Transferable Deep Learning Architecture for Simultaneous Prediction of Surface Quality and Dimensional Accuracy Across Multiple CNC Machining Centers

Abstract

This research arises from the need identified during the COVID-19 pandemic, where supply chain disruptions highlighted the importance of developing more adaptable manufacturing systems capable of redirecting production between different sites while maintaining quality standards. For this purpose, the application of artificial intelligence has become an imperative necessity.

This work presents the development of a transferable Deep Learning model for simultaneous prediction of surface roughness (Ra) and diameter deviation (Ddev) through process-measured signals (vibrations, forces, acoustic emission, among others) in CNC machining processes, applied across four different manufacturing sites (MS1-MS4) to implement a resilient distributed manufacturing strategy.