Publication

Abstract : Tool downtime prediction requires efficient monitoring because it helps maximize industrial operation quality while maintaining uninterrupted plant operations. The analysis utilizes machine learning algorithms with Decision Trees to identify machine downtime events through operational measurements which include hydraulic pressure and coolant temperature and spindle speed and vibration levels. Data normalization alongside Forward Selection feature selection methods were used on a balanced dataset to boost model performance for prediction purposes. The Decision Tree model functioned to differentiate machine defects into two discrete categories-failure and no failure-thus proving its competence in such predictive classification activities. The study conducted a comparative evaluation between Decision Trees and Random Forest along with other algorithms which included SVM and Logistic Regression and Adaboost in addition to CatBoost and XGBoost and Naive Bayes and KNN using KD-Tree and Ball-Tree variations. The experimental findings show that Decision Trees provide clear predictions coupled with high precision and fair runtimes to support predictive maintenance implementations. The research delivers fundamental findings that help improve downtime prediction methods to boost operational effectiveness.