2026 International Conference on Sensing, Measurement & Data Analytics in the era of Artificial Intelligence (ICSMD 2026)

Deep- and near-bit geosteering measurements provide complementary depths of investigation, but their interpretation can be corrupted by vibration, spatial misregistration, sensor drift, changing drilling mode, and intermittent telemetry. This paper develops a vibration-aware cross-scale sensing and intelligent-diagnosis framework that combines a deep-reading electromagnetic array, near-bit resistivity, directional gamma ray, inclination/toolface, and tri-axial vibration context. The method forms robust vibration features and physics-guided cross-scale innovations after depth, time, and toolface alignment. A data-quality gate prevents missing, frozen, saturated, or low-margin samples from appearing as healthy evidence. Residual groups are fused into an uncertainty-aware health state that distinguishes transient drilling dynamics from persistent measurement inconsistency and supports graded valid, watch, degraded, and invalid decisions. A normalized illustrative study compares nominal operation, a high-vibration transient, depth/time misregistration, and progressive sensor drift; it demonstrates the expected separation between vibration severity and persistent cross-scale disagreement without claiming field fault-classification accuracy. The framework is designed for low-bandwidth downhole deployment and provides a verification path based on hardware-in-the-loop replay, controlled fault injection, grouped well-level evaluation, and maintenance-linked evidence.

cross-scale sensing,downhole vibration,equipment health monitoring,geosteering,intelligent diagnosis,measurement integrity,multi-sensor fusion