Optimising CNC Machining Processes through Artificial Neural Networks: A Case study in a machine tool company

In modern manufacturing environments, production site efficiency is crucial to sustain competitiveness in an ever-changing market. The imperative to reduce rework operations is crucial to improve productivity, reduce operating costs and maximise resource utilisation. This requirement becomes particularly critical when machining large components, which can take a long time, sometimes several days, to complete. To address this challenge, our study presents two single-layer Artificial Neural Networks, each configured with different hyperparameters and number of neurons, designed to accurately determine the material removal volumes required to produce cast iron columns of different lengths. The first model predicts the amount of material to be stripped from the column guides in order to ensure a parabolic profile of the component. The second, on the other hand, predicts the amount of cast iron to be removed on the sides of the column in order to respect parallelism values. In a real industrial context, the algorithm was evaluated on a three-axis CNC machine at the company PAMA S.p.A., which is part of the AIDEAS European project consortium. The effectiveness of the solution was demonstrated by the high score value in predicting the removal parameters (score of both models ≈ 0.9) and was also validated by the previous experience of the company's production manager. © 2024, AIDI - Italian Association of Industrial Operations Professors. All rights reserved.