This thesis explores the application of the machine learning method physics-informed neural network (PINN) for the laser powder bed fusion (LPBF) process of pure copper. Given the process parameters, the accuracy of the predictions for relative densities of this method are compared to the traditional methods of a simple artificial neural network (ANN) and a linear regression. The results show the superiority of the PINN method in all tested scenarios, especially for complex and small data sets. Various data sets containing both literature data using red lasers and experimental data using a green laser are employed and show that a transfer of knowledge from red laser data to green laser data is feasible and beneficial. For red laser data from literature, the PINN method produces predictions with a mean squared error (MSE) of 4.58, as opposed to a MSE of 14.84 and 18.47 for a simple ANN and linear regression respectively. Using the transfer of knowledge for green laser data and reduced set of training data, PINN predictions exhibit a MSE of only 2.46, while a simple ANN and linear regression lead to a MSE of 7.29 and 5.19. These error values mean that for red laser data, the PINN method yields predictions that deviate 2.14 percentage points from the actual value and for green laser data the deviation is only 1.57 percentage points.

In addition, this thesis conducts a thorough literature review on the influence of process parameters on density, microstructure, and mechanical and electrical properties. This knowledge provides valuable insights for the further development of PINN for other outcomes, such as the microstructure and related functional properties. Furthermore, a test series for the LPBF process of copper using a green laser is presented for a laser spot diameter of 50 µm using an analytical approach focused on the geometry of the melting pools. This design of experiments serves as a first step for the systematic exploration of LPBF process parameters to achieve a broad analysis on their effects on density, microstructure, and functional properties.

This study pioneers in the application of PINN on the LPBF process of copper and through that, advances the research concerning the analysis of the in-situ influence through the modification of process parameters. The results help to further the LPBF process optimization in the face of lean process development with implications for the broader field of additive manufacturing.