Multiple-source learning precedence graph concept for the automotive industry

In modern production systems, customized mass production of complex products, such as automotive or white goods, is often realized at assembly lines with a high degree of manual labor. For firms that apply assembly systems the assembly line balancing problem (ALBP) arises, which is to assign optimally tasks to stations or workers with respect to some constraints and objectives. Although the literature provides a number of relevant models and efficient solution methods for ALBP, firms, in most cases, do not use this knowledge to balance their lines. Instead, the planning is mostly performed manually by numerous planners responsible for small sub-problems. This is because of the lack of data, like the precedence relations between the tasks to be performed. Such data is hard to collect and to maintain updated. Klindworth et al. (2012) proposed an approach to collect and to maintain the data on precedence relations between tasks at a low cost, as well as to produce new high-quality feasible assembly balances based on this data. They utilize the knowledge on former production plans available in the firm. However, due to reliance on the single source of information, their concept needs long warming-up periods. Therefore, we enhance the concept by incorporating multiple sources of information available at firms, as the modular structure, and present a guideline on how to conduct valuable interviews. As a result, our approach generates more efficient new feasible assembly line balances without requiring such a long warming-up period.