TRACE-C Detects Relational Anomalies in Multi-Stream Telemetry.

This research introduces TRACE-C, a novel anomaly detection system designed for multi-stream operational telemetry. Unlike methods that only flag individual stream deviations, TRACE-C specializes in identifying "joint anomalies" where multiple streams exhibit unusual relationships or patterns, even if each stream's values remain within expected ranges. The system is auditable and strictly rank-calibrated. TRACE-C processes same-regime rolling median/MAD residuals through three distinct channels: a maximum normalized local sum, a Gaussian copula-form dependence contrast, and a worst standardized AR(1) innovation. The ranks from these channels are then aggregated using Fisher's method and compared against historical aggregates. Evaluations on Great Britain grid streams demonstrated TRACE-C's ability to rank significant events like Storm Atiyah highly, though further analysis revealed specific channel contributions. The research also highlights interpretive limits, emphasizing that the resulting p-values are selection quantities, not event probabilities, and that empirical rank counts are diagnostics rather than proofs of coverage.