Artificial Intelligence in the Microscopic Analysis of Mesophase Pitch

Abstract Mesophase pitch is a critically important precursor in the production of high-quality carbon materials such as high-modulus carbon fibers, synthetic graphite, and needle coke. Traditionally, the key parameter—the content and morphology of liquid crystal mesophase—is monitored by subjective visual analysis of microphotographs. This method lacks accuracy, reproducibility, and speed. In the present work, the introduction of artificial intelligence (AI) and machine learning is shown to ensure objective and rapid analysis of mesophase pitch; its benefits are outlined. The potential of deep learning methods—in particular, semantic segmentation and object detection—in automatic quantitative and morphological analysis of mesophase pitch on the basis of optical microphotographs is considered. The use of AI regression models to establish correlations between the pitch structure and the properties of the final product is also discussed. The introduction of artificial intelligence permits completely automated determination of the mesophase content, with objective and highly accurate results; detailed morphological analysis of spherulites (size, distribution, shape); quality monitoring and early defect diagnostics in real time; and the creation of digital twins of samples and predictive models for optimizing the industrial process. The switch from analysis based on expert assessment to systems based on artificial intelligence and machine learning constitutes a technological breakthrough in quality control of mesophase pitch. It permits a new level of standardization, accelerates research, and shortens the production cycle. That is essential to competitiveness in the market for advanced carbon materials.