Machine Learning Assisted Density Prediction of Cathodes Alloy Matrix

The porous structure of the dispenser cathodes matrix provides a fast channel for the supplement of active substances, which greatly improves the emission performance of the cathode. However, controlling the porosity of the matrix is still challenging. In this work, the dataset consists of the density of two hundred binary tungsten alloy matrices. The Pearson correlation coefficient and recursive elimination method were used as feature selection methods to select potential influencing factors of matrix density. Comparing different machine learning models, it was found that MLP neural networks have accurate predictions of density. By training the learning model, the prediction accuracy (R ² ) is as high as 0.918. In addition, the predicted density of the porous matrix is in good agreement with the experimental values.