TY - JOUR
T1 - In situ conductometry for studying the homogenization of Al-Mg-Si alloys and predicting extrudate grain structure through machine learning
AU - Österreicher, Johannes A.
AU - Živanović, Dragan
AU - Walenta, Wolfram
AU - Maimone, Stefan
AU - Hofbauer, Manuel
AU - Hovden, Sindre
AU - Tükör, Zuzana
AU - Arnoldt, Aurel
AU - Cerny, Angelika
AU - Kronsteiner, Johannes
AU - Antić, Miloš
AU - Zickler, Gregor A.
AU - Ehmeier, Florian
AU - Mikulović, Milomir
AU - Kunschert, Georg
PY - 2024/6/17
Y1 - 2024/6/17
N2 - In industrial practice, no sensors capable of obtaining microstructural information in situ during thermomechanical processing of Al alloys are commonly employed. Inductive electrical conductivity measurement is safe, inexpensive, and capable of acquiring valuable information about precipitation and dissolution processes. However, commercial eddy current sensors work only at low temperatures near room temperature and are thus not suitable for in situ conductometry during heat treatments of Al alloys. We designed a high -temperature eddy current sensor and performed in situ conductometry during the homogenization of six Al -Mg -Si wrought alloys, three of which are experimental recycling -friendly alloys with increased Fe content. The results are interpreted with regard to microstructural investigations, and the advantages and limitations of our approach are discussed. As a proof -of -concept, we show how the conductivity curves and extrusion process parameters can be combined to predict final extrudate grain structures using machine learning. To achieve this, we employed finite element simulation of extrusion coupled with microstructural simulation over a wide parameter range, validated by extrusion experiments and metallography, and trained a feedforward neural network. We believe our interdisciplinary approach can lead to improvements in the industrial processing of Al wrought alloys.
AB - In industrial practice, no sensors capable of obtaining microstructural information in situ during thermomechanical processing of Al alloys are commonly employed. Inductive electrical conductivity measurement is safe, inexpensive, and capable of acquiring valuable information about precipitation and dissolution processes. However, commercial eddy current sensors work only at low temperatures near room temperature and are thus not suitable for in situ conductometry during heat treatments of Al alloys. We designed a high -temperature eddy current sensor and performed in situ conductometry during the homogenization of six Al -Mg -Si wrought alloys, three of which are experimental recycling -friendly alloys with increased Fe content. The results are interpreted with regard to microstructural investigations, and the advantages and limitations of our approach are discussed. As a proof -of -concept, we show how the conductivity curves and extrusion process parameters can be combined to predict final extrudate grain structures using machine learning. To achieve this, we employed finite element simulation of extrusion coupled with microstructural simulation over a wide parameter range, validated by extrusion experiments and metallography, and trained a feedforward neural network. We believe our interdisciplinary approach can lead to improvements in the industrial processing of Al wrought alloys.
KW - Aluminium
U2 - 10.1016/j.matdes.2024.113070
DO - 10.1016/j.matdes.2024.113070
M3 - Article
SN - 0264-1275
VL - 243
JO - Materials & Design
JF - Materials & Design
M1 - 113070
ER -