Blockchain-Enabled Digital Twin Model for Advanced Manufacturing Analytics and Simulation in Production Lines

Advanced manufacturing analytics requires a continuously updated simulation model of the physical production line that can predict defects, simulate scheduling alternatives, optimise quality parameters, and model energy consumption-all before the physical process commits to an outcome that is expensive to reverse. This paper proposes a Blockchain-Enabled Digital Twin (BEDT) framework combining a Hyperledger Fabric sensor-provenance layer, a physics-informed neural network digital twin, and an AI analytics engine, deployed across 18 manufacturing plants (automotive$\mathrm{n}=7$, electronics$\mathrm{n}=6$, precision components$\mathrm{n}=5$) evaluating 48,000 digital twin simulation runs (36,000 training/12,000 test, 75/25 chronological split) against 24,000,000 annotated sensor data points. The BEDT achieves simulation accuracy of 0.884 against physical production ground-truth, outperforming Digital Twin only (0.764; +15.7% [(0.884-0.764)/0.764]), Blockchain only (0.724; +22.1%), and rule-based MES (0.584; +51.4% [(0.884-0.584)/0.584]). Deployment outcomes: +34.2% defect prediction accuracy, +28.4% production scheduling efficiency, 84.4% quality simulation fidelity, and 184ms digital twin update latency. Sub-dimension performance: Defect Prediction (F1 0.924), Production Scheduling (0.884), Quality Control Simulation (0.844), and Energy Optimisation (0.784). Sensor calibration drift degrades accuracy to 0.784 (−11.3%) and model-plant divergence to 0.724 (−18.1%), each with a deployed mitigation.

Hali tarjima qilinmagan