Study compares physics and AI approaches for crack detection in powder compaction
Researchers at the Free University of Bozen-Bolzano, Italy, have compared physics-based simulation with machine learning (ML) for detecting cracks during powder compaction, concluding that the two approaches are best viewed as complementary tools rather than competing technologies. Crack formation during compaction remains a challenge in Powder Metallurgy, the authors explain, with defects compromising the mechanical integrity of sintered components. Finite element modelling can predict where cracks are likely to occur by analysing stress development within the compact, but the simulations are too computationally intensive for routine production monitoring. Machine learning, meanwhile, can analyse press sensor data in near real time, although its predictions are often criticised for lacking a clear physical explanation.
Read more.



















