Machine-Learning-Enhanced Prediction of Track Defects in Multilayer Laser Melting of Inconel 718
Adithya Bharath Shah, George Antony, V S Chris, Yaseen Ahammad, Kurian Antony, Jina Varghese and Bibin Jose
In this paper multilayer fabrication of Inconel 718 using SLM, a widely used additive manufacturing technique was conducted. The quality of the fabricated tracks is dependent on laser power, laser speed and spot diameter. The effects of these parameters during SLM process were investigated by numerical analysis as well as by experimentation to determine the optimized set of parameters. However, process optimization using experimentation and simulation is expensive as well as time consuming. Therefore, in addition, two machine learning models have been trained to predict the width and height of the fabricated tracks based on the parameters. By numerical analysis, the variation in temperature distribution was determined by varying these parameters and the effect of these parameters was validated experimentally. The effect of these parameters on the geometry, morphology of the track was analyzed by utilizing tools such as macroscope, SEM and EDAX. Among the two ML models, linear regression models demonstrated superior predictive accuracy with less error compared to RNN models.
Keywords: SLM, Inconel 718 powder, Machine learning, SEM, FEA, RNN
