Cement Clinker Microstructure Segmentation Using Machine Learning

Cement is one of the most consumed construction materials. The quality of cement depends upon the clinker which is manufactured using limestone and aluminosilicate sources. The clinker microstructure comprises four phases: alite, belite, tricalcium aluminate, and aluminoferrite. Identifying different components of microstructures is a challenging task which in turn helps the researchers to control the quality of cement. In this work, we first create a dataset by segmenting alite and belite particles in cement clinker microstructure and use supervised machine learning methods to train models for identifying alite and belite regions in the microstructure image. We demonstrate the capability of machine learning models to accurately segment alite and belite regions. Further, we propose the methodology and guidelines to be taken care of while creating the datasets and training machine learning models for clinker microstructure segmentation tasks.