Gas turbine engine health monitoring (EHM) systems produce rich streamsof health-parameter residuals derived from gas path analysis (GPA),Kalman filtering, and related estimation frameworks. In current practice,these residuals are primarily treated as scalar magnitudes for thresholdcomparison or as inputs to probabilistic prognostic models. As a result,the temporal structure of residual trajectories—specifically theirdirection, persistence, and curvature—is often not explicitly formalizedor exposed for interpretation. This paper introduces the DSFB (Drift–Slew Fusion Bootstrap) StructuralSemiotics Engine as a deterministic, non-intrusive augmentation layerfor interpreting residual dynamics in gas turbine engine healthmonitoring. DSFB operates as a read-only observer over residual streamsproduced by existing EHM pipelines, transforming scalar deviations intotyped structural objects defined by drift, slew, regime-conditionedadmissibility envelopes, and grammar states. These objects enabledeterministic classification of residual trajectories into interpretablestates (Admissible, Boundary, Violation) and associated reason codes,providing auditable, human-readable structural context for degradationpatterns. The framework is instantiated for gas turbine applications and evaluatedon the NASA C-MAPSS (FD001, FD002, FD003) and N-CMAPSS datasets under aread-only protocol. Results demonstrate that DSFB can expose structureddegradation signatures and provide early structural indication relativeto threshold-based interpretations under controlled simulationconditions. These findings are presented as capability demonstrationsof interpretive expressiveness under calibrated configurations, not asclaims of universal performance or superiority over incumbent methods. DSFB does not replace existing GPA, EHM, or prognostic systems, and doesnot estimate latent physical states or remaining useful life. Its role isto augment these systems by providing deterministic structuralinterpretation of residual behavior that is otherwise aggregated,thresholded, or left uninterpreted. The architecture preserves allupstream system behavior through strict non-interference, enforced viaread-only interfaces and deterministic execution. The paper further outlines limitations, failure modes, and a stagedvalidation roadmap, including physics-informed consonance checks,formal verification of the interpretive state machine, andhardware-in-the-loop evaluation. The contribution is the formalizationof residual trajectories as primary interpretive objects in gas turbinehealth monitoring, enabling a complementary, audit-ready layer ofstructural reasoning within existing monitoring ecosystems.
Riaan De Beer (Fri,) studied this question.