AI-Driven Decision Support System Combining CNN Wrist Posture Classification and Real-Time RULA Biomechanical Analysis for Ergonomic Risk Prevention in Automotive Manufacturing

Repetitive movements and awkward upper-limb postures remain common in modern automotive manufacturing, where operators often repeat the same motion pattern for extended periods. Over time, these repetitive constraints may contribute to musculoskeletal disorders (WMSDs), particularly at the wrist. Traditional ergonomic assessments such as RULA are valuable tools, but they are typically applied manually and intermittently, which makes it difficult to detect posture-related risks as they develop in real working conditions. In this work, we introduce a hybrid ergonomic monitoring system that combines two complementary components: a Convolutional Neural Network (CNN) trained to classify wrist postures (acceptable vs. non-acceptable), and an automated real-time RULA evaluation engine based on pose landmarks. Using a dataset of 3,000 wrist images captured in authentic factory settings, the CNN model was trained and achieved a final accuracy of 94%. The combination of CNN-based Classification and RULA scoring provides continuous ergonomic insights and illustrates how AI can enhance preventive interventions in Industry 4.0 environments.