Design Method for Forming Tools for Rotary Draw Bending of Bend-in-Bend Geometries

回转拉弯广泛用于生产弯管或小半径弯管。生成的几何图形由工具的接触面生成。这些成形工具的接触面目前是使用有限元(FE)模拟以及专家的知识来设计的。由于模拟模型没有考虑某些物理影响，实际成形过程的结果通常与计算的几何形状不同。特别是它们在成形过程中的相互作用和瞬时行为被忽略了。这对于折弯中的几何图形更为重要，在折弯几何图形中，两个折弯的序列排列在一起，而没有直的中间部分。对于第二个折弯，第一个折弯的轮廓定义了第二个工具的型腔的几何图形。即使第一个弓的微小不精确也会导致产品的较大几何偏差。本研究项目的目标是通过不同的工具设计来达到更高的工艺稳定性。该方法是在大量的产品周期中监测工艺的传感器数据，并利用该数据库设计成形过程模型。其他传感器将被集成到成形工具和折弯机中，以获得过程的虚拟图像。第一个资助期(第一个FÖP)的目标是数字表示(研究重点A)和弯曲过程的数据驱动、可解释模型(研究重点C)。目标值是相对于折弯的形状和位置偏差的零部件几何形状，以及潜在起皱区域中的零部件轮廓。成形技术主席(UTS)负责弯曲过程的数值模拟、确定导致部件质量特性波动的敏感工艺变量、在成形系统中选择和实施合适的传感器以及开发自动化、时间同步的测量外围设备。此外，正在开发一种关于成形系统参数和真实产品几何形状的刀具几何参数的方法。普适计算主席(IUC)正在以数据管道的形式开发该过程的数字化，该管道能够将过程数据、专家知识和模拟模型结合在一起。这尤其需要调整关系数据库的数据驱动和可视化模型、具体的特征工程以及将基于随机森林方法的分类和回归过程。这项工作的结果是收集了一组可理解的参考数据集，其中基于数据表示了过程序列，并结合领域知识和学习的分析结果首次对可解释的模型进行了改编。

Rotary draw bending (RDB) is widely used for production of bent tubes or pipes with small radii. The resulting geometries are generated by the contact surfaces of tools. These contact surfaces of the forming tools are currently designed using finite element (FE) simulations as well as knowledge of specialists. The results of the real forming processes are usually different from calculated geometries because some physical effects are unaccounted for the simulation model. Especially their interactions and transient behavior during forming are neglected. This is more important for bend-in-bend geometries in which a sequence of two bends lined up without a straight intermediate part. For the second bend, the contour of the first defines the geometry of the cavity of the second tool. Even small imprecisions of the first bow lead to bigger geometrical deviations of the product. The goal of this research project is to reach a higher process stability by a different tool design. The method is to monitor sensor data of the process during a large quantity of product cycles and design a model of the forming process using this database. Additional sensors will be integrated into the forming tools and the bending machine to get a virtual image of the process. The goal of the first funding period (1st FÖP) is the digital representation (research focus A) and the data-driven, explainable model of the bending process (research focus C). Target values are the components geometries with regard to shape and position deviations of the bends as well as the component contour in the potential area of wrinkling. The Chair of Forming Technology (UTS) is responsible for the numerical modeling of the bending process, the determination of sensitive process variables that lead to fluctuating quality characteristics of the component, the selection and implementation of suitable sensors in the forming system and the development of automated, time-synchronized measuring peripherals. Furthermore, a method for parameterizing the tools geometries with respect of forming system parameters and real product geometries, is being developed. This forms the basis for the design of improved contact surface designs, using the data-driven process modeling in the 2nd FÖP. The Chair of Ubiquitous Computing (IUC) is developing the digitization of the process in the form of a data pipeline that, with regards to the 2nd FÖP, is able to combine the process data, expert knowledge and simulation models. This particularly requires an adaptation of data-driven and visual models for the RDB, specific feature engineering as well as classification and regression processes which will be based on random forest methods. The results of the work are a collection of comprehensible reference data sets with the data-based representation of the process sequence and first adaptations of explainable models in combination with domain knowledge and the analysis results learned.