Interface Overview

The Anomaly Detection section is organized into three tabs — Summary, Models, and Anomalies — each providing a different level of visibility into your anomaly detection setup.

The Summary tab is the default view when you open Anomaly Detection. It gives a fleet-wide picture of monitoring coverage, detected anomalies, and recent system activity across all your models.

Four cards at the top of the page show the most important metrics across all anomaly models:

Function Monitoring — shows how well your devices and assets are covered by anomaly models:

Segment	Meaning
Monitoring Enabled	Devices and assets included in an anomaly model with periodic scanning active
Monitoring Disabled	Devices and assets covered by a model but without periodic scanning enabled
Not Scanned	Devices and assets not yet included in any anomaly model

Fleet Stability — shows the health distribution across scanned devices and assets:

Segment	Meaning
With Anomalies	Devices and assets where at least one anomaly was detected
Healthy	Devices and assets that were scanned but no anomalies were found
Not Scanned	Devices and assets not yet included in any anomaly model scan

Detection Impact — shows the total number of anomalies discovered and their distribution over time:

Metric	Description
Anomalies Discovered	Total count of anomaly events detected across all models
From N scans	Number of scan executions that contributed to those discoveries
Timeline	Monthly heatmap — each cell represents one month; grey means not scanned, green means no anomalies, yellow/orange/red indicates increasing anomaly count

Operational Automation — shows alarm activity triggered by anomaly detection:

Element	Description
Alarms Escalated	Total count of ThingsBoard alarms created from anomalies
Severity bar	Breakdown by severity: Critical, Major, Minor, Warning

The Profile Coverage table lists every device and asset profile and summarizes anomaly detection activity across it:

Column	Description
Profile	Profile name and type (Device or Asset)
Items	Total number of devices or assets in the profile
Models	Number of anomaly models configured for this profile
Monitoring	ON if at least one model has periodic scanning enabled
Alerts	ON if at least one model has alarm creation enabled
Track Score	ON if at least one model is saving anomaly scores to ThingsBoard
Anomalies	Total anomaly count detected across all models for this profile

Each profile row has two action buttons:

• + Create — opens the creation wizard with this profile pre-selected.
• View > — opens the Models tab filtered to show only models for this profile.

Click All models in the top-right corner of the Profile Coverage section to open the Models tab with no profile filter applied — showing all models across all profiles.

The System Pulse panel on the right side of the Summary tab shows the status of the last 10 anomaly model scan executions. It gives you a real-time view of whether your models are running successfully and detecting anomalies as expected.

Each entry in the pulse shows the model name, execution time, and one of four status colors:

Color	Meaning
Green	Scan completed successfully — no anomalies found
Orange	Scan completed successfully — anomalies were detected
Red	Scan failed due to an unexpected error
Gray	Scan task was canceled

Click any entry in the System Pulse to open the full execution details for that scan run.

---

The Models tab lists all anomaly models across all profiles in a single table. Access it by clicking the Models tab in the top navigation:

Each row in the table shows:

Column	Description
Name	Model name and the profile it belongs to
Created	When the model was created
Status	Queued — waiting to be processed; Running — training in progress; Ready — trained and operational; Failed — training error; Canceled — build was canceled
Purpose	Detection intent chosen when the model was created (e.g. Quick Scan, Monitoring)
Monitoring	Whether periodic scanning is enabled
Alerts	Whether alarm creation is enabled
Track Score	Whether anomaly scores are being saved as telemetry
Anomalies	Total count of detected anomalies
Entities	Ratio of devices/assets with anomalies to total monitored

By profile — click the Device profiles dropdown to filter by a specific device or asset profile:

By status or configuration — use the filter chips below the search bar to quickly narrow results:

Chip	Shows
All	All models
Failed	Models where training failed
Monitoring ON	Models with periodic scanning enabled
No Monitoring	Models without periodic scanning
Alerts ON	Models with alarm creation enabled

Click any model row to open a detail panel on the right side. The panel shows a summary of the model’s configuration and performance:

Field	Description
Status	Queued — waiting to be processed; Running — training in progress; Ready — trained and operational; Failed — training error; Canceled — build was canceled
Purpose	Detection intent used to build the model — Quick Scan, Monitoring, Compare, or Health
Anomalies	Total count of anomaly events detected by this model
Peak Score	The highest anomaly score recorded across all devices and assets
Affected	Ratio of devices or assets with at least one detected anomaly to the total scanned by this model
Alarms	Total count of ThingsBoard alarms raised by this model
Monitoring	Whether periodic scanning is enabled
Alerts	Whether alarm creation is enabled
Telemetry	Whether anomaly scores are being saved as telemetry to ThingsBoard
Profile	The device or asset profile this model monitors
Entities	Number of devices or assets monitored by this model
Last Run	Time elapsed since the most recent scan
Created	Date the model was created

Open Model — navigates to the full Result view for this model where you can inspect detected anomalies in detail:

Run Scan — opens the Reprocess Task dialog to trigger an on-demand scan of historical data:

There are two ways to rename a model:

Option 1 — Three-dot menu. Click the ⋯ at the end of any model row and select Rename. The model name becomes editable inline in the table. Press Enter to apply:

Option 2 — Detail panel. Click any model row to open the detail panel, then click the model name in the panel header to edit it in place. Press Enter to apply:

Click the ⋯ at the end of any model row and select Delete. A confirmation dialog will appear before the model is removed:

---

The Anomalies tab provides a unified table of every anomaly detected across all your models. Use it to search, filter, and investigate individual anomaly events without opening each model separately.

Each row in the table shows:

Column	Description
Time	Start and end timestamps of the anomaly
Item	The device or asset on which the anomaly was detected
Duration	How long the anomalous behavior lasted
Score	Maximum anomaly score reached during the event
Impact	Anomaly Score Index — combines score with duration to reflect cumulative impact
Model	The anomaly model that detected this event
Alarm	The ThingsBoard alarm raised for this anomaly, if alarm creation is enabled

Three filters let you narrow the anomaly list:

Search by item name — type part of a device or asset name to filter results instantly:

Filter by profile — click the Device profiles dropdown to narrow results to a specific device or asset profile:

Filter by time range — click the date range field in the top-right corner to set the period. Quick presets (Today, Yesterday, Last 7 Days, Last 30 Days, Week, Month, Year) are available alongside a full calendar picker:

Click any anomaly row to open a detail panel on the right side. The panel shows the full context of the anomaly event:

Field	Description
Score	Maximum anomaly score during the event
Start / End	Exact timestamps of the anomaly
Duration	Length of the anomalous period
Impact	Anomaly Score Index value
Model	The model that detected this anomaly
Profile	The device or asset profile the item belongs to

Two action buttons are available in the panel:

Open Model — navigates directly to the Result tab of the model that detected this anomaly, with the relevant item and time range pre-selected:

Review — opens the Review tab of the model, navigated directly to the anomaly chart for the selected item so you can inspect the raw telemetry alongside the anomaly score: