Real-time ergonomic risk assessment of musculoskeletal disorders using machine learning and digital modeling: a case study of repetitive manual tasks

Purpose This study aims to develop and validate a real-time, computational framework for evaluating ergonomic risks associated with repetitive manual tasks, with a focus on musculoskeletal disorders (MSDs). The research seeks to bridge the gap between traditional ergonomic assessments and modern, data-driven approaches by integrating machine learning and digital human modeling techniques. Design/methodology/approach This study introduces a pilot computational framework that combines 2D pose estimation, geometric joint angle modeling, Machine learning (clustering) and automated ergonomic scoring. Using OpenPose, worker postures were extracted from video footage without wearable sensors. Ovako Working Posture Analyzing System (OWAS) was implemented as the primary ergonomic measure, and Rapid Upper Limb Assessment (RULA) and Rapid Entire Body Assessment (REBA) were automated for comparative benchmarking. The pipeline was applied to a case study of packaging bag preparation, a multistage manual task involving repetitive retrieval, cutting and arranging. Findings Results showed sustained trunk flexion, frequent upper-limb elevation, and repetitive forearm cycles, yielding average scores of OWAS Category 3–4 (high risk), RULA 5.27 (moderate risk and action required) and REBA 5.13 (moderate risk and intervention required). These findings align with epidemiological evidence linking packaging tasks to high MSD prevalence. Practical implications This pilot study demonstrates the potential of AI, computer vision and ML to enable scalable, continuous ergonomic monitoring. The framework offers a practical pathway toward data-driven intervention design and proactive prevention of work-related MSDs, pending broader validation. Originality/value This study introduces a novel ergonomic evaluation pipeline that offers a reproducible, real-time alternative to conventional observational methods. By using data-driven techniques to assess ergonomic risks, this study provides a foundation for developing strategies to improve worker safety, prevent musculoskeletal disorders and enhance the ergonomics of industrial tasks. In contrast to traditional approaches, the framework enables detailed, quantitative risk assessment in dynamic work environments, making it particularly suitable for industrial applications where wearable sensors are impractical or intrusive.