Abstract:  Aiming at the problems that the form of robot milling chatters was complex and difficult to identify effectively, a robot milling chatter adaptive identification method was proposed. Firstly, to retain the information that could characterize the robotic milling states, the original vibration signals were separated. Then the difference between the power spectrum entropy was used to characterize the frequency distribution characteristics of vibration signals in different milling states, and the standard deviation of the original signals was used to characterize the time domain characteristics of vibration signals in robotic milling. The three-dimensional feature vector was input to the support vector machine to construct the adaptive recognition model of robotic milling chatters. The model was verified by experiments. The results show that the proposed adaptive chatter recognition model may accurately identify regenerative chatters and low-frequency chatters in robotic milling, and the recognition accuracy of stable, early regenerative chatters, severe regenerative chatters, low frequency chatters and no-load state reaches 93%, which is better than existing methods.

Key words:  , robotic milling, chatter recognition, feature extraction, intelligent model

摘要： 针对机器人铣削颤振形式复杂、难以有效识别的问题，提出一种机器人铣削加工颤振自适应识别方法。首先对原始振动信号进行信号分离，保留能够表征机器人加工状态的信息；然后采用不同信号功率谱熵之间的差值表征不同铣削状态下振动信号的频率分布特性，采用原始信号的标准偏差表征机器人铣削振动信号的时域特性。采用上述特征指标构建能够表征机器人不同铣削状态的三维特征向量矩阵，将特征向量输入支持向量机，构建机器人铣削加工颤振自适应识别模型。对模型进行验证，并与现有方法进行对比，结果表明，构建的机器人铣削颤振自适应识别模型能够准确识别机器人铣削加工过程中的再生颤振与低频颤振；对于稳定、早期颤振、剧烈颤振、低频颤振以及空载等状态的识别准确率达到93%，优于现有方法。

关键词: 机器人铣削, 颤振识别, 特征提取, 智能模型