Surface Roughness Prediction Using ANN Regression and Classification Model for S30C Alloy Metals Parts Manufactured by Laser Powder Bed Fusion

In Additive Manufacturing (AM), the Laser Powder Bed Fusion (LPBF) method are often used in creating detailed metal parts. This technique utilizes a high-powered laser to melt and fuse the layers of metal powder, allowing for the production of complex shapes with great precision. Although LPBF has many benefits, there are challenges related to the surface quality of the finished parts. Factors like current, scan speed, and line offset play a big role in determining surface quality. Variations in energy density during the process can also lead to uneven surfaces. To study these effects, 32 cube samples made of S30 alloy were produced using LPBF, with data taken from existing studies. A Supervised machine learning algorithm such as Artificial Neural Network (ANN), was used to predict the surface roughness of the metal parts made by LPBF. The accuracy of this ANN classification model was compared with an ANN regression model and experimental results. The ANN classifier was better at predicting surface roughness, which intern help to increase the surface characteristics of S30C alloy. The ANN classifier outperforms the ANN regressor in the tasks primarily because it is specifically designed for such tasks, with appropriate loss functions, output structures, and architectures that are better suited to making discrete predictions. The modeling and testing of these machine-learning algorithms were done using Python 3.2.