Control systems and automation
T
he only advice for positioning operator alarm limits from
EEMUA 191 [1], ISA SP18.2 [2] and others was to put them on
‘the boundary of Normal Operation’ of a process. This was not
very useable advice in the absence of a method to find the boundary
of normal operation until we provided a simple method to find the
boundary in a previous paper [3] where we established that operator
alarms are not independent of each other but are inter-related through
the process operating envelope.
The method is being adopted at an accelerating rate, dramatically
reducing the time needed for alarm rationalisation and at the same
time providing much better alarm systems' performance and alarm
systems' quality.
‘Alarm quality’ has not received the same attention in the last 15
years as ‘alarm performance’ but is arguably even more important.
Using ‘alarm performance’ alone as an alarm system objective invari-
ably leads to an alarm system in which only a fraction of the alarmed
variables will ever alarm.
This fractional annunciation has been observed as fact by so
many observers in so many processes that it has wrongly become
accepted as normal. A little thought soon reveals that it is an inevitable
consequence of attempting to improve the performance of an alarm
system one variable at a time. Such fractional annunciation systems
are a process safety exposure since they do not allow the operator
alarm system to function properly as ‘the first line of defence’.
This article reviews some of the factors causing us to focus on
alarm quality and proposes the Alarm Quality Index (AQI) as an
easily calculated and easily understandable metric. AQI can be used
to quantify alarm quality and track progress in identifying and elimi-
nating fractional annunciation alarm systems. It can also be used to
monitor improvement progress after the rationalisation project when
alarm performance and improvement have been handed over to
process stewardship.
It represents a recognition that no matter how good, automated
process control can never anticipate all possible circumstances and
emphasises the strong inter-dependency between control automation
and operator alarming.
It is expected that the alarm will help the operator in understand-
ing why intervention is required as well as listing the possible operator
action. It is generally accepted in recent thinking that an alarm which
has no defined action should not be an alarm but a status variable.
An alarm on a single variable represents one bit of information
that the operator has to integrate with the single bits of information
from other alarms - and with the current and recent past history of the
many process variables available in his process control system - as
well as his own unique knowledge of the process and its foibles. His
knowledge will be unique if only because human beings learn much
more from their own mistakes, whether of omission or commission,
than they do from the mistakes of others. Fortunately most mistakes
are recoverable and many pass unnoticed by others.
There will usually be a choice of actions for the operator to
perform as there may be more than one causative event sending an
individual variable into alarm. If there is only one desired action and
Measuring
the quality
of operator
alarms
By R Brooks, J Wilson and A Mahoney, Process Plant Computing Limited
(PPCL)
A review of some of the factors causing a focus on Alarm Quality and a proposal
that an Alarm Quality Index is an easily calculated and easily understandable
metric.
Operator alarm:
A request by the process control system for the
operator to intervene with human intelligence and judgement in
response to symptoms that may indicate an abnormal situation
developing in the process.
Process control:
Used in the widest sense and includes the operator
as a part of the control capability.
Control automation:
Refers to electronic or computer-based control
by a DCS or PLC.
Electricity+Control
December ‘13
4
