不同工况振动谱的空间滚动轴承寿命同尺度表征和预测方法研究

Space rolling bearing is one of the most critical activity components that affect the using life and high reliability to spacecraft. In view of the common bearing friction torque, temperature test, etc. can not effectively predict the space rolling bearing life, a novel method concerning life identification and prediction for space rolling bearings based on vibration spectrum under different working condition with the same criterion is proposed. The research stratagem of this proposed method is as follows. A novel idea, which integrates the vibration spectrum acquisition, the life state feature vector extraction, the life state identification and prediction under the multi-condition with same criterion, will be proposed to form the new principle and the new methods for space rolling bearing life state identification and prediction. A novel method of life feature vector mapping under the multi-condition for space rolling bearing will be proposed. This method can solve the problem for space rolling bearing that the life state feature vectors are not comparable under different working conditions, and achieve space rolling bearing life state representation under the multi-condition with same criterion. A novel method of life state identification and prediction under small samples with different classes will be proposed for space rolling bearing prediction. Finally, a system prototype of space rolling bearing life state representation and identification based on vibration spectrum will be developed and system performance will be evaluated. This proposed method is of great significance that it can guide the design and manufacture for space activities parts, establish the evaluation criteria for service life and improve the orbit life and the using performance for the follow-up spacecrafts.

空间滚动轴承是影响空间飞行器高可靠长寿命的最关键的活动零部件之一。针对常用的轴承摩擦力矩、温度等测试手段不能有效预测空间滚动轴承寿命的问题，提出利用轴承振动信号进行不同工况下空间滚动轴承寿命状态同尺度表征和预测的新方法。该方法的研究思路为：提出集微弱振动特征提取、同尺度表征、寿命状态识别与预测于一体的空间滚动轴承寿命表征和预测理论体系；提出不同工况条件下轴承寿命特征向量间的映射新方法，突破不同工况下轴承寿命状态特征向量间不具有可比性的难题，实现不同工况下空间滚动轴承寿命状态同尺度表征；提出小子样非类同条件下的空间滚动轴承寿命状态识别与预测方法，实现空间滚动轴承寿命的预测；最后研发基于振动谱的空间滚动轴承寿命状态表征与预测系统原型，进行系统性能评估。该方法对于指导空间活动零部件的设计与制造、建立使用寿命的评估准则、提高后续空间飞行器的在轨寿命和使用性能都具有重要的意义。

针对常用的轴承摩擦力矩、温度等测试手段不能有效表征、识别和预测空间滚动轴承寿命的问题，本项目提出了利用轴承振动谱进行不同工况下空间滚动轴承寿命状态同尺度表征和预测的新方法。通过项目执行，完成了以下研究工作：（1）提出了地面模拟真空环境中背景干扰的EEMD滤除方法，有效滤除了背景干扰噪声。通过优化参数得到良好的EEMD分解结果，并根据背景干扰和振动信号的IMF分量间的相关性，设计了自动选取IMF分量算法以重构信号，实现了背景干扰的有效滤除；（2）提出了空间滚动轴承寿命敏感特征集构建和基于流形学习的寿命状态识别方法。针对不同域特征量对寿命敏感性的差异性，设计出了敏感性指标算法来进行寿命敏感特征的选择以构建寿命敏感特征集。同时，利用流形学习算法对高维寿命敏感特征集进行维数约简后，将约简后分辨特性好的低维寿命敏感特征集输入最近邻分类器进行分类识别完成了空间滚动轴承寿命状态的有效识别；（3）提出了基于幅值标准化处理和相似性度量的量化特征提取方法，实现了不同工况下滚动轴承退化状态进行同尺度表征。并提出了基于BP神经网络—马尔科夫模型的轴承寿命退化预测方法，实现了轴承寿命退化预测；（4）提出了基于深度神经网络的空间滚动轴承寿命特征自适应提取与智能识别方法，通过多层无监督逐层自学习和有监督微调，完成寿命特征自动提取与表达，并实现了不同工况下寿命的智能识别。最后研制出基于振动谱的空间滚动轴承寿命评估系统，并进行了初步应用。该项目的研究对于指导空间活动零部件的设计与制造、建立使用寿命的评估准则、提高后续空间飞行器的在轨寿命和使用性能都具有重要的意义。