Anomaly Detection watches for behavior that’s unusual for that specific device, rather than checking against a fixed number. Where an Alert Rule fires when a metric crosses a threshold you set (e.g. “temperature > 90”), Anomaly Detection compares each metric against that device’s own recent history and flags it when it drifts away from its normal pattern – useful for catching problems you wouldn’t know to set a fixed threshold for. The two work well together: keep Alert Rules for known hard limits, and let Anomaly Detection catch the unexpected.
This page is available to integrator accounts; client accounts don’t have access to it.
What Runs Automatically
A set of detection rules run continuously in the background, with no configuration needed:
- Statistical Anomaly – compares each metric against its own time-of-day baseline, and distinguishes a brief blip from a sustained deviation.
- Rate-of-Change Trend – flags a metric that’s been consistently drifting in one direction over the past week.
- Predictive Forecast – for a metric with a known critical value (e.g. a temperature limit), projects roughly how many days out it’s headed for that limit at its current trajectory.
- Multi-Metric Correlation – recognizes combined patterns that mean something specific together, even if no single metric alone looks alarming: a Fan Failure Signature (temperature up + fan speed down), Laser Degradation (power draw up + laser output down), Thermal Overload (temperature, CPU, and power draw all up together), and a Network + Performance signature (latency and CPU both up).
- Health Score – a single 0–100 score per device combining every metric’s deviation, with a notable drop logged on its own.
- Anomaly Fingerprinting – recognizes when the same pattern of anomalous metrics has happened before, and flags it as a recurring issue after it repeats a few times.
Anomaly History
The table at the bottom of the page shows the last 100 detected events across your fleet – when, on which device, and what was actually flagged. Click a device’s name to jump straight to its page.
Detection Settings
Fine-tuning individual thresholds and cooldowns (which metrics are watched, how sensitive each one is) is managed by Axion Displays. If a particular metric is firing too often or not sensitively enough for your fleet, let support know which metric and what you’re seeing – they can adjust it.