plant where the alarm originates
ABB’s vision for alarm management is derived
7 User Operator viewing or operating on an alarm typically a
descriptive variable indicating operator name or operator from the company’s extended involvement with,
work station eg John or OL-PCD\erlst
and demonstrated commitment to, standards
8 Category Event category ID, typically an ordinal value such as like EEMUA 191 and ISA SP 18.2 that address the
666371
importance of a robust alarm management

04

—
ABB’s IAM approach provides
process industries, like metals
processing, with the potential to
coordinates →05b; criteria such as practicality,
readability, and grouping efficacy was evaluated
make predictions.
and the best views for specific analysis types
were developed [2].
capability in process plants, and their vast experi-
ence with power generation plants. Each alarm
ABB’s alarm portfolio and vision
should alert, inform and guide; alarms should be
The IAM system has been productized and
presented at a rate that operators can deal with,
is currently being incorporated into the ABB
detectable problems should be alarmed as early
Ability™ Symphony® Plus software [3]. Basically,
as possible, and the cost/benefit of alarm engi-
to perform alarm analysis, historical data is
neering should be reasonable. The Ability® Sym-
imported into the IAM as text files of various
phony™ Plus HMI, S+ Operations, implements
formats; data is transformed with the help of
ABB’s newest IAM technologies to make this
a mapping file, either created or imported, to a
vision a reality.
global alarm model (GLAM), which gives the least
common multiple of historical alarm data and
ABB’s intelligent alarm system consists of two
provides for storage →06. The software currently
critical yet distinguishable parts that cannot
defines 10 different alarm aspects; however,
be separated: the technical optimization of
up to 52 different individual aspects can be set
the alarm system and the optimization of the
depending on the needs of the customer.
human machine interface. This approach to
alarm development not only leads operators to
The new system, ABB’s AlarmInsight® and Safety-
greater awareness, faster response times, and
insight™, has been designed to help achieve safe
better decisions, it provides process industries
and reliable operations in the process automa-
with the potential for prediction. ABB’s Sym-
tion industry throughout the lifecycle of a plant
phony Plus HMI, S+ Operations with IAM has
and covers the initial engineering phase and the
abilities that are leaps-and-bounds beyond the
subsequent operational phase.
descriptive capabilities of currently available

The engineering phase relies on digitalized

alarm management systems. •
engineering data for tools to document hazards,
Acknowledgements
define appropriate barriers, and design the Special thanks are extended to the original IAM team members:
“instrumented” safety and alarm system. The Jinendra Gugaliya, Marcel Dix, Veronika Domova, and Mithun
Acharya for their valuable contribution to the IAM project. Also,
second, or operational phase, relies on contex- Roland Weiss and Alf Isaksson are thanked for their support as
tualized operational data (IT/OT). This serves to program research managers during the project.
 02| 2020 Inte ll ige nt al arm manage me nt 13

—
04 Table of eight
critical attributes used
to develop the alarm
management system.

—
05 The GUI views
were chosen to enable
maximal readability and
usability.

05a A prototype of the

UI: a list view for critical
event analysis is shown
on the right side while
a parallel coordinates
filter is displayed on the
left side.

05b The sunburst visual-

ization display includes
the sunburst (right
side) and the parallel
coordinates filter (left
side): the bigger the
sunburst segment, 05a
the greater is the
number of sequences
that start with the
alarm in question.
The size of the circle
segment represents
the amount of alarms.
The bigger the segment
the more alarms of
this name appear.
Detailed information
about the sequences
that contain the
hierarchy that has been
selected is obtained by
mouse-hovering.

—
06 A workflow diagram
for importing historical
alarm data into the
system for various
analysis types: HRA –
Hiding / Masking Rules
Analysis, DC – De-Chat-
tering, SF – Similar
Alarm Floods, a type
of sequence analysis,
PCA – Parent-Child 05b
Analysis, SM – Sequence
Model Analysis and
CEA – Critical Alarm
PGIM export Import Global model Analysis results
Analysis.

Mapping
HRA DC

ZIP GLAM SF PCA

06 SM CEA

—
References

[1] A. Dagnino, “Data [2] V. Domova, “Intel- [3] ABB Ability™ Sym-
Mining Methods to ligent Alarms Man- phony® Plus, “Intelligent
Analyze Alarm Logs agement: Sequences Alarm Management
in IoT Processes”, in analysis visualization: Technical Preview 1:
IEEE 15th International Guided report on Release Notes in Power
Conference on Auto- implementation”. ABB Generation & Water”,
mation Science and Internal Report, 2018, ABB Internal Report,
Engineering, August pp. 1-37. 2018, pp. 1-57.
22-26, Vancouver,
Canada, 2019, pp. 1-14.