Deep learning revealed statistics of the MgO particles dissolution rate in a CaO-Al2O3-SiO2-MgO slag

Accelerated material development for refractory ceramics triggers possibilities in context to enhanced energy efficiency for industrial processes. Here, the gathering of comprehensive material data is essential. High temperature-confocal laser scanning microscopy (HT-CLSM) displays a highly suitable in-situ method to study the underlying dissolution kinetics in the slag over time. A major drawback concerns the efficient and accurate processing of the collected image data. Here, we introduce an attention encoder-decoder convolutional neural network enabling the fully automated evaluation of the particle dissolution rate with a precision of 99.1%. The presented approach provides accurate and efficient analysis capabilities with high statistical gain and is highly resilient to image quality changes. The prediction model allows an automated diameter evaluation of the MgO particles' dissolution in the silicate slag for different temperature settings and various HT-CLSM data sets. Moreover, it is not limited to HT-CLSM image data and can be applied to various domains.