In the scope of understanding the factors controlling the defects that commonly appear in the microstructures of materials produced by Additive Manufacturing (AM), a first step is the prediction of the thermal history at which the material is subjected by means of different printing parameters. The extremely large amount of different combinations of printing parameters (laser power, hatch distance, laser scanning speed…) adds to the different printing geometries which also can vary the influence of such printing parameters. Its computation by means of Finite Element Modelling is the logical choice, but is time consuming and cannot be used in large computational optimization processes.

This work therefore aims at developing a fast and accurate predictive methodology to calculate temperature history at during the AM process. First, a Finite Element Method (FEM) model was developed to extract thermal profiles at key points for Selective Laser Melting processes, optimized for computational efficiency for all the different printing parameters combinations. Subsequently, different Machine Learning (ML) methodologies are tested as predictive models to determine the most robust approach. The models employed include several three based methods and Deep Learning methods (i.e., Neural Networks). These models were fed with the dataset, yielding optimal predictive outcomes for both ERT and FNN with very good performance. It demonstrates that it is viable to develop models based on ML for predicting thermal profiles in AM using as input the printing parameters. It also determines the performance difference among several of the ML models. This has led finally to a predictive model that offers quasi-instantaneous predictions with minimum error as compared to the computationally expensive FEM simulations.