OptMeSys — Cross-domain optimization for the automated redesign of mechatronic actuation systems

机电驱动系统是实现各种任务自动化的基础。在其开发过程中，可以利用现有产品，因为由于客户特定要求而对先前产品进行的更改和重新设计相对较小。重用先前产品开发周期和优化迭代中的现有隐性知识可以显著减少重新设计的工作量，但由于缺乏结构化获取、链接和提供知识的方法而受到严重阻碍。相反，有必要从力学、电子学和信息学的各个领域收集信息，必须重新创建或手动调整跨领域的功能和行为模型，并且必须重新设置和解决优化问题。因此，本研究项目的目标是使产品开发过程的各个步骤自动化，这对于机电驱动系统的跨领域优化是必要的。一致的、基于知识的工程方法旨在避免耗时的重新设计迭代。这需要开发合适的自动化过程，用于识别特定于领域的产品特征，用于跨领域关系的基于模型的映射，以及用于连接不同的信息和模型，例如需求、解决方案原则，以及具有系统和行为模型的产品特征，等等。文本挖掘、本体、图形数据库和系统建模的使用以及它们的协调应该能够为时变机电系统的基于模型的设计自动定义优化问题。为了能够有效地解决由此产生的复杂优化问题，本研究项目侧重于利用以前产品开发步骤和优化的知识开发新的优化策略。除了选择合适的元启发式算法并对其进行设置，结合基于元模型的优化策略等提高其效率的方法外，对先前优化中获得的中间解的知识进行重用，旨在显著降低优化问题的复杂性。除了在不同项目阶段伴随该方法发展的代表性用例，例如在3D打印机中对齐挤出机和平台的电动窗口调节器和驱动系统，对进一步机电一体化产品的最终研究将证明其可转移性、潜力和重新设计中所面临的变化类型和程度的局限性。

Mechatronic actuation systems are the basis for the automation of a wide variety of tasks. In their development, it is possible to make use of existing products, since the changes and thus the redesign of predecessor products, such as due to customer-specific requirements, are comparatively small. Reusing existing, implicit knowledge from previous product development cycles and optimization iterations could significantly reduce the effort in redesign, but is strongly impeded by the lack of methods for its structured acquisition, linking and provision. Instead, it is necessary to regather the information from the individual domains of mechanics, electronics, and informatics, the cross-domain function and behavior models must be recreated or manually adapted, and the optimization problems have to be set up and solved once more. Thus, the goal of this research project is to automate the individual steps of the product development process, which are necessary for the cross-domain optimization of mechatronic actuation systems. A consistent, knowledge-based engineering method intends to avoid time-consuming redesign iterations. This requires the development of suitable automation processes for the identification of domain-specific product characteristics, for the model-based mapping of cross-domain relationships, and for the linking of diverse information and models, such as requirements, solution principles, and product characteristics with system and behavior models, among others. The use of text mining, ontologies, graph databases and system modeling as well as their harmonization should enable an automated definition of optimization problems for the model-based design of time-variant mechatronic systems. In order to be able to solve the resulting complex optimization problems efficiently, this research project focuses on novel optimization strategies using knowledge from previous product development steps and optimizations. Besides the selection of suitable metaheuristic algorithms and their settings in combination with methods to increase their efficiency, such as metamodel-based optimization strategies, the reuse of knowledge from intermediate solutions obtained from previous optimization aims at significantly reducing the optimization problem complexity. In addition to representative use cases accompanying the development of the method during the different project phases, such as electric window regulators and actuation systems to align extruders and platforms in 3D printers, final studies on further mechatronic products will demonstrate its transferability, potential and limitations with regard to the type and degree of changes confronted in redesign.