The growing complexity of next-generation aircraft architectures developed to support sustainable aviation demands pervasive sensing layers that stream high-rate data to Structural Health Monitoring (SHM) systems for diagnostics, prognostics, maintenance planning, and safety assurance. Existing SHM frameworks focus on inferring structural integrity while assuming nominal sensor behavior, leaving the sensing layer itself largely unevaluated. This assumption becomes problematic when sensor faults and progressive structural degradation produce similar signatures in the measured response. Under these conditions, diagnostic and prognostic logic becomes implicitly conditioned on a potentially corrupted measurement space, increasing the risk of false alarms, inconsistent damage localization, and delayed or missed detection of evolving structural faults. This work introduces a real-time digital twin of the sensing network that assimilates onboard structural and sensor data, learning a self-consistent physics-informed representation of the sensor layer. The twin enables autonomous distinction between structural and sensing anomalies, recovery of sensor-corrupted signals, and continuous probabilistic estimation of sensor reliability. The digital twin combines a reduced order representation of the coupled structure-sensor system, an optimal sensor placement strategy, and three scientific machine learning modules that jointly classify the source of anomalous measurements, recover corrupted observations through projection onto physically admissible manifolds, and assign each sensor a continuously updated reliability score formulated as a probability-calibrated trust metric. By treating the sensing network as an active component of the monitoring loop, the twin delivers real-time corrected signals and quantified measurement reliability, ensuring a validated information flow to downstream SHM logic. The approach is demonstrated on a carbon fiber wing panel subjected to simultaneous structural damage and sensor degradation. The results show that the twin can accurately discriminate between structural and sensing faults, restore strain measurements to physically consistent responses reflecting only the true panel state, and track the reliability of individual sensing channels in real-time.