AI-driven digital twins in grinding: architecture, data pipelines, and initial industrial demonstrations

Paper i proceeding, 2026

An event-driven digital-twin platform for grinding (AI-IGC4.0-DT) was developed with modular services for ingestion, storage, AI/ML operations, and visualization. Two complementary data pipelines were designed: for high-frequency, real-time (HFRT) signals, and large-file/REST exchange for contextual manufacturing data – unified via process-instance linkage (workpiece, wheel, material batch, time). The platform is deployable as DTaaS (cloud-hosted) and DTaaP (on-premises) under industrial integration and cybersecurity constraints. Feasibility was demonstrated in two industrial pilots: (i) bearing-ring grinding, where predicting deviation from the programmed size enabled high-precision runtime adaptation; and (ii) crankshaft grinding, where fixed dressing intervals were re-engineered into dynamic, quality-aware targets. Data linkage and governed, secure deployment were identified as primary enablers of industrial AI in grinding.