Model-based increase of flexibility and robustness of an aerodynamic part feeding system for high-performance assembly

为了克服传统送料系统在送料性能、可靠性和变型灵活性方面存在的缺陷，IFA开发了一种气动零件送料系统。该系统的特别特点是其进料速度高达每分钟1,000个工件和高技术可用性。针对不同工件的零件进给系统的建立仅限于四个系统参数的自适应。然而，最优参数值的识别是非常耗时的。因此，在以前的研究活动中，已经开发了一种遗传算法。该算法使零件进给系统能够自主识别最优参数值，并通过相应的硬件进行自动调整。因此，调整零件进给系统的时间可以大大减少。在接下来的研究方案中，为了应对日益个性化的产品挑战，需要通过采用遗传算法和零件进给系统来扩大被喂工件的范围。如今，不同工件几何形状的仿真需要手动调整仿真模型。这就需要相应的仿真专家的存在。因此，目标是扩展仿真模型，使其能够在不需要手动调整仿真的情况下模拟不同的工件。此外，还应对仿真模型进行扩展和优化，使其结果能够减少在实际中识别最优参数值的时间。为了能够在不降低进给速度的情况下进给更广泛的工件，应将遗传算法的最佳设置（例如突变率或种群大小）确定为要进给的工件特征的函数，并应开发环境条件变化时经济地触发重新参数化。因此，待进给工件的频谱将由相关工件特性及其相关限制的目录来定义。

To counter existing deficits of conventional feeding systems regarding feeding performance, reliability and variant flexibility, an aerodynamic part feeding system has been developed at the IFA. This system is particularly characterized by its feeding rate of up to 1,000 workpieces per minute and its high technical availability. The setting-up of the part feeding system to different workpieces is limited to the adaption of only four system parameters. However, the identification of optimal parameter values was very time-consuming. In previous research activities, therefore, a genetic algorithm has been developed. This algorithm enables the part feeding system to autonomously identify optimal parameter values and to adjust them automatically via corresponding hardware. As a consequence, the time to adjust the part feeding system could be reduced significantly. In this subsequent research proposal, the spectrum of workpieces to be feeded should be expanded by adapting the genetic algorithm and the part feeding system in order to meet the challenges of increasing product individualization. Nowadays, the simulation of different workpiece geometries necessitates manual adjustments in the simulation model. This requires the existence of corresponding simulation experts. Therefore, an aim is to extend the simulation model in a way that enables the simulation of different workpieces without the need to adjust the simulation manually. Furthermore, the simulation model should be extended and optimized in a way that its results lead to a reduction of time to identify optimal parameter values in practice. For being able to feed a wider spectrum of workpieces without decreasing the feeding rate, optimal settings of the genetic algorithm (e. g. mutation rate or population size) should be identified as a function of the characteristics of the workpieces to be feeded and an economic triggering of a reparameterization when ambient conditions change should be developed. Thereby, the spectrum of workpieces to be feeded is to be defined by a catalog of relevant workpiece characteristics and their associated limits.