Quick start
Choose a path
Use a demonstration, retrieve real history, or import your own TLE or OMM records.
Run the analysis
Demonstrations run immediately. Historical retrieval loads and analyses the selected interval.
Read the result
Start with the summary, then review data quality, charts and any grouped candidate event.
Run the LEO candidate demonstration. It requires no account, file or external data request.
Analysis paths
Built-in demonstration
Synthetic histories show a LEO candidate, GEO station-keeping candidate, quiet control and sparse-history abstention.
Real historical data
Enter a NORAD catalog number and date range. The server retrieves authenticated Space-Track GP history and starts the analysis.
Your own records
Paste or import checksum-valid TLE history or validated OMM JSON for one catalog object.
Historical data
- Open Analyse a real satellite using historical data.
- Enter a numeric NORAD catalog number.
- Select the start and end dates. The current maximum is 31 calendar days, including both dates.
- Press Load real history and analyse.
- Wait for the result banner and screening summary.
The browser calls a same-site PHP endpoint. Space-Track credentials remain on the server and are not included in page source, browser storage or exported analysis files.
The CelesTrak snapshot control returns one current OMM record. A single record cannot support maneuver screening. Append it to existing history; replacement is disabled while a multi-record history is loaded.
Input formats
| Format | Requirements | Typical use |
|---|---|---|
| TLE | Valid line 1 and line 2 pairs with correct checksums, epochs and matching catalog identifiers. | Paste or import historical two-line element records. |
| OMM JSON | Records containing epoch, catalog identifier, mean motion, eccentricity, inclination, RAAN, argument of pericenter and mean anomaly. | Space-Track history, CelesTrak snapshots and reproducible imports. |
| Benchmark JSON | Cases containing orbital history, labelled event intervals and source citations. | Event-level validation metrics. |
Append and merge rejects mixed catalog objects and conflicting records at the same epoch. Exact duplicates are removed.
Understanding results
No clear candidate event
No eligible observation passed the current multivariate screening gates. This does not prove that no maneuver occurred.
Candidate event
One or more adjacent anomalous observations passed the configured element-change or corroborated state-residual gates.
Abstain
The workbench suppresses candidate reporting because history, identifiers, propagation or data quality are inadequate.
Weak, Moderate and Strong are internal screening tiers based on anomaly magnitude and corroborating signals. They are not maneuver probabilities.
The sticky result bar preserves the object, classification and data-quality status while you inspect charts and technical details.
Data quality
The workbench checks:
- Catalog identifier consistency
- Malformed or rejected records
- Exact and conflicting duplicate epochs
- Future or stale element epochs
- Oversized sampling gaps
- Successful SGP4 pair comparisons
- Enough sequential evaluation points after calibration
Detector method
- Parse and validate TLE or OMM history for one object.
- Infer or apply an orbit-specific detector profile.
- Propagate consecutive element sets to the later epoch using SGP4.
- Compute radial, along-track and cross-track state residuals.
- Combine state residuals with semi-major-axis rate, mean-motion, eccentricity-vector, inclination-vector and B* changes.
- Fit rolling robust baselines using only preceding eligible pairs.
- Apply magnitude and corroboration gates.
- Group adjacent anomalous epochs into candidate events.
If satellite.js is unavailable, the workbench can fall back to elements-only analysis. The detector mode used is displayed in the result and technical details.
Orbit profiles
Automatic mode classifies the orbit from its element history and applies profile-specific signal floors and thresholds.
| Profile | Why it differs |
|---|---|
| LEO | Drag, frequent catalog updates and dense constellation maintenance can affect normal variation. |
| MEO | Longer periods and different navigation or transfer-orbit behaviour require different residual scales. |
| GEO | Station-keeping produces small, structured changes and long orbital periods. |
| HEO | High eccentricity creates strongly varying orbital speed and geometry. |
Candidate events
Adjacent anomalous observations are grouped into one event rather than reported as separate maneuvers. An event can include:
- Start and end epochs
- Strongest observation
- Supporting observation count
- Net semi-major-axis change
- Screening tier and anomaly score
- Data-quality qualification
- Before-and-after orbit context
The 3D orbit comparison is approximate visual context and is not operational orbit determination.
