<![CDATA[Background: Industrial gas leakage remains a critical safety hazard that can trigger fatal accidents, infrastructure damage, environmental pollution, and substantial economic losses. Conventional gas detection systems relying on static threshold mechanisms often generate false alarms and lack adaptability in dynamic industrial environments. Objective: This study aims to design and evaluate an Internet of Things (IoT)-based early gas leakage detection system integrated with Machine Learning (ML) algorithms to enhance detection accuracy and responsiveness. Research Method: A Research and Development (R&D) approach was employed by implementing a Sensor–Microcontroller–Server–Dashboard architecture for real-time gas monitoring at five-second intervals. A total of 2,400 data samples were collected under normal and simulated leakage conditions. The dataset was divided into 80% training and 20% testing. Random Forest was selected as the classification model based on comparative performance evaluation using confusion matrix metrics. Results: The experimental findings demonstrate strong classification performance, achieving 95.2% accuracy, 94.3% precision, 96.3% recall, and 95.3% F1-score. The high recall rate indicates effective minimization of false negatives, which is critical in industrial safety systems. Overall: The integration of IoT and ML provides a more adaptive, accurate, and responsive gas leakage detection solution compared to conventional threshold-based systems, supporting smarter and data-driven industrial safety management.]]>
Copyrights © 2024
