A modern process plant easily has over 10.000 real-time database points. Each of these points can be used to generate multiple alarms in case of non-normal situations. This development is usually technology driven: whenever a component is capable of sending alarm, it will be implemented, just to be ‘on the safe side’.
A common problem is that far too many alarms and off-normal messages are presented to the operator. The effect on the process operator is the opposite: too many messages lead to not looking at them anymore.
Below are the general guidelines for the acceptable number of alarms:
< 2 alarms/10 minute interval per operator position, for normal process conditions;
< 10 alarms/10 minute interval per operator position, for up-set process conditions;
< 10 standing alarms per operator position.
It is recommended to develop an operational philosophy including:
- Rules to rationalize alarm messages, in a way each comparable process unit and process condition leads to the same type of messages
- Categorization of events; such as alarm, off-normal message, log entry only, etc.
- Conventions about alarm information presentation; for example do we need a dedicated alarm screen, or do we integrate off-normal information in process graphics.
For existing process plants, some cleaning up should be considered. Simply collect alarm data over a one month period and list the Top 10 of most frequent alarms. Probably, most of the Top 10 alarms are not really ‘alarming’. An instrument may be broken down, an alarm setting is wrong, etcetera. Solving these items will reduce the alarm frequency considerably. A next step in reducing alarms, is to look for double messages: two or more alarms related to one event (for example: valve closed alarm, and a no-flow alarm). Either combine both alarms or remove one of them.
Learn more: see Alarm Systems, A guide to design, management and procurement, EEMUA Publication 191 (3th edition).