Deep Multi-Layer Perceptron based Prediction of Energy Efficiency and Surface Quality for Milling in The Era of Sustainability and Big Data

Abstract

In advanced industries such as aerospace, medical and automotive, high precision machining is increasingly required for many parts made by difficult-to-cut alloys. Machine tool manufacturers respond to this demand by developing more advanced machine tools that have advanced sub-systems for attaining high-precision and wide flexibility, with the expense of energy efficiency. Unfortunately, worldwide primary energy resources continue to run out. Furthermore, GHG emissions mostly related to energy, remain to be a global issue with the ever-increasing economic expansion in many developed and developing economies. In the meantime, increasing use of sensors and Internet of Things (IoT) technologies in shop-floors set off a data explosion, warranting the use of emerging Deep Learning techniques to cope with “Big Data” reality of manufacturing. Therefore, in this study a Deep Multi-Layer Perceptron (DMLP) based algorithm for predicting surface roughness and specific cutting energy - major measures of precision and energy efficiency-, has been developed for slot milling of AL7075. Design of Experiment is used to collect the required data and train DMLP based model. The finalized prediction algorithm estimated quality and energy efficiency with 91.5% and 90.7% accuracy rates in slot milling, verified by additional machining and data collection.