Deciphering Material Histories: Estimating Exposure Temperature and Fracture Causes through Microscopic Imaging

抄録

In the evolving landscape of material science, the intersection of microscopic imaging and machine learning is proving to be a game-changer. This presentation seeks to illuminate how these modern technologies can substantiate and even surpass insights previously reliant on human expertise. The first segment explores our ground-breaking methodology for estimating exposure temperatures in power plant piping. We have achieved evidence-based temperature estimates with a remarkable accuracy of ±10 degrees by utilizing machine learning algorithms to analyze precipitates on metal surfaces. Moreover, we identify the critical image features that are most contributory, providing quantifiable evidence to support expert assessments. The second part focuses on discerning the causes of metal fractures. Using machine learning, we delve into identifying and justifying image features that can effectively infer the reason for material failure. These methods empower researchers with evidence-based arguments, previously dependent on experiential interpretation, thereby revolutionizing material analysis and diagnosis. Real-world case studies will be presented to demonstrate these technological advances' practical applications and transformative potential.