Learning curve modelling of work assignment in mass customized assembly lines

In customized markets, a large variety of product models is demanded by customers. That requires fast setup of production resources to comply with specifications of the next model to be produced. Such compliance, however, may cause considerable production and quality losses owing to workers’ poor performance during the initial production runs of a new model. Therefore, modelling workers’ learning upon exposure to each product model may help production managers to define the best assignment scheme for models and workers, such that losses in the initial stages of production are minimized. The current paper presents a method that uses learning curves to guide the best assignment of product models to teams of workers. Product models are first clustered into families based on their similarities, aiming at a smaller data collection to generate the learning curves. Allocation of product families to teams is then carried out, based on the analysis of their corresponding learning curves. Two courses of action are proposed for this, depending on whether the production batch will lead to longer or shorter production runs. The proposed methodology is illustrated in a case study from the shoe manufacturing industry.