2026 Volume 65 Issue 3 Pages 239-244
3D printing methods such as powder bed fusion (PBF) and material extrusion additive manufacturing (MEX-AM) can fabricate complex three-dimensional shapes of metals and polymers, and their applications are expanding from rapid prototyping to end-product applications. However, in the case of lasers, many combinations of power and scan speed are possible, and in the case of heater nozzles, many combinations of temperature and scan speed are possible. Since the melting and solidification of materials occurs locally in a short time, high-throughput exploration of process parameters is crucial. On the other hand, there are machine learning methods such as Bayesian optimization that can explore parameters that satisfy the desired characteristics with a small number of trials. However, since fabricating test specimens in 3D printing is time-consuming, we proposed and demonstrated Intermediate Feature Values (IFV). This paper introduces examples of IFVs obtained by real-time observation: side images, top images, and spatter images.
