Predictive Maintenance Based on Embedded AI and Wireless Monitoring of Working Environment

Traditional fault diagnosis methods are often inefficient and costly, hindered by limited computing power that compromises accuracy. Their lack of flexibility also poses challenges in establishing effective multi-device operational monitoring systems. To address these issues, we developed a fault diagnosis and predictive maintenance system based on Lightweight-Time-Delay Neural Networks (Lightweight-TDNNs). Leveraging their robust capability in processing time-series data, TDNNs excel in learning features across various time delays. By optimizing AI algorithms for lightweight deployment, we successfully implemented embedded AI for predictive maintenance of machinery and equipment faults at the edge. Utilizing a Zigbee-based IoT approach, we enabled wireless monitoring of equipment status and environmental conditions, overcoming the constraints of wired installations. Validation on the STM32F429 chip demonstrated that our model achieves over $\mathbf{9 0 \%}$ accuracy on test sets, demonstrating its suitability for industrial applications.