Predicting the Tool Wear of a Drilling Process Using Novel Machine Learning XGBoost-SDA

Area of Science:

• Manufacturing Engineering
• Machine Learning Applications
• Predictive Maintenance

Background:

• Tool wear significantly degrades workpiece quality in drilling operations.
• Accurate tool wear prediction is essential for maintaining optimal machine performance and preventing defects.
• Existing methods require improvement for precise wear forecasting.

Purpose of the Study:

• To introduce a novel hybrid machine learning approach for predicting tool wear during drilling.
• To optimize the Extreme Gradient Boosting (XGBoost) algorithm using a Spiral Dynamic Optimization (SDA) algorithm (XGBoost-SDA).
• To evaluate the predictive accuracy of XGBoost-SDA on copper and cast-iron datasets.

Main Methods:

• Developed a hybrid model combining XGBoost with SDA for hyperparameter optimization.
• Conducted simulations using copper and cast-iron drilling datasets.
• Performed comparative analysis against Support Vector Machines (SVM) and Multilayer Perceptron Artificial Neural Networks (MLP-ANN).

Main Results:

• XGBoost-SDA achieved high accuracy in predicting flank wear for copper workpieces (MAE = 4.67%, R² = 0.9973).
• For cast iron, XGBoost-SDA accurately predicted surface roughness (MAE = 5.25%, RMSE = 6.49%, R² = 0.975).
• The proposed XGBoost-SDA method demonstrated superior performance compared to SVM and MLP-ANN.

Conclusions:

• XGBoost-SDA is an effective hybrid machine learning method for accurate tool wear prediction in drilling.
• The model ensures high predictive accuracy for flank wear and surface roughness, crucial for manufacturing quality.
• This approach offers a significant advancement in predictive maintenance for drilling processes.