Industrial vibration anomaly detection plays an important role in evaluating the condition and operational performance of rotating industrial equipment. Traditional maintenance practices may not identify developing mechanical faults at an early stage, resulting in increased downtime and operational risks. This work introduces a complete IoT and AI-based framework intended for continuous predictive maintenance applications. The system utilizes high-precision MEMS accelerometers integrated with an IoT gateway to capture tri-axial vibration data. To process this data, we developed a hybrid deep learning model—combining Convolutional Neural Networks (CNN) for spatial feature extraction and Long Short-Term Memory (LSTM) networks to capture temporal dependencies in vibration patterns. This architecture allows for the autonomous identification of deviations from "normal" operating signatures, such as misalignment, bearing failure, or imbalance. Experimental results demonstrate that the proposed AI-driven approach significantly outperforms manual threshold-based monitoring. The system achieved an anomaly detection accuracy of 98.5%, with the IoT integration ensuring a latency of less than 100ms for edge-to-cloud data transmission. By replacing periodic manual inspections with this continuous, automated methodology, industrial plants can achieve higher reliability and a measurable reduction in maintenance overhead, proving the system's viability for Industry 4.0 applications.