Selection of mechanical properties of uranium and uranium alloys after corrosion based on machine learning

Uranium and uranium alloys play a vital role as service materials in strategic equipment, but their mechanical properties can be adversely affected by corrosion. Therefore, accurately predicting the mechanical properties of uranium and uranium alloys after corrosion holds significant importance. In this study, we created a database to investigate the impact of oxygen corrosion on the mechanical properties of uranium and uranium alloys. Uti-lizing this database, we developed a feature-guided decision tree algorithm to predict various tensile properties, including yield strength, tensile strength, elongation, and cross-section shrinkage. Our research highlights three key findings. Firstly, we established a machine learning modeling framework that effectively predicts tensile properties and exhibits potential for predicting other properties of uranium and uranium alloys. Secondly, through feature engineering, we uncovered crucial correlations involving reaction time, reaction temperature, alloy type, phase structure composition, and phase number. These correlations significantly enhanced the per-formance of machine learning models in predicting tensile properties after oxygen corrosion. Lastly, by employing the decision tree algorithm guided by feature engineering, we successfully predicted the mechanical properties of uranium and uranium alloys after oxygen corrosion with a prediction error of less than 5%.

Automated summary

This study investigates the impact of oxygen corrosion on the mechanical properties of uranium and uranium alloys and develops a feature-guided decision tree algorithm for prediction. The research highlights crucial correlations discovered through feature engineering, improving the performance of machine learning models. The mechanical properties of uranium and uranium alloys after oxygen corrosion are successfully predicted with a prediction error of less than 5%.