Alarm Acknowledgement Latency

Definition

Alarm Acknowledgement Latency (AAL) is the elapsed time, in seconds, between the instant an alarm is asserted by CHLORA and the instant a duty operator acknowledges it at the console. Acknowledgement is the deliberate human action that confirms the alarm has been seen and ownership accepted — it precedes any corrective work. Each alarm yields one AAL sample; the metric is reported as the P95 of latencies over the trailing hour so that worst-case responsiveness, not the easy median, governs the thresholds.

Why it matters

AAL is the facility's measure of how long a hazard sits unseen. A drifting reactor temperature or a containment pressure excursion does not wait for a human; every second between assertion and acknowledgement is a second of unmanaged risk. A rising AAL signals operator overload, alarm flooding, or staffing gaps — the conditions under which a critical alarm is most likely to be missed. It is a core human-factors metric on the Operations Dashboard and a direct input to shift-staffing decisions.

Formula

AAL is a direct per-alarm duration; the reported figure is the trailing-window P95:

aal_i = t_ack,i − t_assert,i        (seconds, per alarm i)

Reported value (governs thresholds):
  AAL = P95( { aal_i : asserted in trailing 60 min } )

Also published:
  AAL_median, AAL_max, n_unacked (alarms open > crit threshold)

Severity weighting:
  P1 alarms carry a 30 s hard ceiling regardless of window P95;
  any P1 unacknowledged > 30 s triggers AUDIBLE_REASSERT and
  auto-pages the on-call lead (Mycelial Data Fabric channel).

Latency measured at the console event bus; clock-sync to MDF
master ± 50 ms, so sub-second jitter is not material.

Inputs

Units & Scale

AAL is measured in seconds to one-second precision. The headline figure is the trailing-hour P95; median and max are published for context. It is not averaged across consoles — each console's P95 is reported separately and the facility figure is the worst console's P95, because a single overloaded console is the operational risk. A LOW_SAMPLE flag is attached when fewer than ten alarms occurred in the window.

Sampling & Source

• Sampled per alarm; the reported P95 is recomputed continuously over a trailing 60-minute window.
• Assertion and acknowledgement events both timestamped on the CHLORA console event bus.
• Clocks synchronized to the Mycelial Data Fabric time master within ± 50 ms.
• Stale / low-sample handling: < 10 alarms in window → LOW_SAMPLE; any P1 unacknowledged > 30 s → audible reassert and auto-page.

Thresholds

Recent Trend

Worst-console P95 acknowledgement latency in seconds, last 14 windows:

Interpretation Guidance

Related Metrics

Related SOPs

• SOP Library — alarm acknowledgement and console-staffing procedures.
• Alarm-flood rationalization & supervisor takeover on CRIT — see the SOP Library.
• Alarm assertion, audible reassert, and on-call paging — Monitoring Systems.