A Top-Down Approach for an Automatic Precedence Graph Construction under the Influence of High Product Variety

This paper describes a top-down method for an automatic precedence graph construction that can cope with high variant products. The concept generates a joint precedence graph including all variants of a product directly. The graph is automatically derived from the bill of materials and buildability rules as well as existing solutions for the assignment of tasks to workstations. The presented method is very error prone and can improve the practical applicability of many assembly line balancing problems, that could not be used in practice yet. Producing companies have to cope with an increasing product variety and frequently changing demands. Despite this development assembly lines are used to gain an efficient production process. The main focus in setting up and reconfigurating an assembly line is the assignment of tasks to workplaces. Sophisticated algorithms have been developed for this problem [1, 2]. Almost all existing concepts assume that a precedence graph, describing the relations which tasks have to be done before others, exists. A manual definition of a precedence graph for products with high variety fails, due to the complexity. Some authors developed intelligent methods for an automatic assembly sequence generation [3, 5–11]. Most methods proposed have in common that they start bottom up. Initially no relation exists and all tasks are independent. Restrictions between tasks are added successively. This proceeding is very error-prone as forgetting a relation can lead to an unfeasible setup solution, when the graph is used as an input in an assembly line balancing algorithm. Furthermore, none of the former approaches can handle a product portfolio without an explicit variant definition like it is predominant e.g. in the automobile industry. We propose a method that starts top-down with the existing feasible assignment and sequencing solution and uses automatic algorithms to break up restrictions successively. The algorithms use data that already exists in many producing companies. The first advantage of the top-down approach is the guarantee that in any stage of the precedence graph development, valid assignment solutions can be derived from the graph. There are no possibilities to change the order of tasks for assignment algorithms in the graph that could lead to unfeasible 2 Altemeier, Brodkorb, Dangelmaier assignment solutions. Second, the approach can deal with a practical number of variants as the joint-precedence graph, consisting of all models produced in an assembly line, is created directly. Third, the method uses the existing product documentation.