Keedian operations · Standard procedure

VAV loop hunting — damper and zone temperature oscillating around setpoint.

A VAV box's control loop is hunting — the damper command and the space-temperature error oscillating around setpoint during steady load instead of settling on it. It wears the modulating damper actuator with continuous travel, unsettles the zone (temperature drifting up and down, drafts as airflow swings), and pushes needless airflow swings back to the AHU's static and fan control. Caught early, a future failed-actuator no-cooling call becomes a planned retune or a scheduled actuator replacement.

01 Detail

Trigger, action plan, and escalation criteria.

optimization_metric hvac medium essential tier AI: hybrid

Related metric: vav_airflow_delivery
Trigger: oscillation_index > hunting_threshold during steady_window (raises loop-hunting)
Value drivers: Asset lifespan Customer experience Energy savings
Preconditions: Per-box damper command and space-temperature error mapped from the terminal controller (or the BMS under Model B), with an oscillation index derived over a settled window. Each box's size, control tuning, and a settled baseline recorded at onboarding, and the steady-load window that hunting is scored over defined, so a genuine load change or a morning warm-up ramp is not scored as instability. hunting_threshold and steady_window are configurable server-side attributes.
Human role: Primary at Managed/Optimized — Keedian reads the hunting signature (fast-small vs slow-wide), confirms steady load, and either retunes the loop remotely or refers a failing actuator. At Essential the customer's team sees the hunting box and retunes it themselves. A field visit is only for a genuine actuator or sensor failure, scoped separately.
Action plan: (1) Read the hunting signature before prescribing a fix — a fast, small oscillation is usually loop tuning or a noisy sensor; a slow, wide swing is usually an oversized box dithering near its minimum or a sticking actuator; confirm the load was steady over steady_window first. (2) Confirm the oscillation index has stayed above hunting_threshold during steady load for a sustained window, not through a load change or schedule transition. (3) Separate a retune from a device scope, so a controller gain is not adjusted where an actuator is failing, nor a device replaced where the loop simply needs tuning. (4) Tiered response — at Essential surface the hunting box for the customer; at Managed triage the signature and dispatch where a device is at fault; at Optimized retune the loop remotely (adjusting the box's control parameters with an audit trail) and refer only a failing actuator for a scoped field visit.
Escalation: (1) A box that begins hunting hard after being stable → often a failing actuator or a sensor going noisy; scope it before it fails open or closed, not at the next cadence review. (2) Instability on a critical zone (a boardroom, a trading floor) where the comfort swing is felt immediately → direct action rather than a queued retune. (3) Many boxes sharing a control strategy hunting together → a tuning-standard or sequence review carried into the MBR rather than box-by-box retunes.
Prevention: (1) Record each box's control tuning, size, and settled oscillation baseline at onboarding, and confirm the space-temperature sensor and damper actuator are healthy so instability is not a hidden device fault. (2) Define the steady-load window hunting is scored over and alert on a sustained elevated oscillation index during steady load, not on single corrections. (3) Trend the oscillation index over time so a slow rise (actuator wear) is separated from a fixed tuning offset, and keep a per-box stability log so a recurring sizing or default-tuning issue is visible.