Diagnose a sensor disagreement

Enter redundant readings, reliability priors, and probe costs. DEIC estimates the state, names the most suspect sensor, and recommends the next confirmation without forcing a conclusion.

Sensor	Reading	Reliability	Probe cost

Unit

Expected noise (+/-)

Warning threshold

Critical threshold

Confirmatory probes

This worksheet demonstrates bounded inference and query selection. It is not certified control logic and must not directly command machinery, vehicles, clinical equipment, or safety systems.

Estimate state without trusting every sensor equally A thermal-pack example is loaded. Run the diagnostic to inspect the decision trace.

Diagnostic result

ABSTAIN

Estimated state

Consensus

Confidence

Suspect sensor

Conflict weight

Sensor evidence map

Next best confirmation

Auditable sensor trace

Sensor	Reading	Reliability	Deviation	Leave-one-out	Suspect score	Probe value

The estimator uses a reliability-weighted median, robust dispersion, leave-one-out consensus, explicit threshold uncertainty, and value-per-cost probe ranking. Real deployments require calibrated noise models, temporal dynamics, authenticated telemetry, and domain validation.