IIoT integrated system for monitoring and maintaining the performance of a part-error evaluation model by considering machine tool thermal drift

Thermal deformation of milling machine tools (MMT) causes milling cutter drift (MCD) that results in part-error. Finite element (FE) simulations and machine learning (ML) models are used for thermal-error prediction. The FE-simulations are computationally expensive, while ML-models demand drift measurements. Hence, the present work proposes a novel IIoT-based framework that enables an energy-balance-based thermo-mechanical FE model, which eliminates the need to account for geometry variation due to changing thermal conditions, and is deployed in the cloud to acquire real-time hotspot temperatures for MCD simulation. The simulated MCD is validated against the milled part error, eliminating the need for direct drift measurement. Validated simulations generate data for an ML-model (trained in the cloud-module) that is deployed on the edge-module for real-time prediction. ML-model’s reduced inference time (~0.5s) against FE simulations (~3 h) supports the proposed approach’s potential for real-time IIoT-based lifecycle maintenance. Demonstration on an indigenously developed MMT with nine-inputs showed strong agreement between the part error and simulated MCD, with a mean absolute percentage error (MAPE) of 9.7%, whereas ML-model predicted part-error with 5.98%. Proposed correlation-based strategy reduced the number of temperature inputs from nine to three that reduced sensor requirements by 66.7% with an increase in MAPE by only 0.42%.