Machine learning methods for adaptive process planning of 5-axis milling

5轴铣削自适应工艺规划的机器学习方法

• 批准号: 424298653
• 负责人: Professor Dr.-Ing. Berend Denkena, since 1/2022
• 金额: --
• 依托单位: Institut für Fertigungstechnik und Werkzeugmaschinen (IFW)
• 依托单位国家: 德国
• 项目类别: Research Grants
• 财政年份: 2020
• 资助国家: 德国
• 起止时间: 2019-12-31 至 2023-12-31
• 项目状态: 已结题
• 来源: https://gepris.dfg.de/gepris/projekt/424298653?language=en
• 关键词: Machine; learning; methods; adaptive; process

The proposed project aims to research a framework for a learning 5-axis compensation of shape errors in milling processes based on a process-parallel material removal simulation and sophisticated machine learning strategies. Moreover, we aim to investigate the ability of knowledge transfer between different workpiece geometries, milling tools and machine tools for an enhanced process planning. For this purpose, we will establish a framework that encompasses the functionalities needed to support a flexible and real-time-capable filtering, fusion and storage of data streams with different characteristics. Next, fundamental knowledge about the performance of different machine learning algorithms for building up process knowledge and design suitable supervised learning methods is provided. Based on this knowledge a method that identifies novel process situations automatically and decides whether a new model domain is necessary or if existing knowledge can be transferred, is researched. Finally, we plan to develop a compensation strategy for shape errors that combines an adjustment of the toolpath using 5-axis of the machine tool with a local adaption of the feed rate. Since production data is only available to a very limited extent to the scientific community, the experimental data sets and labels are made accessible online to the scientific community. This will allow other research groups to reproduce our findings and evaluate their own methods.

该项目旨在研究一种基于过程平行材料去除模拟和复杂机器学习策略的铣削过程中形状误差学习五轴补偿框架。此外，我们的目标是研究不同工件几何形状，铣刀和机床之间的知识转移能力，以增强工艺规划。为此，我们将建立一个框架，该框架包含支持具有不同特征的数据流的灵活和实时过滤、融合和存储所需的功能。其次，提供了关于不同机器学习算法性能的基本知识，用于建立过程知识和设计合适的监督学习方法。在此基础上，研究了一种自动识别新过程情况并决定是否需要一个新的模型域或是否可以转移现有知识的方法。最后，我们计划开发一种形状误差的补偿策略，该策略结合了使用机床的5轴调整刀具路径和局部适应进给速度。由于生产数据只能在非常有限的范围内提供给科学界，因此实验数据集和标签可以在线提供给科学界。这将允许其他研究小组复制我们的发现并评估他们自己的方法。