齿轮动力学参数反求及其在故障机理研究和定量诊断中的应用

When faults occur in gear, the corresponding vibration signal would change for the dynamics parameters such as time-varying mesh stiffness and gear surface friction coefficient changed. Therefore, the study of the relationship among the fault state, the dynamics parameters and the vibration signal is of important significance for mastering the gear fault mechanism and fault diagnosis. Aiming at the existing problems in the gear fault mechanism research and fault diagnosis, the gear fault mechanism research and fault quantitative diagnosis methods based on the dynamics parameters reverse are researched in this project, which provides new ideas and approaches for the fault mechanism and diagnosis research. Firstly, the reverse computation method of dynamics parameters based on the tooth root stress is researched, and the accurate dynamics parameters for the establishment of various fault dynamics models are provided. Secondly, the gear fault mechanism research method based on the dynamics parameters reverse is proposed, base upon which the quantitative relationship between the dynamics parameters and the fault state is studied and the fault evolution law is revealed. Thirdly, the evaluation indexes of the characteristic parameters are introduced and the normalized feature vector extraction and analysis method based on the multiple transform domains is proposed; furthermore, the quantitative and mapping matrix relationships among the fault state, dynamics parameters and feature vector are constructed. Finally, the feature information enhancement and the multisource blind deconvolution separation methods of the gearbox vibration signal are studied, and the gear fault quantitative diagnosis method based on the mapping matrix relationship between the dynamics parameters and feature vector is proposed.

当齿轮发生故障时，振动信号发生变化的主要原因是时变啮合刚度和齿面摩擦因数等动力学参数发生了改变，因此研究故障状态、动力学参数和振动信号三者之间的关系对于掌握齿轮故障机理和故障诊断具有重要意义。项目针对目前齿轮故障机理研究和故障诊断中存在的问题，开展基于动力学参数反求的齿轮故障机理与故障定量诊断方法研究，为故障机理研究和故障诊断提供新的思路和方法。研究基于齿根应力的动力学参数反求计算方法，为建立各种故障动力学模型提供准确的动力学参数；提出基于动力学参数反求的齿轮故障机理研究方法，研究动力学参数与故障状态之间的定量关系，揭示齿轮故障演化规律；引入特征参数评价指标，提出基于多变换域的振动信号归一化特征向量提取和分析方法，构建故障状态、动力学参数和特征向量三者之间的定量和映射矩阵关系；研究齿轮箱振动信号特征信息增强、多源盲解卷积分离方法，形成基于动力学参数和特征向量映射矩阵的齿轮故障定量诊断方法。

项目开展了基于动力学参数反求的齿轮故障机理和齿轮故障定量诊断方法研究，揭示了齿轮故障机理，实现了齿轮定量故障诊断。主要研究工作和创新性成果有：（1）齿轮动力学参数反求及故障机理研究。基于动力学参数反求理论，研究了时变啮合刚度和时变摩擦因子计算方法，提出了基于参数反求的齿轮裂纹时变啮合刚度、齿轮点蚀故障下时变摩擦因子的计算方法。建立了各种故障齿轮动力学模型，研究了齿轮各种不同故障的振动信号响应特性，揭示了齿轮故障的演化规律。（2）齿轮定量故障诊断方法研究。研究了基于多变换域的振动信号归一化特征参数提取方法，将萤火虫算法（FA）、主模态分析（PMA）、变分模态分解（VMD）、多变量预测模型（VPMCD）等方法相结合，提出了基于FA-PMA-VMD和VPMCD的变转速齿轮故障定量诊断方法；基于振动信号的时频分布包含更丰富的故障信息，采用GP-奇异谱分解时频图，实现了变工况下的齿轮箱故障诊断；针对不同工况下的齿轮裂纹故障，提出了基于改进迁移学习的齿轮裂纹定量诊断方法。（3）齿轮箱振动信号自适应降噪和特征信息增强方法研究。针对目前已有降噪方法的缺陷，提出了基于拉格朗日乘子的辛奇异值模态分解降噪方法；针对待降噪信号中可能存在多种未知或已知噪声的情况，提出了基于变分贝叶斯学习的辛变换降噪方法；针对齿轮微弱故障特征提取问题，提出了基于复数微分算子和快速谱的信息增强方法。（4）齿轮箱盲源分离算法研究。从不同应用层面，分别提出了基于自适应辛几何模态分解、基于群分解和平均差值形态算子、基于改进的变分模态分解和稳健型广义形态分量、基于GP-奇异谱分解的齿轮箱振动信号的盲源分离方法。项目研究了齿轮啮合刚度和摩擦因子等动力学参数的变化对齿轮振动响应的影响，揭示了齿轮各种故障的演化规律，提出了适合于齿轮箱振动信号的降噪及信息增强、盲源分离和归一化特征参数提取方法，实现了齿轮的定量故障诊断。