Prediction of surface roughness and porosity in SLM-fabricated SS316L components using an artificial neural network approach

¹ Department of Mechanical Engineering, Faculty of Engineering & Technology, M.J.P. Rohilkhand University, Bareilly, India
² Department of Mechanical Engineering, Faculty of Engineering & Technology, Jamia Millia Islamia, New Delhi, India
³ Department of CSIT, Faculty of Engineering & Technology, M.J.P. Rohilkhand University, Bareilly, India
⁴ Department of Management Studies, Graphic Era University, Dehradun, India

^* Corresponding author: sharma_amit0408@yahoo.com

Published online: 7 January 2026

Abstract

Selective Laser Melting (SLM) is a sophisticated additive manufacturing process utilizing a high-energy laser to selectively melt layers of metallic powder according to a CAD design. The online monitoring and control of defects is a challenging task in SLM fabrication which is necessary to maintain the functionality and longevity of parts. No universally accepted model has been developed, for prediction of surface defects and porosity in SLM-fabricated parts. Image processing using artificial neural networks (ANN) has been successfully used in various fields such as in quality control, human computer interaction, remote sensing, and construction. In this paper, ANN model was utilized for estimation of porosity and surface roughness (Ra) of SLM-fabricated components. The SEM images were used as input for pattern extraction by image processing using ANN model. The integrity of the ANN model to predict the optimum process parameters for the minimum Ra and porosity was assessed. Both testing as well as training data sets showed lower values of the mean absolute error (MAE) & mean square error (MSE) at 0.9662 and 1.6296, indicating a well-trained ANN model. 2.45% and 3.07% average error was obtained between actual and predicted values of Ra and porosity. This research has developed a fast and accurate method for prediction of Ra and porosity to improve functionality and longevity of SLM-fabrication components.