Back to Article
Unified Data Lineage and Quality Governance Framework for Multi-Source Sensor Streams in Heavy-Duty Powertrain Manufacturing
Online Journal of Mechanical Engineering
| Vol 1, Issue 1
Table 1. Comparative Overview of UDLQ FrameworkArchitectures
| Architecture Type | Key Features | Limitations |
| Rule-Based Lineage + Threshold Quality (Model A) | Simple rule logic, low implementation cost, static quality thresholds | No real-time adaptation, high false-alarm rate (18–25%), no provenance depth, no drift handling |
| Batch ETL with Manual Lineage (Model B) | Improved data completeness, periodic quality checks, partial provenance | High latency (196 ms), no streaming support, manual lineage annotation required |
| Streaming Ingestion + Partial Governance (Model C) | Near-real-time ingestion (118 ms), automated completeness checks, partial provenance | No cross-source fusion, separate modules for each quality dimension, limited lineage depth |
| Proposed UDLQ Framework (Model D) | Unified ontology-based lineage, adaptive quality governance, 91.8% F1, 52 ms latency, 93.4% provenance coverage | Requires cloud infrastructure provisioning and initial ontology engineering effort |