This paper presents an embedded machine learning approach for real-time fault detection in industrial fans. The system monitors fan vibration using an accelerometer and employs a Convolutional Neural Network (CNN) model to assess anomalies. The CNN model demonstrates high accuracy without the need for feature extraction, achieving optimal results within 30 epochs. The trained model was deployed on an embedded system, achieving accuracy rates of 99.8%, 99.9%, and 100.0% for Fan-Fault, Fan-Off, and Fan-On states, respectively, on a validation dataset. Real-time testing confirmed high accuracy scores ranging from 90% to 100% across all operational states. The research addresses challenges such as algorithm selection, real-time deployment, hyperparameter tuning, sensor integration, and energy efficiency. The proposed methodology is promising for efficient and accurate fan fault detection, with potential applications in broader industrial and smart sensing contexts.This paper presents an embedded machine learning approach for real-time fault detection in industrial fans. The system monitors fan vibration using an accelerometer and employs a Convolutional Neural Network (CNN) model to assess anomalies. The CNN model demonstrates high accuracy without the need for feature extraction, achieving optimal results within 30 epochs. The trained model was deployed on an embedded system, achieving accuracy rates of 99.8%, 99.9%, and 100.0% for Fan-Fault, Fan-Off, and Fan-On states, respectively, on a validation dataset. Real-time testing confirmed high accuracy scores ranging from 90% to 100% across all operational states. The research addresses challenges such as algorithm selection, real-time deployment, hyperparameter tuning, sensor integration, and energy efficiency. The proposed methodology is promising for efficient and accurate fan fault detection, with potential applications in broader industrial and smart sensing contexts.