Respond to right alarms
When alarms are going off left and right, the noise becomes deafening—metaphorically speaking anyway. Operators turn a deaf ear because they are overwhelmed by constant alarms.
With no documented alarm management process, Gary Godfrey, senior staff engineer in the control systems department at Saudi Petrochemcial Company in Saudi Arabia, knew this was not a good way to keep the plant running safely. Operators ignore alarms because there are just too many to deal with. “If you’re being told 20 times a minute something’s going wrong, you have no think time to settle in and know what to do about that alarm,” he said.
One of the problems is a management system via paper, and no alarm system performance measures. “Today if it’s not electronic, people don’t do it,” he said.
You need a philosophy for your plant well documented and well written, or you cannot do alarm management.
Audible alarms must have operator action. “If your alarm philosophy contains that one line, you’re 80% the way there. If there’s no action for the operator, and there is nothing they can do about it, then don’t tell them.”
You need a governing document, and you need site-wide buy-in. An alarm philosophy is not one size that fits all. It is going to be unique to your plant and practices.
Incorporate industry guidelines. “One thing we embraced was EEMUA, a U.K.-based organization that has the industry standard for alarm management” Godfrey said.
Include basic definitions: What is an alarm?
Be aware of alerts vs. alarms. Alarms you have to take quick action on. Alerts you may need to check every 12 hours.
Make sure you have alarm dead bands. This is key so you do not get chattering alarms. If it’ is just going to sit there and go in and out, you need a dead band there. Define values in percents or EU. Use PV clamping on indication points.
Set alarm priorities, such as emergency/high/low. Only 5% of your alarms should have emergency priority. High should be 1%. For that, you probably have a minute to a five-minute reaction time. Alarms that are low priority (i.e., the filter needs changing) can be attended to next shift. “You don’t have to get out there in the next two minutes. And remember; alarms are for the operators, not engineers.”
Pilot study gives “incredible” results
Godfrey and his team did a pilot study for six months, and “the numbers were incredible,” he said. There was an 83% reduction on one of the consoles in six months, and operators could respond in a timely fashion.
Information was kept in Excel with major columns including alarm description, reason for alarm, consequence of not alarming, expected operator action, response time defines priority), PVLO, PVHI, and other LCN parameters.
Only look at bad actors, and consider the 80/20 rule, he said. “We did a weekly bad actor review. Somebody has to own the alarm management. Empower someone to make decisions.” Owners in this project were control system engineers, production/process engineers, and operations representatives. Team membership continuity is crucial. Justify all bad actors. “Over time, you’ll build up a database of rationalized alarms online for operators in the future,” he said.
With one alarm per 10 minutes on average, Godfrey was pleased with the results of the study. But he cautioned alarm management is an ongoing process. “Don’t just do it once and walk away. You’ve got to walk before you run, so keep it simple.”
Other gems of advice included having management support for alarm management.
“In the future, we’re going to address standing alarms, reduce inhibited and disabled alarms, operator changes (top changed points and why), and operator changes per audible alarm,” he said.
— Ellen Fussell Policastro
With no documented alarm management process, Gary Godfrey, senior staff engineer in the control systems department at Saudi Petrochemcial Company in Saudi Arabia, knew this was not a good way to keep the plant running safely. Operators ignore alarms because there are just too many to deal with. “If you’re being told 20 times a minute something’s going wrong, you have no think time to settle in and know what to do about that alarm,” he said.
One of the problems is a management system via paper, and no alarm system performance measures. “Today if it’s not electronic, people don’t do it,” he said.
You need a philosophy for your plant well documented and well written, or you cannot do alarm management.
Audible alarms must have operator action. “If your alarm philosophy contains that one line, you’re 80% the way there. If there’s no action for the operator, and there is nothing they can do about it, then don’t tell them.”
You need a governing document, and you need site-wide buy-in. An alarm philosophy is not one size that fits all. It is going to be unique to your plant and practices.
Incorporate industry guidelines. “One thing we embraced was EEMUA, a U.K.-based organization that has the industry standard for alarm management” Godfrey said.
Include basic definitions: What is an alarm?
Be aware of alerts vs. alarms. Alarms you have to take quick action on. Alerts you may need to check every 12 hours.
Make sure you have alarm dead bands. This is key so you do not get chattering alarms. If it’ is just going to sit there and go in and out, you need a dead band there. Define values in percents or EU. Use PV clamping on indication points.
Set alarm priorities, such as emergency/high/low. Only 5% of your alarms should have emergency priority. High should be 1%. For that, you probably have a minute to a five-minute reaction time. Alarms that are low priority (i.e., the filter needs changing) can be attended to next shift. “You don’t have to get out there in the next two minutes. And remember; alarms are for the operators, not engineers.”
Pilot study gives “incredible” results
Godfrey and his team did a pilot study for six months, and “the numbers were incredible,” he said. There was an 83% reduction on one of the consoles in six months, and operators could respond in a timely fashion.
Information was kept in Excel with major columns including alarm description, reason for alarm, consequence of not alarming, expected operator action, response time defines priority), PVLO, PVHI, and other LCN parameters.
Only look at bad actors, and consider the 80/20 rule, he said. “We did a weekly bad actor review. Somebody has to own the alarm management. Empower someone to make decisions.” Owners in this project were control system engineers, production/process engineers, and operations representatives. Team membership continuity is crucial. Justify all bad actors. “Over time, you’ll build up a database of rationalized alarms online for operators in the future,” he said.
With one alarm per 10 minutes on average, Godfrey was pleased with the results of the study. But he cautioned alarm management is an ongoing process. “Don’t just do it once and walk away. You’ve got to walk before you run, so keep it simple.”
Other gems of advice included having management support for alarm management.
“In the future, we’re going to address standing alarms, reduce inhibited and disabled alarms, operator changes (top changed points and why), and operator changes per audible alarm,” he said.
— Ellen Fussell Policastro
posted by ISA at 12:49 AM
<< Home