Improving surface roughness in barrel finishing process using supervised machine learning

The Barrel finishing process is a finishing method applied for cleaning, polishing, improving surface quality, Deburring, and rounding corners of both metallic and non-metallic parts. There are Several factors affect the final surface integrity of the barrel finished samples such as initial surface roughness, piece length, operation time, and different abrasive materials (i.e. aluminum oxide, steel balls, and ceramic). On the other hand, each factor has different levels, and handling this amount of data to reach desired results is approximately impossible due to the “curse of dimensionality”. Machine learning is a promising method to pave this avenue for computing huge amounts of data and predicting the future state of the system. Accordingly, in this study, it is attempted to apply a supervised machine learning algorithm, an artificial neural network- to improve surface quality in the barrel finishing process. Python is used to code the program and extract several simulations and related graphs. Results show that time has the greatest effect on surface roughness, moreover, among the different abrasive media, steel balls have the best performance to improve surface roughness and the combination of 75% steel balls and 25% aluminum oxide has the effective effect. The simulation results have an acceptable compatibility with experimental ones.