Conductivity and Hardness Prediction of Cu-Cr-Zr via Feature Augmentation on Geodesic Curve

Cu-Cr-Zr alloy has garnered significant attention for its outstanding performance in high-temperature and high-radiation environments, such as nuclear fusion reactors. However, accurately predicting its key properties, such as conductivity and hardness, remains a challenging task. To address this issue, a pioneering approach is presented, integrating machine learning and shape space theory. We propose a performance prediction method for Cu-Cr-Zr based on feature augmentation on Geodesic curves, referred to as F AGC. The methodology initiates with the extraction of image features from Cu-Cr-Zr using a neural network model. Subsequently, multiple image features are projected into a Pre-shape space, where Geodesic curves are constructed to seamlessly fit and generate features. Experimental findings reveal that, in predicting the conductivity and hardness of Cu-Cr-Zr alloy, both attain an R2 value of 0.916 and 0.923, respectively.