转速大波动机械特征敏感参数提取与评估方法研究

As mechanical equipment including aviation power plant, large-scale engineering vehicles, wind turbine and Marine engineering equipment, etc. usually runs on working condition of large varying-speed, it causes the challenge for conventional condition monitoring technology. It is urgent to build the condition monitoring and diagnosis method for the mechanical equipment with large speed oscillation. Hence, the feature extraction and fault diagnosis method for the mechanical equipment under large varying-speed condition is proposed in this project. Firstly, considering vibration characteristics of mechanical equipment, the high precise time-frequency representation method is designed to uncover the key information buried in the mechanical signal. And then, the data fusion theory and local dynamic path optimization are presented to identify the targeted ridge curve with the complicated shape, respectively. Thirdly, the sensitive mechanical dynamic parameter extraction and the key component signal reconstruction method are studied based on the identified ridge curve information (instantaneous amplitude and frequency). Finally, the mapping relationship between the depth learning theory and the sensitive feature of mechanical equipment is explored, and the condition assessment templete of mechanical equipment is constructed to realize the mechanical equipment accurate diagnosis.

机械设备（航空动力装置、大型工程机械、风力发电机组及海洋工程装备等）运行过程中转速往往会发生剧烈波动，给当前机械状态监控技术带来了巨大的挑战。为此，转速大波动状态下设备监控与诊断成为一项急需发展的技术。本项目立足于实际需求与技术难点，提出转速大波动下的机械设备信息特征提取与诊断理论与方法。首先基于机械设备振动特性，建立转速大波动下机械信号关键信息的高分辨率时频表示技术；然后分别从数据融合理论和局部动态路径优化两个角度研究鲁棒的形态复杂特征脊线识别方法；再以特征脊线信息(瞬时幅值、频率)为基础，开展机械动态信号重构与敏感特性参数提取的方法研究；最后，以敏感动态特性参数为样本，研究大数据驱动下基于深度学习理论的转速大波动机械设备状态评估模板的设计方法，实现转速大波动下机械设备状态的准确诊断。

机械设备（航空动力装置、大型工程机械、风力发电机组及海洋工程装备等）运行过程中转速往往会发生剧烈波动，给当前机械状态监控技术带来了巨大的挑战。为此，转速大波动状态下设备监控与诊断成为一项急需发展的技术。本项目立足于实际需求与技术难点，提出转速大波动下的机械设备信息特征提取与诊断理论与方法。首先基于机械设备振动特性，建立转速大波动下机械信号关键信息的高分辨率时频表示技术；然后分别从数据融合理论和局部动态路径优化两个角度研究鲁棒的形态复杂特征脊线识别方法；再以特征脊线信息(瞬时幅值、频率)为基础，开展机械动态信号重构与敏感特性参数提取的方法研究；最后，以敏感动态特性参数为样本，研究大数据驱动下基于深度学习理论的转速大波动机械设备状态评估模板的设计方法，实现转速大波动下机械设备状态的准确诊断。