深度领域适应的空间滚动轴承变工况寿命状态匹配表征和预测方法

Space rolling bearing is one of the vital kinematic components in the spacecraft, and the length of its life will directly influence the normal operation, planed space missions completion and lifetime of the spacecraft. Therefore, it is in urgent need to identify and select optimal-life space rolling bearing. According to this, aiming at the obstacles of variation operating conditions encountered, such as speed variation, in-band interference, load-variation, etc. which make the life state representation and prediction very difficult, a novel method is proposed in this project for the space rolling bearing feature extraction and lifetime estimation. The research will be conducted mainly on the basis of the deep domain adaptation (DDA) theory and matching representation of space bearing life state under variation operating conditions. The outline of the research contents are as follows. For signal pre-processing purpose, a signal re-sampling method based on tacho-less order tracking and an in-band interference suppression method based on local and global space noise robust mapping, is proposed to process the interference caused by speed variation and the heavy background noise. On this basis, the deep feature extraction model based on DDA theory is constructed for space rolling bearing performance degradation process characterizing. Meanwhile, a breakthrough is hoped to be made in the extraction of rolling bearing features with consistent distribution under load-variation operating conditions, besides, the difficulties in matching the estimated bearing performance degradation trend with the law of actual degradation will be addressed. And then, the life of each sample is predicted and identified. On this basis, the life of the space rolling bearings to be selected is evaluated based on bootstrap maximum entropy method. The relevant research achievement in this project, will not only provide theoretical and technical support for the spacecraft to successfully accomplish the missions, but also enlighten the development of solutions for solving the bottleneck problems in rotary machinery life state representation and prediction under variation operating conditions.

空间滚动轴承是航天器的关键活动部件，其寿命的长短关系着航天器能否实现预定功能并达到预期寿命，因而需要进行空间轴承的寿命鉴定和筛选。本项目针对由变速、同频干扰以及变载等变工况干扰引起的空间轴承寿命状态表征和预测困难的问题，研究基于深度领域适应的轴承寿命状态匹配表征和预测新方法。该方法的研究思路为：提出无键相重采样的变速干扰抑制方法和高维空间局部与全局流形噪声鲁棒映射的同频干扰抑制方法；构建基于深度领域适应的空间轴承变工况性能退化特征匹配表征模型，重点突破变载工况轴承特征分布领域不一致问题，以及性能退化趋势空间和寿命退化规律不匹配的难题。对轴承样本进行寿命预测，并据此进行待选空间轴承短期跑合试验，实现基于自助最大熵小子样估计的待选轴承寿命评估。本项目研究将为保障航天器遂行飞行任务提供理论与技术支撑，也为旋转机械装备变工况寿命状态表征和预测方面亟待解决的瓶颈问题提供解决方案。

本项目以空间滚动轴承为研究对象，以待选空间轴承寿命评估为目标，研究空间滚动轴承变工况寿命状态表征和预测方法。项目执行过程中，研究了空间轴承地面模拟真空环境试验过程中的转速干扰抑制机制，形成了一套转速波动干扰无键相抑制方法，该方法在航空发动机附件齿轮系统、风力发电机增速齿轮箱等传动系统得到了应用验证；建立了深度平衡领域适应网络模型，通过对深度网络进行逐层领域适配的方式，解决了由变载工况导致的空间轴承磨损振动特征分布不一致的问题，实现了变载工况条件下轴承特征的有效提取；项目负责人所在课题组完成了地面模拟真空环境条件下的空间轴承寿命截尾样本和全寿命试验，建立了轴承寿命样本数据库；针对空间轴承全寿命样本总体稀少而无标记寿命样本较多的问题，建立了基于关联网络的小样本数据扩充和模型增强学习策略，实现了小样本条件下的轴承寿命评估；对本项目涉及的关键技术进行了试验验证和有效性评估。截至目前，基于本项目已在国内外著名学术期刊或学术会议发表论文12篇，其中，SCI论文8篇，主要发表在国际著名学术期刊《IEEE Transactions on Industrial Informatics》、《Mechanical Systems and Signal Processing》、《IEEE Transactions on Instrumentation and Measurement》等。.项目负责人基于本项目研究成果，结合关于直升机附件齿轮传动系统相关的军工科研项目研究结果，发现航空高速锥齿轮在运行工程中面临着与本项目十分相关的共性问题，为此结合前期技术积累和关键问题的探索发掘，申请获批了2021年度的国家自然科学基金面上项目，该项目将进一步依托本基金项目取得的各项研究成果，针对直升机附件齿轮传动系统早期故障迁移诊断面临的动力学机理分析、溅射噪声干扰、多源行波信号复杂耦合以及受限样本条件下的迁移诊断等方面的问题，开展进一步深入研究攻关，通过对行波声学信号进行深度解耦分析处理，以突破传统基于振动分析的附件齿轮监测诊断方法对早期故障不敏感的问题。.在本项目不仅为空间轴承寿命状态评估提供技术和理论支撑，也对解决航空装备故障诊断面临的问题具有借鉴意义。