Automatic fault detection, timing, and classification.
Detego automatically detects fault events in your COMTRADE recording and classifies them by type. You do not need to configure anything -- the detection and classification algorithms run automatically when a recording is loaded. The results appear throughout the app: as annotation markers on the waveform, in the dedicated Fault ID tab, in the AI analysis report, and in the readout bars of protection tabs (Distance, Overcurrent).
This page explains what the fault detection and classification results mean and how to interpret them. The underlying mathematics are covered in the linked theory pages.
Detego monitors all current and voltage channels for sudden changes that indicate a fault has occurred. Two key time points are identified for each fault event:
The moment when the fault begins. Detego detects this by looking for sudden changes in current or voltage using superimposed quantities -- essentially comparing each sample against the value from one power frequency cycle earlier. Under normal steady-state conditions, this difference is zero. When a fault occurs, the difference spikes sharply, providing sub-cycle detection accuracy.
If the superimposed method cannot be used (e.g., on de-energized channels), a fallback method based on RMS magnitude changes is applied.
The moment when the circuit breaker interrupts the fault current and the system returns to a stable post-fault condition. Detego detects clearance by looking for the last significant transient spike followed by a sustained period of low change -- indicating the system has settled. The approximate clearance is then refined using wavelet analysis to pinpoint the exact breaker-operation transient.
Clearance detection prioritises current channels over voltage channels, because voltage recovery dynamics (oscillations, inrush) can produce misleading late events. If the recording is voltage-only, all channels are used.
The difference between clearance and inception times gives the total fault duration, which includes the relay operating time plus the breaker interrupting time.
Power systems often use auto-reclose schemes: the breaker trips to clear a fault, waits, then recloses. If the fault was transient (e.g., lightning), the reclose succeeds. If permanent, a second fault occurs and the breaker trips again. This produces multiple fault events in a single recording.
Detego detects all fault events in the recording. When multiple events are found, annotations are numbered: "Inception #1", "Clearance #1", "Inception #2", "Clearance #2", and so on.
Once a fault is detected, Detego classifies the fault type using two complementary analyses:
The three-phase currents are decomposed into positive, negative, and zero sequence components. The ratios between these components directly indicate the type of fault:
The specific faulted phases are identified by comparing individual phase magnitudes against the three-phase average. Faulted phases show characteristic voltage dips and current rises relative to the healthy phases.
Classification is not a single snapshot. Detego scans the fault classification at half-cycle steps within a protected window that skips the first cycle after inception and the last cycle before clearance. This avoids DC offset settling at inception and breaker transient noise at clearance, both of which can produce spurious classification results. Each sample produces an independent classification, and consecutive identical results are merged into segments. Short-lived classification changes (under approximately 30 ms) are filtered out as transient noise, keeping only sustained fault type changes.
When the recording has enough pre-fault data, Detego uses superimposed (delta) quantities — the change in voltages and currents relative to pre-fault load — so load current does not dilute the fault signature. This makes classification more accurate on heavily loaded lines. When no pre-fault period is available (e.g. the recording starts at the moment of the fault), Detego falls back to classifying from total phasors and notes this in the popover.
If the fault changes type during the window — for example an SLG-B fault that evolves into a DLG-BC fault — the timeline captures both segments and the transition point. The classification is marked as evolving.
The classification appears as an orange badge on the info bar at the bottom of the viewer. The badge text reflects the timeline:
SLG-B.SLG-B → DLG-BC. A ◆ marker appears before the text.Unclear — click it to see the full timeline.Click the badge to open a popover with the full segment timeline, the classification method, and any contextual notes. The popover closes when you click outside it, press Escape, or click the x button.
The popover footer shows the classification mode that was used, such as "Delta sequences from V + I (highest accuracy)". This tells you two things:
Fault classification badge and popover
| Time | Duration | Type | Phases |
|---|---|---|---|
| 87→119 ms | 32 ms | SLG-B | B |
| 119→171 ms | 52 ms | DLG-BC | B, C |
Adjust the fault window if classification looks wrong
Detego classifies faults into the following types:
Transformer differential recordings
| Type | Designation | Description |
|---|---|---|
| Single Line-to-Ground | SLG (e.g. A-G) | One phase faults to ground. The most common fault type (~70-80% of all faults). |
| Line-to-Line | LL (e.g. AB) | Two phases fault together without ground involvement. |
| Double Line-to-Ground | DLG (e.g. AB-G) | Two phases fault together with ground involvement. |
| Three-Phase | 3P (ABC) | All three phases affected equally. Balanced fault. |
| Three-Phase-to-Ground | 3P-G (ABC-G) | All three phases affected with ground current. |
Fault ID Tab
Fault detection and classification results appear in several places throughout the app:
When a fault is detected, vertical annotation markers are automatically placed on the waveform viewer at the inception and clearance times. These markers show the channel name and event type (e.g., "Inception (IA)", "Clearance (IA)"). For auto-reclose recordings with multiple fault events, markers are numbered (e.g., Inception #1, #2). You can click on these markers to navigate to the exact time point.
The Fault Location tab uses both fault timing and fault classification. The detected inception and clearance times define the measurement window for impedance calculation, and the classified fault type (SLG, LL, DLG, 3P) determines which measurement loop is selected for the distance estimate.
The Overcurrent tab uses the detected fault inception and clearance to shade the fault duration on the TCC plot and highlight the fault period on the RMS Trend chart. The operating point is automatically positioned at fault inception to evaluate relay performance.
The Directional tab uses fault detection to automatically narrow the analysis window to the fault duration. It detects the fault on the residual current channel (or falls back to phase currents) and bounds the directional torque calculation to the fault period only, improving accuracy.
The Report Editor can include a fault events block that lists all detected inception and clearance times across every analog channel in a timeline table. This block is optional and can be toggled on or off in the report.
The AI analysis engine calls fault detection and classification as measurement tools during its analysis. The AI report includes fault timing, classification, measured fault currents, and commentary on relay performance relative to the detected fault.
Tips
How it works