Modelling the 2D object recognition task in manufacturing context: An information-based model

In the last decays, manufacturing systems evolved to meet the high product variety required by the market. Different products can be manufactured in the mixed-model assembly lines, with an increase in the process complexity. In these production systems, the required flexibility is mainly provided by operators in the final assembly stages. Here, human errors could lead to high economic losses. A lack is observed in available research concerning a formal quantification of manufacturing complexity considering the joint effect of shape complexity and similarity in the mix variety. This paper focuses on operator decision-making in 2D object recognition tasks, since this is the most critical task performed in mixed model assembly systems. A novel model to quantify the information content in 2D object recognition task is proposed. The model is based on the Shannon's Entropy theory and considers both shape complexity and object similarities. Numerical experiments are provided, and results obtained show the effectiveness of the model in capturing the joint effect of shape complexity and similarities on the task information content. The proposed model can be adopted in a production environment for re-allocating tasks/sub-tasks to avoid the high amount of information to be processed affecting operators' performance.