滚动轴承作为重大装备和主机产品的关键基础件，广泛应用于航空航天、轨道交通、风电设备等领域。滚动轴承局部缺陷激励机理、信号特征提取以及故障识别成为机械故障诊断领域的基础性关键问题。本项目拟研究滚动轴承缺陷三维形貌特征参数与振动响应波形特征之间的定量关系，建立轴承系统的非线性动力学模型，揭示滚动轴承局部缺陷及其动态扩展的激励机理；研究基于不对称高斯线调谐小波的自适应小波基函数、多共振分量、多参数全局优化方法相结合的共振稀疏分解特征提取方法，实现滚动轴承微弱耦合特征的有效提取；研究基于稀疏特征的多源异构域迁移学习诊断策略，通过最小化特征分布差异以及伪标签学习误差，构建滚动轴承迁移学习诊断模型，挖掘出仿真数据和实验数据中蕴含的诊断知识，用于解决滚动轴承小样本故障诊断的难题，并进行实验研究与工程应用。项目成果将为滚动轴承的故障诊断和状态监测提供基础理论和关键技术，具有重要的学术意义和工程价值。

As a key component of the important equipments and main machines, rolling element bearings are widely used in aerospace, rail transit, wind power equipment and other fields. The local defects excitation mechanism, signal feature extraction and fault identification of the rolling element bearings have become the fundamental key issues in the field of mechanical fault diagnosis. This project intends to study the quantitative relationship between the three-dimensional shape feature parameters of local defects and the waveform characteristics of the vibration response, establish a nonlinear dynamic model of the bearing system, and reveal the excitation mechanism of local defects. The Asymmetric Gaussian Chirplet Model-based (AGCM) adaptive wavelet functions, multiple resonance components and global multi-parameter optimization are studied to extract the weak and coupled feature of rolling element bearings. The transfer learning diagnosis strategy based on sparse features in multi-source heterogeneous domains is studied. The transfer learning diagnosis model of the rolling element bearing is constructed by minimizing the difference of feature distribution and the learning error of pseudo-label. Mining the diagnostic knowledge contained in the simulation data and experimental data to solve the small sample problem in the fault diagnosis of rolling element bearings, and experimental research and engineering application are carried out. The results of this project will provide basic theory and key technologies for fault diagnosis and condition monitoring of the rolling element bearings, with important academic significance and engineering value.