Published Paper Details:

Machine vibration-based health monitoring has emerged as a vital technique for predictive maintenance and fault diagnosis in rotating and industrial machinery. This review conducts a detailed comparative analysis of recent research papers focusing on vibration signal processing and machine learning-based diagnostic models. Studies published between 2017 and 2025 demonstrate a significant evolution from traditional signal and frequency domain methods toward data-driven frameworks leveraging deep learning (DL), convolutional neural networks (CNN), long short-term memory (LSTM), and transfer learning approaches. Benchmark datasets such as the Case Western Reserve University (CWRU) and Paderborn datasets remain the primary sources for model evaluation. Despite the impressive accuracy achieved by deep models, several challenges persist, including handling noisy sensor data, ensuring model generalization in varying operational conditions, and achieving real-time deployment on low-power devices. Moreover, there is limited exploration of multimodal sensor fusion and domain adaptation techniques. The study concludes that future research should focus on integrating IoT-enabled vibration monitoring systems with edge AI, adaptive learning, and cloud-based predictive analytics to advance intelligent fault diagnosis and ensure industrial reliability in the era of Industry 4.0.