**Authors:** Muxiu Yang, Fengzhou Li, Wenfeng Qiu
**Published:** _IEEE Access_, 2025
**Link:** [IEEE Xplore](https://doi.org/10.1109/ACCESS.2024.3523519)
---
### Summary
This study presents a real-time monitoring system for **dental equipment** using [[AutoGluon]], an automated machine learning (AutoML) framework. The system classifies equipment performance as **Optimal, Warning, or Failure** based on real-time sensor data like **temperature, vibration, and pressure**. It aims to prevent unexpected equipment failure and optimize maintenance schedules through predictive analytics.
---
### Key Points
- **[[AutoGluon]] for Real-Time Monitoring:**
- Automatically selects and trains [[ML]] models for performance classification.
- Models are fine-tuned with **hyperparameter optimization** and **feature selection**.
- Achieved **100% accuracy** in classifying equipment states using the **Weighted Ensemble model**.
- **Sensor Data:**
- Uses temperature, vibration, and pressure sensors to detect anomalies.
- **SHAP Analysis** showed **temperature** and **vibration** as the most influential features.
- **Model Performance:**
- Weighted Ensemble achieved **100% accuracy**, **balanced accuracy**, and Matthews Correlation Coefficient #MCC of **1.0**.
- Other models like [[LightGBM]], [[RandomForest]], and [[CatBoost]] also performed exceptionally well.
- **Scalability and Deployment:**
- Designed to be scalable and deployable in **live clinical environments**.
- Future improvements include **edge computing** for faster real-time analysis and expanding the system to more types of dental equipment.
---
### How It Works
1. **Data Collection:**
- Real-time sensor readings (temperature, vibration, pressure) are streamed from dental equipment.
2. **Preprocessing and Training:**
- Data is preprocessed with **label encoding** and **normalization**.
- [[AutoGluon]] handles **model selection**, **training**, and **hyperparameter tuning**.
3. **Model Evaluation:**
- Models are tested for #Accuracy, #Precision, #Recall, and #F1-score
- The best-performing models are ensembled for optimal predictions.
4. **Deployment:**
- The system is deployed for **real-time monitoring** and **predictive maintenance** in dental clinics.
---
### Strengths
- **High Accuracy:** Achieves 100% classification accuracy in real-time.
- **Low Maintenance:** [[AutoGluon]] handles model selection and optimization automatically.
- **Scalable:** Can be expanded to include more sensor types and equipment.
---
### Limitations
- **Data Dependency:** Requires high-quality sensor data for optimal performance.
- **Limited Equipment Types:** Currently focused on X-rays, dental chairs, and sterilizers.
- **Sensor Integration:** Needs seamless integration with proprietary dental equipment.