In construction, workers are frequently exposed to ergonomic risks that can lead to musculoskeletal disorders. To prevent ergonomic injuries, proper assessment of ergonomic risk is a key to identifying risk factors and modifying work practice in a timely manner. In field observation, however, difficulties in visually estimating human postures (e.g., body joint angles) required for ergonomic analysis have led to inconsistent results due to the subjectiveness of observers. This study thus proposes a fuzzy logic approach to posture-based ergonomic evaluation tools. Rapid Upper Limb Assessment (RULA) is selected as a case study to describe the fuzzy logic modelling of RULA scoring systems and discuss the application to modular construction shops. The results of validation comparing correlations with biomechanical analysis — used as a ground truth — reveal that the proposed system produces more accurate results than traditional methods and hence helps minimize human errors in observation for reliable on-site ergonomic assessment.