A strategy combining machine learning and multiscale calculation to predict tensile strength for pearlitic steel wires with industrial data

Using Machine Learning to Predict Oxygen Evolution Activity for Transition Metal Hydroxide Electrocatalysts

“…Density functional theory provides a path to indirectly infer the OER activity trends and facts by calculating the adsorption energetics of different intermediates, but it is difficult to directly give a quantitative prediction of OER activity, especially for the large potential candidate space comprising the chemical element, phase, morphology, and so on. Machine learning (ML) provides a data-driven methodology to capture the complex relationship for multi-dimensional factors, thus accelerating the exploitation of new materials. − For example, ML has recently been exploratively used in OER overpotential prediction of single-atom catalysts and Ni–Fe–Co–Ce oxide catalysts and for evaluating the chemistry factors of OER activity. , Despite the great potential, the use in transition metal hydroxide catalysts under the multi-dimensional space has been notably absent.…”

A machine learning aided interpretable model for rupture strength prediction in Fe-based martensitic and austenitic alloys

“…Another article showed that by incorporating thermodynamics data generated from the computational thermodynamics study into the machine learning model, highly accurate models can be obtained for creep prediction in ferrous materials 27 . In a recent article, Jiang et al showed that a machine learning model can be successfully employed to accurately predict the tensile strength in pearlitic steel wires 28 . However, the assessment also revealed a number of weaknesses in the existing ML models infrastructure, e.g., (1) small dataset size, in particular, a large amount of published data has not been used in building these models 30 , (2) model accuracy is currently not sufficient to make reliable predictions because of the use of inadequate algorithms, and (3) accurate interpretation of developed models for the inverse design of novel alloy materials is not straightforward.…”

“…With the advent of the Materials Genome Initiative, highly sophisticated database management system (DBMS), and unprecedented improvement in machine learning algorithms and computational power, machine learning has enabled development of highly accurate and fast predictive models that are accelerating the identification and subsequent deployment of superior materials for a variety of applications 24,25 . In a recent assessment of the literature, a data science based approach has been found to be more accurate than just a physics-based one, or one using thermodynamics-based models, for the prediction of rupture life or rupture strength [26][27][28][29] . Gaussian Process regression with Matérn kernel has been successfully utilized to predict the creep rupture life with 56% overall prediction performance by synergistically exploiting the experimental findings with the state-of-the-art computational machine learning methods 26 .…”

Machine Learning‐Aided Process Design: Modeling and Prediction of Transformation Temperature for Pearlitic Steel

“…The ML approaches commonly exhibit excellent performance in dealing with the complex multivariate nonlinear relationship between input and output variables in view of the existing data. [38] With respect to the complex and nonlinear interaction between alloying elements, nevertheless, ML has become an efficient technical means to get sufficient phase transition temperature, which can be determined on continuous cooling…”