WEB Data: Linkage of the world of Materials ScienceThursday (24.09.2020) 09:00 - 09:30 M: Modelling and Simulation 2 Part of:
Data is a core element in the big data era and data science as the forth paradigm of the sciecne discovery witnesses its ever-increasing application in nearly all the fields. Materials databases have a century of history in the world and a history of four decades in China. They have been paving the way for data mining on material data and data-driven material design to extract knowledge since the launch of Materials Genome Initiative. In this case, progresses on materials databases construction and data sharing are introduced. And we provide a machine learning approach and a few case studies based on the data from steel pruduction line and the experimental results collected from our team integrating with the data from the literatures. Data behave as the medium to transfer the information among the steel production processes so that the across processes calculation models to predict the microstructure and properties are successfully realized, which demostrates a prototype for the intelligent manufacturing of steel. The chemical composition selection, the characteristics of the raw materials, together with the processing parameters are discussed jointly to predict the performance of the structural materials by selecting the regression algorisms and constructing the model. We tried with the experimental dataset for appropriate model selecting, and multilayer perceptron model was found to work well for its distinguished prediction performance with high correlation coefficient and low error values. The accuracy of prediction by machine learning is evaluated. The predicted results agreed very well with the experimental one for each material, with the inaccuracy less than 3%.
Keywords: Data, database, data mining, intelligent manufacturing, regression