Skip to main content
Predictive maintenance

Know before your drivers do.

ChargeOS monitors 17 distinct fault and performance signals per connector, applies industrial-grade statistical methods, and alerts your team with specific, actionable recommendations — often before anyone on site notices a problem.

How it works

Four steps. Fully automatic.

From raw OCPP data to actionable maintenance recommendations — no manual configuration, no data science degree required.

01

Charger sends data

Every heartbeat, status change, meter value, fault code, and session event is captured automatically over OCPP.

02

ChargeOS learns the baseline

Over the first 20 charging sessions, the system establishes each charger's individual normal behavior.

03

Drift is detected

CUSUM and EWMA statistical monitors continuously compare current behavior to the baseline, detecting subtle shifts.

04

You get a recommendation

An alert appears in your dashboard with the triggering signal, the severity, and exactly what to do about it.

Health scores

One number. Full breakdown.

Every connector gets a health score from 0 to 100. The charger-level score is the minimum across all its connectors — a single weak connector brings the whole unit's score down.

85 — 100
Healthy
No action required.
65 — 84
Watching
Monitor closely.
40 — 64
Degrading
Schedule maintenance.
15 — 39
At Risk
Immediate attention recommended.
0 — 14
Critical
Connector may be auto-disabled.

Health scores start at 100 and decrease as faults and anomalies accumulate. Penalties range from 5 points (minor anomaly) to 40 points (ground fault). Scores recover over time when issues are resolved — this is not a one-way ratchet. Safety-critical penalties persist until an operator acknowledges the alert after a physical inspection.

Signal inventory

17 signals across four categories.

Every signal is independently monitored per connector. Each has a defined detection method, penalty weight, and recovery period.

Safety-critical
3 signals
Ground fault Auto-action
Electrical grounding failure. Connector taken offline automatically.
Overcurrent fault Auto-action
Current exceeding safe limits. Connector taken offline automatically.
Tamper event
Physical tampering or unauthorized access detected.
Hardware
4 signals
Thermal fault
Elevated temperature events from the charger's thermal monitoring.
Connector lock fault
Failure of the connector locking mechanism (cable retention).
Power meter fault
Energy meter malfunction — affects billing accuracy.
Reader fault
RFID card reader or authentication device failure.
Performance
4 signals
Energy delivery ratio
Ratio of actual energy delivered versus expected.
Session abort
Sessions ending within 60 seconds with no energy delivered.
Current draw variance
Unusual variation in current during charging sessions.
Session duration
Anomalous session length patterns.
Communication + stability
6 signals
Unexpected reboot
Charger restarted without a preceding reset command.
EV communication error
Failure in pilot signal communication between charger and vehicle.
Connectivity loss
Gap in heartbeat communication exceeding expected threshold.
Weak signal
Cellular or network signal strength below acceptable levels.
Heartbeat irregularity
Variance in timing between heartbeat messages.
Internal error
Charger-reported internal software or hardware error.
Detection methods

Statistical process control. Not a black box.

ChargeOS uses methods proven in industrial manufacturing for decades. Every alert is fully explainable — you always know what was detected, how it was detected, and what to do about it.

Threshold rules

Immediate safety response

Threshold rules fire on the first occurrence of a safety-critical event. No learning period. Active from the moment a charger connects. When a ground fault or overcurrent is detected, the connector is taken offline automatically and your team is notified.

7 signals monitored with threshold rules. Active from session 1.
CUSUM

Detecting gradual drift

CUSUM (Cumulative Sum) detects when a fault type is occurring more frequently than the charger's historical baseline. It accumulates the evidence of each small increase and signals when the cumulative drift becomes statistically significant. Used in manufacturing quality control for over 50 years.

13 signals monitored with CUSUM. Detects sustained drift within ~10 observations.
EWMA

Smoothed anomaly detection

EWMA (Exponentially Weighted Moving Average) tracks continuous performance metrics by maintaining a smoothed average that gives more weight to recent observations. A single unusual session does not trigger an alert, but a persistent trend does. The same method used for monitoring production line quality in semiconductor manufacturing.

4 signals monitored with EWMA: energy delivery, current draw, heartbeat, session duration.
Alert design

Every alert tells you what to do.

No log files. No mystery codes. Each alert includes the triggering signal, detection method, health impact, and a plain-language recommendation your field technician can act on.

CUSUM Medium severity
Connector lock fault frequency trending upward
Signal
Connector lock fault
Current rate
4.2x baseline
Health impact
92 → 71
Recommendation

Inspect locking mechanism for debris, wear, or mechanical damage. Check cable retention spring and solenoid operation.

Dashboard — instant push via SSE. Always active.
Slack — webhook integration. Configure in tenant settings.
Email — transactional email alerts to team members.
Learning period

20 sessions to learn. Safety rules from session one.

When a new charger is added, ChargeOS establishes that specific charger's individual baseline over its first 20 charging sessions.

During learning

Safety rules active immediately
Threshold rules active immediately
CUSUM and EWMA run silently
Health score shows "--"

After 20 sessions

Full statistical monitoring active
Health scores calculated and displayed
Drift detection with charger-specific baseline
All 17 signals fully monitored