基于内置编码器信息的关节减速器多源扭振行为分析与智能健康监测

The wide equipment of built-in encoders opens a promising way for the health monitoring of modern electromechanical equipment. However, the fault responses of encoder signals to different kinds of defects are not yet clear, and the corresponding information decomposition and feature enhancement algorithms are relatively limited. Moreover, the applicability of encoder signal analysis in complicated working conditions is rather weak. The above shortcomings have seriously restricted the future applications of encoder signals in the field health monitoring. Taking robot joint reducer as the research object, this project aims to reveal the generation mechanism, propagation principle and signal characterization of multi-source torsional vibration via lateral-torsional vibration modeling. Based upon this, multi-scale difference, tunable harmonic sideband matching decomposition and target path compensation methods will be investigated and proposed to enrich the adaptive decomposition and feature enhancement approaches for encoder signals. Further research on information fusion and transfer learning strategy based on deep convolution neural network will be made thereafter, and through which, the mapping between the deep characteristic of encoder and the performance degradation of reducer is clarified, and then the health monitoring method of joint reducer under variable conditions is built. The research work in this project could not only extend the current scope of monitoring and diagnosis technology but also help to improve the predictive maintenance ability and intelligence level of industrial robots, which has important theoretical significance and engineering value.

内置编码器的广泛配备可为复杂机电装备的健康监测开辟新的途径。然而，编码器信号的故障响应机理尚未明晰、信息解耦和特征增强算法相对薄弱、以及复杂服役工况的适应能力不足等问题，严重制约了编码器信息在健康监测领域的应用前景。本项目以机器人关节减速器为研究对象，通过多激励源平移—扭转动力学建模，揭示关节减速器多源扭振行为的产生机理、传播机制及其编码器信号表征；通过多尺度差分、可伸缩谐波—边频匹配分解及靶向路径补偿方法研究，构建编码器信号的自适应分解和特征增强算法，提升早期故障和微弱损伤的辨识能力；进一步研究基于深度卷积神经网络的多工况信息融合和迁移学习策略，阐明编码器深层特征与和减速器性能退化之间映射规律，并以此构建关节减速器的变工况健康监测方法。本项目的研究工作不仅可拓展现有监测诊断技术的应用范围，也有助于提升工业机器人的预测性维护能力和智能化水平，具有重要的学术研究意义和工程应用价值。

鉴于关节减速器独特的服役工况和复杂的内部结构，其状态监测技术不仅需要建立新的信息传感手段及信号处理方法对其动态特征进行准确捕捉和辨识，还需要从机理上对早期损伤和其瞬态扭振特征之间的关联规律进行深入探索。为此本项目开展了以下三个方向的研究工作。1.在故障激励和编码器响应方面，构建了行星关节传动系统的多激励源扭转动力学模型，研究了不同裂纹损伤下编码器信号的响应特征，提出了基于数据驱动的减速器健康状态自评估方法，克服了传统诊断方法对先验信息和领域知识的依赖。2.在编码器信号解读和消噪方面，本项目研究了编码器信号的周期组稀疏差分特性，建立了基于周期组稀疏差分算法的编码器信号消噪技术，提出了编码器信号的广义差分滤波方法，实现了大噪声背景下角位置信息的精确、高效差分，为编码器信号的解读提供有效手段。3.在复杂工况的适应能力方面，项目探索了基于深度耦合密度网络的智能监测技术。该方法能够将振动和角加速度两种互补信息在高维非线性空间进行耦合，有效提高了复杂服役工况的适应性。除此之外，本项目还探索了编码器数据修复，以及基于编码器信息的冲击自感知问题。研究工作不仅能够突破现有监测方式的成本与技术瓶颈，也有助于提升装备核心部件的服役安全性和运行可靠性，具有重要的学术意义和工程价值。