A Study on Fault Prediction of Simulated Die Press Machine Based on Machine Learning

Reliability and fault prediction are essential for enhancing the operational performance of simulated tablet presses. In this study, multiple machine learning algorithms are applied to fault prediction to improve both prediction accuracy and computational efficiency. The algorithms employed include K-nearest neighbor classifiers, support vector machine classifiers, Naive Bayes classifiers, Gaussian process classifiers, a parameter-tuned Random Forest model, and a SMOTE-BO-XGBoost ensemble model. A comprehensive evaluation is conducted using multi-dimensional performance metrics, including accuracy, precision, recall, F1 score, and first error rate, to compare the effectiveness of the models. The results indicate that the parameter-tuned Random Forest model achieves superior performance in terms of both accuracy and F1 score, demonstrating that parameter optimization plays a significant role in enhancing model performance. In addition, the SMOTE-BO-XGBoost ensemble model exhibits excellent performance in minimizing the first error rate, underscoring its robustness and predictive accuracy. These findings confirm the feasibility and reliability of applying machine learning techniques to fault prediction for simulated tablet presses, and they offer valuable guidance for the selection and optimization of predictive algorithms in future applications.