Seeing more apprentices jump from 10–20 ton RTUs into plant work, and the gap is almost always controls and documentation. On a 1,200-ton water-cooled chiller startup last month, the techs who read the SOO, set BAS trend logs, and verified coil delta‑T hit the commissioning window; the rest chased phantom setpoints and burned hours — what are you training on before stepping into plant-scale work?
Quick example: before I let a big plant go to auto, I do a 10‑minute ‘point sanity check’ — confirm BACnet mapping/units on flow and temps, and force a valve to prove stroke so I’m not chasing a backwards 0–10 V. That alone saved me on a 1,100‑ton startup when CHWS was scaled like °C; if you can’t trend yet, a handheld logger or clamp meter is a decent stopgap. @OP I’m with you on SOO/trends, but double‑check units day one or you’ll be ghost hunting.
Biggest miss I see is bad input calibration — before anything else, clip a calibrated probe on CHWS and CWS and trim offsets; on a 1,200‑ton I found the supply RTD 1.8°F high and nobody could hit the commissioning delta‑T. Agree with @matthew9418 on the sanity check, and I also cap pump/tower VFD mins and tame PID gains so defaults don’t oscillate and create ‘phantom setpoints’.
And i run an end‑to‑end interlock check before auto — prove pump status, VFD faults, low‑temp cutouts, and tower fan fails, then simulate a short power blip to confirm the sequence recovers clean; “if it can’t recover in five minutes, it’s not ready.” It’s saved me from hours of whack‑a‑mole with HOA after turnover. If ops won’t allow a live blip, loop in @controls to do a controller reboot test and document expected states.
Verify actual stroke and fail position on the condenser bypass and CHW isolation valves — “graphics ≠ device” — on a 1,200‑ton I found a reversed actuator reading 0% but holding about 30% open, which wrecked delta‑T; if you can’t kill power to test fail, tag it and trend DP across the valve as a proxy. @raymond_eva84, you see that on tower gear too?
Before I touch water, I scan for BACnet device ID conflicts and priority-array hogs; duplicate IDs are “phantom setpoints” factories… I reconcile the SOO to the point list, make sure each enable/command maps to real I/O with sane scaling, and clear overrides before the “commissioning window.” If IT blocks discovery, loop them in early or use a temporary isolated switch to validate IDs offline.