ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE, vol.123, no.Part B, pp.1-17, 2023 (SCI-Expanded)
Non-ergonomic working conditions are the leading causes of musculoskeletal disorders that seriously affect
human health. REBA is widely used tool due to its convenience and consideration of all body parts. However,
it heavily relies on the subjective judgments of the assessor, leading to inconsistencies in results, and lacks
sensitivity in detecting small changes in ergonomic risk factors. Therefore, there is a need to improve
the REBA method by integrating it with new technologies. While a few studies have proposed integrating
ergonomic risk measurement tools with ANNs, there is a research gap in comparing different types of neural
networks and membership functions to determine the most effective approach for improving the performance
of REBA. Additionally, there is a need to apply these integrations to real-life case studies to demonstrate
their effectiveness in practice. This study proposes a comparative neural network and neuro-fuzzy-based REBA
method that includes various types of neural networks and membership functions. The proposed method is
applied to service employee who have experienced increased workloads due to the Covid-19 pandemic. The
results show that the neuro-fuzzy method is more accurate than the REBA and provides greater flexibility in
defining which member belongs to which risk level cluster. This study is critical because it addresses research
gaps in integrating neural networks and REBA and applies these integrations to a real-life case study. By
comparing different types of neural networks and membership functions, the study provides insights into which
approaches are most effective for improving the performance of REBA.