Alarm Rationalization by the Numbers

An operator who gets 45 alarms in ten minutes stops reading them. At Milford Haven in 1994, two operators faced 275 alarms in the last eleven minutes before the explosion. Most plants never come that close, and most never measure how close they are. This book measures first.

It takes the ISA-18.2 lifecycle, the IEC 62682 framework, and the EEMUA 191 benchmarks and turns them into numbers you can pull from your own Ignition alarm journal: alarm rate per operator, floods, chattering and fleeting alarms, stale and standing alarms, priority mix, and the ten alarms doing most of the damage. Then it shows how to run the rationalization those numbers call for, and how to prove the result.

You will build a journal parser that rebuilds alarm episodes from Ignition's alarm_events table, a metrics engine and flood detector, chattering, fleeting, and stale alarm reports, a top-ten bad actor report, a master alarm database workbook for rationalization meetings, a shelving and suppression audit, and a before-and-after report that proves the work paid off.

Every chapter works on Red Mesa Gas Plant, a cryogenic gas plant with two console operators and an alarm system that is bad on purpose. Six months of its alarm journal, generated in the real Ignition table layout, improves chapter by chapter as the fixes go in. The code is Python 3.14 on Windows, standard library first, free and open source (MIT) at github.com/cbvance/AlarmsByTheNumbers.