The material's microstructure is the output of the material composition and production process;simultaneously, it is the key factor in the material's performance and plays an important role in the material dev...The material's microstructure is the output of the material composition and production process;simultaneously, it is the key factor in the material's performance and plays an important role in the material development cycle.The digital features of microstructures that can be recognized by a computer are obtained through the quantitative description of the material's microstructure, and a mathematical model is used to associate them with the composition, process, and properties.Combining the evolution mechanism of microstructure features can help quickly to predict microstructure properties, which can be used to optimize the composition and process, accelerate product development iteration, and solve the long cycle problem of traditional research and development mode while lowering the high cost of high-frequency trial production.This paper introduces the development and application of the combination of image knowledge, computer knowledge, and expert experience in the quantification of microstructures and the practical cases of recent studies of the Mask-Regions with Convolutional Neural Network Features(RCNN) method combined with machine learning in the digital research and development of heavy plates in Baosteel, and prospects the potential application of the quantification of microstructures of materials.展开更多
文摘The material's microstructure is the output of the material composition and production process;simultaneously, it is the key factor in the material's performance and plays an important role in the material development cycle.The digital features of microstructures that can be recognized by a computer are obtained through the quantitative description of the material's microstructure, and a mathematical model is used to associate them with the composition, process, and properties.Combining the evolution mechanism of microstructure features can help quickly to predict microstructure properties, which can be used to optimize the composition and process, accelerate product development iteration, and solve the long cycle problem of traditional research and development mode while lowering the high cost of high-frequency trial production.This paper introduces the development and application of the combination of image knowledge, computer knowledge, and expert experience in the quantification of microstructures and the practical cases of recent studies of the Mask-Regions with Convolutional Neural Network Features(RCNN) method combined with machine learning in the digital research and development of heavy plates in Baosteel, and prospects the potential application of the quantification of microstructures of materials.
