Feature-Driven Decision Tree Approach for Microhardness Enhancement in nanopowder Mixed Electrical Discharge Machining

Abstract

In the proposed approach, the use of the decision tree-based algorithm establishes a strong prediction of micro hardness for the materials produced under specific machining parameters. With the dataset consisting of variables such as gap current, pulse ON time, pulse OFF time, voltage, and powder concentration, the model attained perfect fitting during training with a root mean-square error (RMSE) of 0.00 and a coefficient of determination (R²) of 1.0000. Recursive Feature Elimination (RFE) in conjunction with cross-validation was used to select features optimally and to avoid overfitting. Accordingly, the reduced model depicted the major key process parameters with gap current, pulse time, and powder concentration being the main influencing process parameters to microhardness values. A clarity of interpretation was attained by this decision-tree structure capturing nonlinear dependencies and resulted in recommending an optimal microhardness of 1382 HV  for input parameters, gap current of 6 A, pulse ON time of 200 µs, and powder concentration of 0.6%. Such results mark decision-tree models as a means by which intelligent design and quality control can be performed in high-tech machining.