Selective laser melting of metals optimized by deep learning
The prerequisite for exploiting the full potential of additive manufacturing (AM) is the rapid and cost-effective fabrication of defect-free components. This is no easy task, since the AM process is complex and controlling it challenging. Aside from feedstock characteristics, numerous process parameters must be carefully selected to yield the optimum parameter set. Each newly processed material usually requires the identification of the optimal set, a cost and time-consuming process, mostly conducted by trial and error. Here, we propose an alternative optimization strategy based on artificial intelligence (AI). Ti5553 components are manufactured using selective laser melting. A model to predict their density from experimental data is developed by means of deep learning. This supervised model utilizes an artificial neural network that enables the identification of the optimum parameter set via a specific recommender system. Our approach opens the way to a substantial decrease of resource consumption in additive manufacturing via AI-driven process optimization.