复杂曲面制造的颤振特征敏捷稀疏表示与工艺健康性定量诊断

Chatter vibration is a harmful effect in manufacturing of complex surface. In this project, we attempt to make insightful analysis to online condition monitoring data, separate incipient chatter vibration symptom from non-stationary, non-linear, multi-component coupled dynamic signals, and to reveal failures on cutting tools, mechanical components, operational performance degradation rule of manufacturing process. This project attempts to search improvements in the following aspects: (1) Agile condition monitoring system: A fix observation point based bid data collection system for vibration signal is established, and the vibration signal is associated with its simultaneous machined surface image and process information. (2) Novel tight frames of non-conventional time-frequency topology are investigated. These frame is equipped with centralized multi-resolution, time-frequency directional sensitivity, and inter-scale analyzing ability. Moreover, the performance of higher dimensional multi-scale geometric analysis is enhanced. The proposed tight frames are used as sparse dictionary. (3) Nonlinear geometric observation operator and non-convex penalty function are introduced into sparse representation problem so as to enhance the intrinsic dynamic regularization. Parametric treatment form non-convex problem to convex problem is investigated. As a result, combined with the proposed sparse dictionary, simultaneous multi-component sparse separation to the coupled signal is enabled. (4) Investigation of dynamic feature evolution based on large time-scale iso-manufacturing process information so as to realize quantitative evaluation of machine-tool operational performance and manufacturing process health. The research results are beneficial to the development of non-chatter process environment and also to promote high accuracy and high efficiency manufacturing of complex surface. This project is of theoretical significance and engineering value.

颤振是危害复杂曲面加工的顽疾。本课题深入分析切削过程动态数据，从非平稳、非线性及多重特征耦合响应信号中分离微弱颤振特征，揭示不合理工艺参数、刀具零部件失效和系统运行特性劣化现象。研究固定观测点振动信号的完整无遗漏“拟大数据”采集方法，建立其与加工表面图像、工艺参数按工步关联的敏捷监测系统；基于双树复小波研究具有中心极化、时频方向敏感性、跨时间尺度分析特色的非常规时频拓扑过完备紧框架，并以此提升高维多尺度几何分析性能，通过它们增强稀疏字典学习；引入非线性几何观测器和非凸罚函数强化信号内禀动力学信息的正则约束，研究参数化的整体凸优化稀疏表示方法，使构造的字典形成对信号中多重耦合特征的同步稀疏优化能力；根据“工步同构-动态信号可对比性原理”，研究大时间尺度下动态特征驱动的工艺健康性定量诊断研究。研究成果有益于发展复杂曲面低颤振工艺环境，促进复杂曲面高精高效加工，具有重要理论意义和工程价值。

本项目针对复杂曲面加工过程的正问题及反问题进行了深入的研究。通过数值环境模拟了复杂曲面加工过程的颤振预测；通过无人工干预/少人工干预的方法对制造过程的大数据进行了智能分析；提出自适应的框架字典及稀疏分解算法对复杂曲面加工过程的动态信号进行了分析；在多轴联动数控机床上对提出方法及研发的算法进行了加工验证。