Converging material flow simulation models as digital twins of production systems

在日益动荡的市场环境中，制作公司必须不断适应新的环境条件，对外部事件做出反应，并在巨大的时间压力和不确定性下评估替代行动方案。生产系统的物流仿真模型可以有效而精确地支持这一点，该模型可以处理系统中复杂的相互作用、动态行为和随机过程。由于保持这种模拟模型以所需的高精度提供所需的模型维护是非常劳动密集型的，尤其是技术密集型的，因此在本项目中这将是自动化的。为此目的，制定了使用实际生产数据(由于数字化而越来越多地可用)使模型自动适应现实(所谓的收敛)的程序。因此，传统的模拟模型成为生产系统的数字孪生兄弟。为此，首先必须开发一个整体概念，其中除了整个过程之外，还必须定义数字双胞胎开发的初始状态，即手动创建的种子模型。在此基础上，必须描述数据收集和处理的必要方法。因此，需要开发用于定义许多设置参数的系统学，例如所使用的验证度量、阈值和更新周期长度。由于在实践中，模拟模型通常是在生产系统的规划阶段创建的，而数字孪生兄弟应该伴随生产系统的整个生命周期，因此该研究项目也在研究数字孪生兄弟从计划阶段到运行阶段的过渡。该程序及其算法将被实施，并针对来自行业和研究所自己的学习工厂的示范用例进行测试。通过这种方式，可以更详细地检查数字孪生兄弟在实施后的行为，以便进一步了解其开发和使用。

In an increasingly volatile market environment, production companies must constantly adapt to new environmental conditions, react to external events and evaluate alternative courses of action under great time pressure and uncertainty. This can be efficiently and precisely supported by a material flow simulation model of the production system, which can handle the complex interactions, dynamic behavior and stochastic processes in the system. Since the model maintenance required to keep such a simulation model available with the required high accuracy is very labor-intensive and, in particular, know-how-intensive, this shall be automated in this project. To this end, procedures for using real production data (that is increasingly available because of digitalization) for the automated adaptation of the model to reality (the so-called convergence) are developed. The conventional simulation model thus becomes a digital twin of the production system. For this purpose, first a holistic concept in which, in addition to the overall procedure, the initial state of the digital twin development, the manually created seed model, is defined, must be developed. Based on this, the necessary methods for data collection and processing must be described. Therefore, systematics for the definition of numerous setup parameters, such as the validation metrics used, threshold values and the update period length, are to be developed. Since in practice simulation models are often created in the planning phase of production systems and digital twins should accompany the entire life cycle of the production system, the transition from the planning to the operating phase of the digital twin is also being investigated in the research project. The procedure and its algorithms will be implemented and tested for exemplary use cases from industry and the institute's own learning factory. In this way, the behavior of a digital twin after implementation can be examined in greater detail in order to gain further insights into its development and use.