Intelligent Real-Time Thermal Monitoring System for Industrial Environments Using Machine Learning and Thermal Imaging

Fault detection has become an indispensable strategy in the current developments of technical features as newer machinery grows increasingly complex, aiming to prevent costly downtimes and safety risks. Traditional periodic inspection-based maintenance methods usually result in unexpected breakdowns and resource utilization in an inefficient manner. Former studies on predictive maintenance, which were based on sensors such as vibration or current monitors, have been effective in this regard but lack early anomaly detection and are prone to latency issues in fast-paced industrial settings. The above-mentioned limitations are targeted by proposing a real-time fault detection system, combining thermal imaging and machine learning. Thermal cameras stream video from the real-time mapping of temperature distributions and delays are further reduced using WebRTC. Temperature variations will be analyzed with the help of a Convolutional Neural Network (CNN), machinery detection will be done with You Only Look Once (YOLOv5), and then the Isolation Forest algorithm will spot the anomalies as soon as possible. The results showed very high accuracy of as high as 97.5%, with only 30 ms latency. This solution is easily scalable and very promising for industrial applications shortly. Research into optimizing edge device deployment is ongoing.