旋转机械振动信号成分复杂，在时变工况下具有非平稳特点。如何从复杂非平稳振动信号中有效提取故障信息，是旋转机械故障诊断中的一个关键共性科学问题。以具有典型代表意义的转子和行星齿轮箱为研究对象，针对振动信号的成分复杂性以及时变工况下的非平稳特点，通过理论分析、仿真计算和实验研究等手段，研究自适应迭代广义解调信号分析方法，解决旋转机械故障诊断中涉及的复杂多分量非平稳振动信号分析理论、方法及应用层面的关键问题，包括信号真实分量的自适应辨识、任意时变带状时频通域滤波、单分量迭代分解、高性能时频分析、增强阶比分析和瞬态轴心轨迹分析等方法。挖掘发挥自适应迭代广义解调方法的独特优势，突破常规方法的局限，避免主观因素的影响以及对经验知识的依赖，抑制背景噪声干扰，有效分析复杂非平稳信号中蕴含的丰富信息，准确提取幅值、频率、轴心轨迹等参数的变化特征，全方位多角度揭示旋转机械的健康状态，识别故障原因。

Rotating machine vibration signals feature complex multi-components and nonstationarity under time-varying running conditions. How to effectively extract fault information from such complex and nonstationary signals is a common yet important topic in rotating machine fault diagnosis. In this project, with rotors and planetary gearboxes as the representative research target, and regarding the complexity and nonstationarity of their vibration signals, adaptive iterative generalized demodulation method will be investigated, and its application issues in rotating machine fault diagnosis will be addressed, through theoretical analysis, numerical simulation and experimental validation. Research topics include adaptive identification of true signal components, time-varying filtering within arbitrary time-frequency region, iterative mono-component decomposition, quality time-frequency analysis, enhanced order spectrum analysis and transient orbit analysis. By exploiting the unique advantages of adaptive iterative generalized demodulation method, the limitations inherent with most conventional methods will be overcome, the background noise will be suppressed, the influence due to subjective factors and the dependence on a priori knowledge will be avoided. The rich characteristic information contained in complex nonstationary vibration signals is expected to be effectively extracted, e.g. the fundamental yet key parameters such as amplitude, frequency, and orbit, as well as their instantaneous changes. Therefore, the fault vibration mechanism of rotating machine will be thoroughly revealed, and the health condition will be well identified.