The Gospel according to Rothenberg...

April 17, 2007
Doug Rothenberg, "subject matter expert" for alarm management at Invensys gave a tutorial on Alarm Management. "Telling you that an alarm isn't important is like telling you that one of your children is superfluous." "The alarm management issue is an "iceberg problem" because alarm problems are the part you see--but the underlying deficiencies are deep into the infrastructure of your plant. When you start doing alarm management effectively, you are going to impact lots of areas in the enterpri...
Doug Rothenberg, "subject matter expert" for alarm management at Invensys gave a tutorial on Alarm Management. "Telling you that an alarm isn't important is like telling you that one of your children is superfluous." "The alarm management issue is an "iceberg problem" because alarm problems are the part you see--but the underlying deficiencies are deep into the infrastructure of your plant. When you start doing alarm management effectively, you are going to impact lots of areas in the enterprise, and people will be at least temporarily very unhappy with you," Rothenberg noted. Short term management always wins unless executive management is focused on long term gain. But once you start doing alarm rationalization, you will find all those submerged deficiencies. "The goal of alarm management," Rothenberg said, "is only this: to improve the performance of the operator. We need to make alarm information available to the operator in real time, with less than five alarm activations per hour, and very few alarms active over long times. Major 'alarm flood' situations need to be managed." On a priority basis, Rothenberg said, noting the EMUA studies, we need to have <20 emergency, about 5% high, 15% medium and 80% low priority alarms. We should have <1 alarm per control loop, <50% of analog measurements, etc. Six stages to effective alarm systems (starting from where you are):
  1.  identify current processes and repair
  2.  reduce nuisance alarms
  3. consistent alarm limits and priorities
  4. alarms for standardized parts of the process
  5. full rationalization
  6. knowledge-based alarming
With each step you get batter operation, better "flood control" and better operator guidance. What's the problem with stopping at step 5? Most people do. Step 6 says, "Let me make the alarm system smart enough so that I only give you an alarm when it is useful to the plant. Starting from ZERO you have four steps:
  1. Assess and repair, just like above
  2. Design alarms for all the subsystems of the process
  3. finish rationalization
  4. knowledge-based alarming
This gets you as good as you can get. But most people stop before step 4, or step 6. Seven Attributes of successful alarm management:
  1. timely activation -- when you have enough time to do something about it
  2. designate importance-- prioritize the alarm
  3. provide information-- the encapsulated knowledge of the entire plant
  4. unique information
  5. understand the alarm
  6. provide guidance to the operator
  7. take action
If that alarm does not require explicit operator action, and that action needs to be taken, then it shouldn't be an alarm. Once you have done this process, you have to keep it alive. Audit and correct. Audit and correct. Alarm management is no longer optional.