Development of optimized ensemble machine learning-based prediction models for wire electrical discharge machining...

Area of Science:

• Manufacturing Engineering
• Computational Intelligence
• Materials Science

Background:

• Wire electrical discharge machining (WEDM) is crucial for manufacturing complex profiles on hard-to-machine materials.
• Accurate prediction of WEDM process responses is vital for optimizing production and material quality.

Purpose of the Study:

• To develop optimized heterogeneous ensemble models for predicting WEDM process responses.
• To enhance prediction accuracy by combining multiple machine learning algorithms.

Main Methods:

• Developed ensemble models by integrating predictions from Random Forest, Support Vector Machine, and Ridge Regression.
• Formulated optimization problems to minimize prediction errors (RMSE, MAE) for weighted ensemble creation.
• Evaluated model performance using nine statistical metrics and a Multi-Response Signal-to-Noise (MRSN) ratio.

Main Results:

• Optimized ensemble models demonstrated higher prediction accuracy compared to individual base models.
• The Multi-Response Signal-to-Noise (MRSN) ratio confirmed the superior performance of the developed ensembles.
• The study utilized two experimental datasets from WEDM processes.

Conclusions:

• Optimized heterogeneous ensemble models offer superior prediction accuracy for WEDM processes.
• Ensemble modeling provides a robust approach for improving the reliability of manufacturing process predictions.
• The proposed method is effective for optimizing complex machining operations.