Hi all,
Treatment plant, CompactLogix 1769-L33ER on V32 with FactoryTalk View SE on top, and about 1200 alarms sitting in ALMD and ALMA instructions.
Every startup after a shutdown or a power blip, roughly 200 of them fire at once, all because nothing's running yet. The operators hit ack all without reading any of it, and last month they acked a real high chlorine alarm along with the flood, which is what's got me looking at this now.
I put a global inhibit on for 5 minutes after power-up and that kills the flood, but it hides the real ones with it and the plant manager said no. So I raised a few of the low limits instead and that just moved them later into the startup.
What do people actually do here? Any pointers would be good, I'd rather fix it in the PLC than in the SCADA.
How many of those 200 need an operator to do something? Genuinely. That's the number to start from.
Honestly, maybe 20. The rest are "that pump isn't running" in different words. The radio link to the north tank drops a couple of times a week as well, and each drop is 30 alarms on its own.
Which of the 1200 are on the P&ID as required alarms, and which got added because the tag existed and the range was there? A low flow alarm on a filter that's off isn't an alarm. It's a description of a filter that's off.
Two jobs, different sizes. The small one you can do this week: every process alarm gets a condition, the equipment it belongs to is running, plus a delay long enough to settle. A TON at 30000ms for a pump, a few minutes for a filter, in the ALMD's In condition, not a global inhibit. That isn't suppression, that's what the alarm means.
The big one is a rationalisation and I reckon that's your winter. Sit with the operators, go through the 1200, keep the ones that need an action, the rest become events in the historian. ISA-18.2 is the method. The north tank link gets one alarm, comms loss, and the 30 tags under it get bad quality. There's an article with the alarm section: https://plctr.com/exploring-plc-applications-in-the-chemical-and-petrochemical-industry-process-control-and-safety/
Deadband on the analog ones as well, or the 20 keepers will chatter.
Pump alarms are done. Running bit AND the 30000ms TON in the In condition of each ALMD, filters next week at three minutes. This morning's restart after the grid work gave us 12 alarms, all real, and the operators read every one.
So the alarms never knew what state the equipment was in, a stopped pump looked the same as a failed one, and the running bit plus a delay in each ALMD is the whole fix. The rationalisation meeting is booked. Nobody's looking forward to it.
The north tank link is one alarm now, but the 30 tags under it still show stale numbers and I can't find where in SE the bad quality greying lives. Thanks scott_m and valh7.
Twelve real ones from two hundred. That's the whole job.