基于多源迁移学习的变位姿刀尖点模态参数预测

Pose-Dependent Tool Tip Modal Parameters Prediction Method Based on Multi-Source Transfer Learning

切削颤振会导致被加工工件表面质量变差、材料去除率降低以及刀具磨损增加等问题。刀尖点模态参数是构建稳定性叶瓣图、选取无颤振加工参数必不可少的输入。然而在加工过程中刀尖点模态参数随刀具位姿而变化且刀具更换频繁,经典锤击试验方法效率低、成本高,如何准确高效地预测变位姿下的刀尖模态参数成为切削加工中亟待解决的问题。本文结合迁移学习思想,提出一种基于多源迁移学习的变位姿刀尖点模态参数预测方法。当更换新刀具后,仅需通过锤击试验获取少量位姿下的刀尖点模态参数,再结合已有多把刀具的模态参数数据进行多源迁移得到新刀具的刀尖点模态参数预测模型。最后,在实际五轴机床上进行试验,试验表明所提方法是有效的。

Cutting chatter can lead to problems such as poor surface quality of the machined workpiece,reduced material removal rate and increased tool wear.The tool tip modal parameters are essential inputs for constructing stability lobe diagrams and selecting chatter free machining parameters.However,in the machining process,the tool tip modal parameters change with the tool pose and the tool changes frequently,and the classical impact test method has low efficiency and high cost,so how to accurately and efficiently predict the tool tip modal parameters under the changed pose has become an urgent problem to be solved in the cutting process.In this paper,combined with the idea of transfer learning,a method of modal parameter prediction based on multi-source transfer learning is proposed.When a new tool is used,the tool point modal parameters under only a few positions need to be measured through impact test,and then the tool point modal parameter prediction model for the new tool can be obtained by multi-source transfer combined with the modal parameter data of multiple existing tools.Finally,a practical experiment on a five-axis machine tool shows that the proposed method is effective.