If it needs an SOP to be safe, it needs an interlock
Spent my morning digging through HMI fault logs again. Another pump seal roasted because the unit started with the discharge closed and the seal flush blocked in. Well, there’s your problem. I keep hearing that we’ll train the operators and update the SOP. My answer is the same as always: just interlock it.
We added simple permissives: no start unless discharge valve proof open, seal flush pressure proven, suction strainer DP below limit, and the VFD will not ramp if NPSHa proxy is low. Trips latch on high vibration or low seal pot level. None of this is fancy. It just enforces physics.
Result: alarm noise dropped, seal life improved, and nights got quieter. Yes, too many interlocks can make a plant brittle, but nuisance alarms and cooked equipment are worse. If it takes a paragraph in an SOP to explain a safe start, that logic belongs in the PLC.
Where do you draw the line between operator judgment and hard interlocks? What single permissive or trip has saved you the most downtime?
Agree the logic belongs in the PLC for equipment protection, but it is worth deciding which layer it lives in. A start permissive in the BPCS (discharge valve proven open, seal flush pressure proven) is fine for asset protection. If the same function is being credited in a LOPA to prevent a safety consequence, it becomes a SIF under IEC 61511, and the standard limits a BPCS-based protection layer to a risk reduction of no more than 10. More than that needs independent sensors, logic and final elements plus proof testing.
Two failure modes I would design out: a limit switch on the discharge valve proves stem position, not flow, so a broken stem or a closed downstream block still passes. A low-flow or discharge-pressure confirmation a few seconds after start closes that gap. Second, permissives that get bypassed for a stuck start tend to stay bypassed; time-limited bypasses with an alarm and a bypass log reviewed at shift handover keep them honest.
Where I draw the line: operator judgment for degraded-but-safe states, hard interlocks where the consequence develops faster than an operator can reasonably detect, diagnose and act.
Two failure modes I would design out: a limit switch on the discharge valve proves stem position, not flow, so a broken stem or a closed downstream block still passes. A low-flow or discharge-pressure confirmation a few seconds after start closes that gap. Second, permissives that get bypassed for a stuck start tend to stay bypassed; time-limited bypasses with an alarm and a bypass log reviewed at shift handover keep them honest.
Where I draw the line: operator judgment for degraded-but-safe states, hard interlocks where the consequence develops faster than an operator can reasonably detect, diagnose and act.