Configure event detection at the edge. Define signal alerts, derived features, log watchers, and anomaly detection rules—then deploy them to your entire fleet with centralized management. Achieve 90-99% telemetry reduction while maintaining full diagnostic capability.
Threshold-based monitoring: battery_temp > 45°C for 10s, soc_level < 15%, coolant_flow < 3.0 L/min. Severity levels (critical, warning, info), duration requirements, and hysteresis to prevent flapping.
Custom metrics computed on-edge: pack_temp_slope, energy_efficiency, thermal_gradient. Use formulas like (max(cell_temp) - min(cell_temp)) to create compound metrics from raw signals.
Regex patterns on diagnostic logs: "ERROR.*HV_BATTERY", "WARN.*THERMAL". Context capture (50 lines before/after match), rate limiting (10 events/hour), and stack trace extraction.
ML-powered edge models: Isolation Forest, LSTM Autoencoders. Trained on historical data, deployed as ONNX models. Detect subtle deviations invisible to threshold rules.
Define what telemetry to upload when events trigger. Specify signals (K=100-400), time window (t-10min to t+5min), and compression level. Balance diagnostic value vs. bandwidth cost.
Centralized catalog of all deployed rules. Version control, enable/disable toggles, target cohorts (model, region, software version), deployment status per VIN.
Test rules on historical data before deployment. Simulate rule behavior, measure event frequency, estimate bandwidth impact, and validate thresholds with shadow mode.
Over-the-air rule updates to entire fleet. Staged rollout (10% → 50% → 100%), rollback capability, deployment progress tracking, and zero vehicle downtime.
Simple: battery_temp > 45. Compound: (battery_temp > 45 AND coolant_flow < 3). Duration: sustained for 10s. Hysteresis: exit when value falls below 42.
Rate of change: battery_temp_slope > 1.2°C/min for 5 minutes. Sudden jumps: delta(pack_voltage) > 5V in 1 second. Gradual drift: soc_accuracy deviation over days.
Sequence detection: charge_cycle_start → battery_temp spike → early_cutoff. Event correlation: HV_BATTERY error within 2min of THERMAL warning.
Upload only what matters. From 17.3B to 17.3M data points/day for 1,000 vehicles while maintaining full diagnostic capability.
Edge processing enables sub-second event detection. No cloud round-trip latency. Critical alerts trigger immediately.
Deploy new rules to entire fleet in minutes via OTA. Update detection logic without vehicle visits or firmware changes.
See how Sensor Studio enables sophisticated event detection at the edge while reducing telemetry costs by 90-99%.