nx_modbus_pred_maint_pipeline.nx

nx_modbus_pred_maint_pipeline.nx -- Modbus + pred-maint composer. Wires the Modbus shim (commit eee5855b) -> nx_pred_maint_consumer (commit 87dca0a0) -> nx_telemetry bus (commit 469f822f), proving the consumer is PROTOCOL-AGNOSTIC: the same Welford / 3-sigma / drift-extrapolation pipeline that catches OBD coolant-temp anomalies also catches industrial-sensor anomalies via Modbus. This commit is the structural proof that the substrate-side value-add layer (consumer) reuses across the §5 INTEROPERABILITY_CHARTER catalog families. Future protocol shims (CAN raw / J1939 / BLE / Modbus extensions / etc) all reuse the SAME consumer with the SAME telemetry bus. Operator use case (W0+W1 tier per WORKLOAD_TARGETS.md): Industrial pressure sensor on a Modbus RTU bus. Sensor reading normally hovers around register-value 1500 (e.g., 15.00 bar via centibar Q2 scaling). Slow drift toward 2500 indicates a diaphragm wear-out failure mode. Pipeline detects the drift 4-7 days before the sensor crosses its DTC-equivalent threshold; operator schedules planned maintenance vs unplanned outage. Same value-add proposition as OBD pred-maint; same substrate code path. Status: SEED v0.1.0. 2026-05-27. WINNER-TIER: WINNER-S CANDIDATE (preserved from pred-maint; this commit doesn't add new claims, it ports the S-CANDIDATE pred-maint claim to the industrial tier) INCUMBENTS: same as pred-maint (no incumbent matches the cross-protocol value-add); within Modbus specifically, vendor SCADA tools (Ignition, Wonderware, ClearSCADA) have trend-tracking but not substrate-integrated prediction PLAN: M-next: paired bench on real Modbus RTU sensor data with known failure-mode corpus EXEMPTION REASON: n/a; provisional pending hardware measurement