Control-Relevance-Aware Sensor Integrity in a Position-Controlled DC Motor: Surrogate Evaluation and an Exploratory Direct-Simulation Crossover Test

Control-relevance-aware sensor integrity (CRASI) combines heuristic relevance, common-evidence subtraction, finite trust retention and bias tracking. We evaluate an internally specified 2520-episode DC-motor surrogate campaign calibrated from Proteus characterization. In the held-out modeled 70% load domain, CRASI reduced integral absolute error (IAE) versus residual-plus-uncertainty by 37.6% for encoder bias, 19.9% for drift, 15.9% for current under-read and 4.4% for current bias. Fixed control performed better for drift/current bias, under-read matched fixed, and dropouts were effectively null; adaptive Huber was 2.0% better for drift. Revision diagnostics identify comparator-dependent attribution, simultaneous-fault blindness, current self-masking, inactive position-uncertainty derating, and no hard-safety superiority. Matched-observer, detection/recovery, parameter-sensitivity and guard-stress analyses therefore restrict mechanism claims. A revision-era direct Proteus audit is not treated as validation because its acquisition/analysis gates were amended after partial outcome inspection and protection trips materially affected several 50% cells; the audit is retained only as a transparent research record and is not used as evidence for the manuscript conclusions. Findings support selective tracking comparisons under modeled sensor/plant confounding, while fresh pre-specified direct Proteus/HIL/physical validation remains necessary.