The Process
From transformer to scheduled repair in four steps
Magnefy works by reading the physics of the transformer itself. The electromagnetic signature encodes what is happening inside the winding and core before any thermal or chemical indicator registers the same condition.
Step 1
Continuous electromagnetic measurement
A Magnefy sensor node installs at the transformer in under an hour. It uses non-invasive current and voltage probes on the secondary terminals. No primary-side access, no oil sampling, no downtime for the asset.
The node samples all three phases at 10-second intervals and captures harmonic content through the 25th order. That resolution is necessary to see the early-stage distortion patterns that winding faults produce, which temperature monitoring cannot detect until the fault is advanced.
Transmission uses cellular or the utility's existing data network. The node has 48 hours of local buffer storage in case of connectivity interruption.
Non-invasive installation
Clips onto existing low-voltage secondary terminals. No primary-side access required, no oil sampling equipment needed.
Full-spectrum capture
Harmonic analysis through the 25th order at 10-second intervals. The full frequency domain, not just RMS values.
Step 2
Fault signature analysis against learned baseline
Once a transformer is installed, Magnefy builds a unique baseline for that asset under its normal load profile and ambient conditions. The baseline captures what the transformer's healthy EM signature looks like across all operating states.
Each incoming reading is compared against that baseline across 40+ fault signature dimensions. A developing fault shifts the harmonic ratios in characteristic ways. Winding insulation degradation raises the third harmonic. Core anomalies distort the magnetization current envelope. The model detects these shifts early, when they are consistent but still subtle.
Step 3
Fault type classification and severity scoring
Not every EM deviation is a fault. Magnefy's classification layer distinguishes fault-pattern deviations from load-driven variation, switching transients, and measurement noise. The model was trained on labeled field data with known fault outcomes to reduce false positives.
When a pattern matches a fault class, the system assigns a severity level (watch, caution, action) based on the rate of deviation change and the historical progression of that fault type. A slow-developing insulation fault starting at watch will escalate to caution on a predictable timeline.
Fault taxonomy covered
- Insulation class Winding insulation aging, moisture ingress, thermal deterioration
- Winding class Inter-turn short, open circuit, winding displacement
- Core class Core lamination fault, increased core loss, shorted core packet
- Thermal class Cooling system failure, radiator blockage, fan motor fault
- Connection class Loose bushing connection, tap changer wear, terminal oxidation
Step 4
Structured alert delivery to your existing tools
Magnefy does not add a new screen to monitor. When a fault alert fires, it goes to whichever endpoint your team already uses: SCADA via OPC-UA, CMMS via REST webhook, or a daily email digest to maintenance coordinators.
Each alert payload carries the asset ID, fault classification, confidence score, severity level, predicted event window, and recommended action. Enough information to create a work order and schedule a crew without any additional research.
Severity changes (watch to caution, caution to action) trigger new alerts automatically. Maintenance coordinators do not have to re-check dashboards as a fault progresses.
Walk through the process on your fleet
We review your transformer fleet profile and show you what Magnefy would surface on your specific assets. No generic pitch.