| Abstract Scope |
The Q&P steel, a high-strength steel grade that simultaneously improves strength and ductility, is produced by the quenching and partitioning process. It consists of multiple microstructures including ferrite, martensite, and retained austenite. Each microstructure's composition and proportion affect the steel's physical characteristics, including its strength, elongation, and formability. Using nanoindentation testing and deep learning, this study proposes an innovative method to automatically identify the complex low-temperature transformation microstructures in Q&P steel that were previously impossible to distinguish, classify, or quantify. For deep learning models, a training dataset of SEM images displaying microstructures corresponding to hardness values obtained from the nanoindentation test was prepared. Then, using various deep learning network and hyperparameter combinations, a model with the highest prediction accuracy was obtained, with a focus on distinguishing between fresh and tempered martensite. Additionally, a novel approach that enhances prediction accuracy was investigated by applying image processing techniques to SEM images. |