基于延拓核主成分分析的空间滚动轴承性能演变趋势预测方法研究

Equipment failure can be divided into several periods: incubation period, infancy period, extended period, expiration period. The space rolling bearing as the most critical activities parts of the spacecraft is also true. The project proposed to use the space rolling bearing vibration signal of incubation period obtained by terrestrial analog vacuum environment to achieve the performance evolution trend prediction in the framework of extension kernel principal component analysis. The specific research ideas are: Proposed the extension kernel principal component analysis theory for space rolling bearing performance evolution trend prediction, based on the theory we can achieve space rolling bearing feature extract, state dynamic assessment and performance evolution trend prediction in the incubation period; Proposed to expand kernel principal component and reconstruct the weak feature of the space rolling bearing in low dimension, achieving the feature extraction of incubation period in strong background noise; Proposed to embed the intelligent decision-making mechanism into the main component of the feature space so as to achieve the state dynamic assessment, break the problem of the feature in the incubation period that status information vague, without obvious regularity, lead to it is difficult to use the feature to achieve the performance evolution trend prediction, so as to dynamic assessment of bearing condition; Proposed embed the forecasting theory into the multicore main component of the feature space , to achieve the across time scales prediction. This proposed method is of great significance that it can guide the design and manufacture for space activities parts, improve the orbit life and the using performance for the follow-up spacecrafts.

设备故障大体可分为几个周期：潜伏期、萌芽期、扩展期、失效期。作为空间飞行器中关键活动零部件之一的空间滚动轴承亦如此。针对典型的预测方法不适于空间滚动轴承性能分析的问题，项目提出在延拓核主成分框架内利用地面模拟真空环境下轴承潜伏期振动信号进行演变趋势预测的新方法。研究思路为：提出基于延拓核主成分分析的集微弱特征提取、状态动态评估及连续趋势预测等功能于一体的轴承演变趋势预测理论体系；提出低维主成分重构的轴承微弱特征提取新方法，实现强背景噪声下轴承微弱特征的提取；提出核主成分空间嵌入智能决策机制的动态状态识别新方法，突破潜伏期特征状态信息模糊、规律不明显，难以被用来进行演变趋势预测的难题，实现轴承状态的动态评估；提出趋势预测理论融入多核主成分特征空间的趋势预测新方法，实现轴承跨时间尺度的连续趋势预测。该方法对指导我国空间飞行器的设计与制造、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。

空间滚动轴承寿命状态评估和演变趋势预测技术对于我国空间航天技术的发展具有重要意义。课题负责人提出了基于延拓核主成分分析的地面模拟真空环境下空间滚动轴承性能演变预测和状态识别模型。该项目进行了地面模拟真空环境下空间滚动轴承振动信号处理、特征信息的提取、轴承状态诊断和分析、演变趋势预测、剩余寿命评估等研究。通过对地面模拟真空环境空间滚动轴承振动信号分析处理，对轴承的振动信号的特征属性进行了深入挖掘，利用人工智能模型揭示了地面模拟真空环境下空间滚动轴承状态的演变规律和寿命的表征形式。具体成果如下：.针对特征信息表征困难，特征提取现有算法无法准确提取轴承状态特征的难题，提出了基于改进核主成分分析的特征信息提取和特征属性表征方法；.针对振动特征状态模糊，研究了在核主成分空间进行失效特征识别和评价的规则，分析了特征空间中围绕轴承失效的特征指标的表现特性和映射规则，建立了映射空间中轴承失效特征属性和状态分布趋势之间的关联机制，提出了基于核主成分分析和模糊C均值聚类算法的地面模拟真空环境空间滚动轴承寿命状态评估模型（轴承状态评估准确率高于75%）；.针对轴承特征趋势无从表征的难题，研究了轴承演变趋势在映射空间的特征属性，提出了特征多迭代实现滚动轴承演变趋势预测的模型，实现了对轴承退化趋势的预测和跟踪；.针对经典的轴承寿命预测理论如L-P模型，I-H模型需获取大量寿命样本数据才能实现，而地面模拟真空环境试验代价大、周期长，获得大量寿命样本是不可行的难题，提出了少样本情况下多支持向量机模型预测误差加权融合的空间滚动轴承剩余寿命预测理论，从信号处理角度解决了在加速寿命环境下的轴承剩余寿命预测问题。.项目主持人以第一作者或通讯作者发表SCI检索期刊11篇，其中一篇论文获得2016年重庆市科协首届自然科学优秀学术论文,EI检索论文1篇，团队成员发表SCI论文5篇，专利2项。