Online sequencing of buffers for automotive assembly lines

Online sequencing of buffers for automotive assembly lines

 In this project, we considered the usage of a buffer to alternate the sequence of orders at an assembly line in the automotive industry. These assembly lines must be flexible enough to produce multiple vehicle models, so the automobile manufacturers face the challenge to produce a huge number of models while using the relatively inflexible production form of an assembly line.

Due to unforeseen disturbances at the assembly line, like machine breakdowns, material defects or last-minute orders, the initial sequence of products is disarranged and must be put back to avoid idle times and overload. To resequence the orders, we considered a buffer which is placed upstream of the final assembly.

In each cycle time, a new (random) product enters the buffer and one of the buffer positions must be selected for the final assembly. The decision (selection) must be made without full information in an online problem. Due to the complexity of the problem, heuristic rules and metaheuristics were used to solve this problem:

1.       Heuristic rules

2.       Local search algorithm

3.       Lookahead search

To pick a job from the buffer, the rules can process the information of the current state, containing processing times, due dates, worker positions, utility work and idle time. To test the different solution approaches, a stochastic simulation, based on three parts: model generation, selection of model and utility work simulation, was set up.

For the simulation 45 instances based on the dataset of Scholl (from assembly-line-balancing.de) with different numbers of station and cycle times were used. The dataset and the obtained results can be found [here].

The contact person for this research project is Celso Gustavo Stall Sikora.