核电管板深孔钻削刀具系统时变异常振动智能识别方法研究

The abnormal vibration of tool system is the key factor that restricts the improvement of deep hole machining quality, reliability and efficiency of nuclear tube sheet. Effective identification method for tool condition can provide evidence to suppress the abnormal vibration of tool timely. According to the time varying and weakness of dynamic response signals of tool, and the nonlinear relationship between abnormal vibration and its symptom, this research project aims to the nonlinear dynamical model of deep hole tool system and experiment are developed to reveal evolution mechanism of tool system abnormal vibration. The mapping relationship between abnormal vibration and its dynamic response is investigated. To realize the extraction of time varying and weak feature from dynamic response signal, the hierarchic matching sparse decomposition algorithm based on full frequency domain dictionary and weak feature enhancement algorithm based on synchronized compression is built. To realize the accurate identification and warning of the tool system abnormal vibration type and severity in nuclear tube sheet deep hole machining, the deep migration learning intelligent identification model is constructed. This research project is intend for providing the key technology for the accurate identification of the deep hole tool condition, which is significant for the suppression of abnormal vibration and the improvement of machining quality, reliability and efficiency of nuclear tube sheet.

深孔刀具系统异常振动是影响核电管板深孔加工质量、可靠性和效率的关键因素。目前由于尚缺乏快速有效的异常振动识别方法，使得深孔刀具系统异常振动难以实时在线抑制。针对深孔刀具系统异常振动特征信号的时变、微弱性，异常振动与其征兆之间关系的复杂非线性等特点，本项目研究基于凝聚建模法的刀具系统局部非线性耦合特性表征方法，构建深孔刀具系统动力学模型，揭示刀具系统异常振动与其动态响应的映射关系；研究基于全频域字典库的逐级匹配稀疏分解算法和微弱特征时频域同步压缩能量增强方法，实现动态响应中时变、微弱异常振动特征信号精确提取；研究构建适用于核电管板深孔刀具系统的深度迁移学习智能识别模型，实现刀具系统异常振动及其严重程度的实时精确识别和预警。本项目可为核电管板深孔刀具系统时变异常振动智能识别提供新方法，对于实现深孔刀具系统异常振动在线抑制、提高核电管板深孔加工质量、可靠性和效率，具有重要学术意义和工程应用价值。

深孔刀具系统异常振动是影响核电管板深孔加工质量、可靠性和效率的关键因素。快速有效的异常振动识别方法，可为深孔刀具系统异常振动在线抑制提供理论基础和依据，对提高深孔加工质量具有重要意义。本项目开展核电管板深孔刀具系统异常振动智能识别方法研究，主要研究内容和成果如下：(1) 考虑深孔刀具系统各局部接触面耦合特性，构建融入局部接触特性和钻削力的深孔刀具系统动力学模型；基于模型探究了深孔刀具系统动态响应特性，可为异常振动特征信号提取提供先验知识。(2) 考虑到背景噪声对主轴转频高倍频信号的污染，以及加工环境中背景噪声的高频和低幅值特性，提出了改进经验小波去噪技术的深孔刀具系统异常振动特征信号提取方法, 实现了动态响应中异常振动特征信号的有效提取。(3) 针对由于单传感器信号中特征信息不全面及特征信息微弱等问题导致深孔刀具系统振动状态辨识难的问题，提出了基于多变量时频同步压缩算法的多传感器信号融合和特征增强方法，实现了异常振动特征信息的全面提取和有效增强。(4) 针对进行大量深孔钻削实验、对大量振动信号进行标注成本高、耗时耗力等问题，提出了基于深度迁移卷积神经网络模型的深孔刀具系统异常振动智能识别方法，实现了深孔刀具系统异常振动的有效识别。本项目研究成果为深孔刀具系统异常振动智能识别提供了新方法。