Industrial ventilation systems are highly susceptible to vibration-induced failures, which can compromise operational efficiency and safety. This paper proposes a predictive maintenance (PdM) framework leveraging vibration-based condition monitoring, integrated with modern data analytics and IoT technologies. We analyze and compare conventional and advanced signal processing techniques, including Fast Fourier Transform, Wavelet, and Cepstrum analysis, as well as machine learning models for fault classification and Remaining Useful Life (RUL) prediction. A simulated testbed based on benchmark datasets (e.g., PRONOSTIA) was used to evaluate model performance. Among various algorithms tested, Random Forest demonstrated the highest accuracy in fault detection. The study further incorporates edge computing, LoRaWAN, and wireless sensor networks (WSNs) into a scalable PdM system architecture. This approach enables real-time vibration monitoring and decision-making to preempt component-level failures. The results show the feasibility of an integrated, intelligent PdM solution for improving equipment lifespan and reducing unplanned downtime in industrial ventilation systems.
Andreica et al. (Fri,) studied this question.