ANOMALY INVESTIGATION · {{kind==='imported'?'LOCAL CSV':'SYNTHETIC DATA'}}
Find the event.
Show the evidence.
Compare timing, separate channel-specific changes, and keep the evidence.
Import your own CSV · stays on this device
Import non-sensitive CSV locally. No file is uploaded. Format: channel,second,value — one reading per channel per whole second; missing seconds remain gaps.
{{importError}}
Local file: {{importName}}
Up to 8 channels, 20,000 readings, 2 MB. Seconds 0–3600, expected cadence 1 Hz. Choose a threshold suitable for your data.
Each channel is scaled to its own observed range for timing comparison. Compare numerical values below.
Compare the evidence
| Channel | Early mean | Late mean | Change | Missing | Jump at | Source |
|---|---|---|---|---|---|---|
| {{e.name}} | {{fmt(e.before)}} | {{fmt(e.after)}} | {{fmt(e.delta)}} | {{e.missing}} | {{e.jump.t===null?'—':time(e.jump.t)}} |
Arbitrary units. Early and late halves split at the time midpoint. Gaps are excluded from means, counted separately, and never joined in the plot.
INVESTIGATION BRIEF · RULE-BASED
{{report.title}}
{{report.observation}}
Data quality
{{report.quality}}
Next checks
{{report.next}}
Demonstration rule: adjacent change above {{threshold}} signal units; shared timing within two seconds. These are illustrative settings, not engineering alarm thresholds.
Source records · {{focus}}
| Record | Time | Raw value | Source line |
|---|---|---|---|
| {{r.id}} | {{time(r.t)}} | {{r.value}} | {{r.line}} |