A comparative study of basic and ensemble artificial intelligence models for surface roughness prediction during the AA7075 milling process

This study reports on how ML algorithms are employed to investigate and predict surface roughness. Experiments were executed with a CNC milling machine, using AA7075 as part material and a “TiCN” coated tool. Feed rates per tooth, cutting speeds, cut depth, and cutting fluid were studied in response to roughness average (Ra) values. In the present study, Ra was measured with contact stylus tracing. Forty-two experiments were executed: thirty-three were used in all models training and nine in tests, and an additional experiment was carried out with diverse cutting parameters to validate the preferred models. This is the first study where thirteen ML algorithms, of which seven are basic and six are ensemble models, have been studied in the context of surface roughness. The study results showed that the voting regression model was the best model according to performance metrics ( R 2 = 0.97, RAE = 0.17, RMSE = 0.0325, MAE = 0.13, and RSE = 0.09) and deviation 5.66%. Manufacturing companies can employ the voting regression model to predict surface roughness to enhance manufacturing efficiency, by harmonizing cutting conditions values against surface roughness.