Exports and reproducibility
Download analysis package is the recommended export. It contains the original input, normalized records, detector settings, method and workbench versions, quality checks, event results and deterministic run identifier.
- Evidence note for reports and review discussions
- CSV containing per-epoch signals and screening state
- JSON containing machine-readable analysis results
- Combined chart PNG
- Reproducibility package containing the complete run context
Validation tools
The collapsed Developer validation tools section provides:
- Eight deterministic regression checks for candidate, quiet, sparse, malformed, mixed-object and grouping behaviour
- Ranking across the built-in examples
- A labelled benchmark runner
The benchmark runner reports event-level precision, recall, F1, false events per object-month, detection delay and abstentions when supplied with a suitable labelled dataset.
Troubleshooting
Historical data load failed with HTTP 503
The same-site history endpoint is temporarily unavailable or cannot authenticate with Space-Track. The existing demonstration or imported input is preserved. Try again later or use a local TLE or OMM history file.
Run analysis appears to do nothing
When a result is current, the control is disabled and labelled as up to date. After changing input or settings, run the analysis and use the sticky result bar or View result button to reach the summary.
The workbench abstained
Open Technical details and review data quality. Common causes are too little history, mixed catalog identifiers, conflicting duplicate epochs, oversized gaps or unsuccessful propagation.
A file was rejected
Check TLE line length and checksum, or ensure OMM JSON contains the required fields. Confirm that all records belong to one catalog object.
SGP4 is unavailable
A browser extension, network filter or unavailable local vendor asset may have blocked satellite.js. Reload the page, verify the same-site asset route, or use elements-only mode. The result states which detector mode was used.
No candidate event was found
This means no eligible observation passed the selected gates. It does not establish that the object did not maneuver. Review the date range, data density, profile and operational context.
Accessibility
The workbench supports keyboard navigation, visible focus, responsive layouts, status announcements and a tabular alternative for the interactive charts.
- Use the skip link to move directly to the analysis result.
- Open Accessible chart data below the charts for the underlying values.
- All primary actions use native buttons, links, inputs and disclosure controls.
An internal WCAG 2.2 AA-oriented review and automated smoke test are included in Workbench 3.4. This is not a formal Section 508 certification or third-party Accessibility Conformance Report. See the accessibility statement.
Benchmark status
Workbench 3.4.1 includes a browser benchmark runner, a small public v0.1 seed manifest covering LEO, MEO, GEO and HEO examples, a same-site builder that retrieves OMM histories, reporting by orbit profile and a deterministic missing-data sensitivity check. The public manifest route serves that seed and runner format; it is not publication benchmark v1.0.
Frozen publication benchmark v1.0 contains 40 satellite-disjoint development, calibration and confirmatory cases, including 10 verified-quiet controls and 154 documented event intervals. Its one-shot confirmatory partition has already been consumed and is permanently closed to retuning as untouched evidence. Physics-ML v1.0 failed the predeclared gate and was not integrated; detector 0.4 remains the public scientific method. Review the benchmark status and evidence distinctions before citing any result.
Local dependencies and security
Plotly is served from a versioned local file and loaded on demand after analysis, avoiding the large renderer on the initial page load. satellite.js is pinned to version 6.0.2 and remains available at startup for SGP4 screening. Cesium has been removed; the optional 3D orbit context is rendered locally with Plotly.
- No analysis library is loaded by the browser from a third-party CDN.
- The page uses a restrictive Content Security Policy and blocks framing.
- Space-Track credentials remain outside the public web root and are never sent to the browser.
Institutional evaluation
This release is suitable for technical evaluation and research screening. It is not an operational space-domain-awareness service and should not be used as the sole basis for maneuver confirmation or operational decisions.
- Use labelled, independently sourced maneuver and quiet-control cases to evaluate performance.
- Retain the analysis package and deterministic run identifier for review.
- Confirm candidate events using orbit determination, additional catalogs and mission context.
Limitations
- The workbench screens catalog-level orbital changes and does not confirm burns, commands, intent or operational purpose.
- TLE and OMM records are fitted element products rather than direct maneuver measurements.
- Catalog updates, drag, sparse sampling and noisy fitting can produce unusual changes.
- Internal screening tiers are not calibrated probabilities.
- Follow-up should include source review, orbit determination, additional catalogs and mission context where available.