Ergonomic Assessment for Maintenance Personnel in Open-Pit Using Artificial Intelligence

Abstract

Musculoskeletal disorders (MSDs) represent a critical occupational health challenge for maintenance personnel in open-pit mining, where awkward postures, heavy load handling, and repetitive tasks substantially increase the risk of injury. This study develops and validates an innovative automated ergonomic assessment system, based on artificial intelligence (AI), for the real-time identification of postural risk. The methodology integrates on-site image capture, processing with the MediaPipe library for joint angle extraction, and classification using a convolutional neural network (CNN). The model, trained on a dataset of 2,450 annotated images of critical tasks (tire changing, nut extraction, etc.), demonstrated robust performance, achieving an accuracy of 89.4% when validated against expert ergonomic assessments using the REBA method. The results show that the system overcomes the limitations of traditional observational methods by providing an objective, quantitative, and scalable evaluation. It is concluded that the integration of this AI system enables not only proactive monitoring and the prioritization of specific ergonomic interventions but also establishes the foundation for improving health outcomes and productivity in this high-risk industrial environment.