Real-Time Detection of Industrial Respirator Fit Using Embedded Breath Sensors and Machine Learning Algorithms.

Area of Science:

• Occupational Health and Safety
• Biomedical Engineering
• Sensor Technology

Background:

• Effective facial seal is crucial for tight-fitting industrial respirators.
• Traditional fit testing methods (QLFT, QNFT) are periodic and cannot detect in-use fit degradation.
• High-risk sectors like mining, manufacturing, and construction require reliable respirator performance.

Purpose of the Study:

• To develop and validate a real-time respirator fit detection system.
• To enable continuous, non-invasive monitoring of respirator-face seal integrity.
• To enhance occupational safety and regulatory compliance through immediate fit feedback.

Main Methods:

• A compact sensor module inside the respirator measures pressure, temperature, and humidity.
• Data is transmitted via Bluetooth Low Energy (BLE) to a smartphone for on-device machine learning inference.
• Multimodal biosensing (pressure, thermal-hygrometric signals) and cycle-synchronous pattern analysis were employed.

Main Results:

• Machine learning models (Random Forest, SVM, XGBoost) achieved F1 scores over 95% in detecting fit/misfit conditions.
• The system demonstrated reliable real-time readout of respirator-face sealing.
• Validation included k-fold cross-validation and leave-one-subject-out evaluation on over 10,000 breathing cycles.

Conclusions:

• The developed system offers a scalable, cost-effective, and field-deployable solution for continuous respirator fit monitoring.
• It relies on internal physiological signals, unlike traditional external measurement techniques.
• Enables real-time alerts during work shifts, significantly improving occupational safety.