Automated Production Network Planning Under Uncertainty by Developing Representative Demand Scenarios

Due to the variety and interaction of volatile influencing factors as well as the increasing requirements resulting from individualization, the prediction of future demand development is becoming increasingly difficult and complex. In manufacturing companies, this leads to a need for shorter and faster production planning cycles. In addition, the production network must be secured against uncertainty. This is possible by scenario analysis integrated into automated planning. In this paper, an automated scenario analysis in combination with deterministic modeling for integrated product allocation and global network configuration is developed to tackle demand uncertainty in a medium-term planning horizon. When creating scenarios, a trade-off arises concerning the completeness of possible developments and the manageability of the set. The objective is to achieve a representative coverage of possible future states by a small number of reasonable scenarios. Therefore, change drivers are defined that can lead to modifications of customer orders. This is followed by an automated simulation of the occurrence of the change drivers using a Monte Carlo simulation with a high number of samples for statistical validation. A cluster analysis with upstream principal component analysis is used to reduce the number of scenarios while maintaining representativeness. Finally, the scenarios are optimized in a production planning tool. The approach is applied to a real use case. The results are used to validate the representativeness of the scenarios, as well as to conclude robust decisions.