Integrating AutoGluon for Real-Time Monitoring and Classification of Dental Equipment Performance

Abstract

This study aims to introduce AutoGluon, an automated machine learning (AutoML) framework that monitors and classifies the performance of dental equipment in real time. The intent is to enable predictive maintenance through data processing automation, model selection and performance classification. Using sensor data, such as temperature readings, vibration readings, and pressure readings from different dental tools, we trained machine learning models using AutoGluon. Data preprocessing consisted of label encoding and normalization. The system evaluates multiple performance metrics: accuracy, balanced accuracy and Matthews correlation coefficient (MCC). We compared models to identify the best approach for real-time monitoring. The Weighted Ensemble model achieved a 100% accuracy, balanced accuracy and MCC score of 1.0 (indicating perfect reliability in classifying equipment states as 'Optimal,' 'Warning,' and 'Failure'). The system was computationally efficient, generalizable across classes, and showed robust generalization, making it suitable for real-time deployment. Temperature and vibration were the most influential features in predicting equipment states based on the SHAP analysis. By taking advantage of AutoGluon, the proposed system significantly increases the reliability of dental equipment monitoring while also performing real-time machine classification and prediction maintenance. This scalable solution ensures quality patient care, optimizes maintenance scheduling and reduces downtime. The system will be enlarged with additional sensors, used to develop the system further in future work, and deployed to live clinical environments.