基于符号动力学熵的航空发动机主轴承早期故障检测与诊断方法研究

The key bearing is the key component in the rotating system of aero-engine. However, the key bearing often services in the harsh environment, such as high temperature, high speed and heavy load, which is prone to occur damage on key bearings. Therefore, the condition monitoring and fault diagnosis of key bearings is crucial to avoid the accident and ensure the safe operation of aero-engine. Firstly, the mapping relationship between the various fault types of key bearings with complexity is constructed based on the symbol dynamic entropy (SDE) theory. Then, the symbolic denoising method and parameter optimization strategy is studied to improve the fault extraction ability of SDE. Secondly, taking the difference between the healthy condition and fault condition as a bridge, an anomaly measure index that can capture the performance degradation information of key bearings would be constructed. Therefore, the physical degradation process of key bearings can be revealed, resulting in the accurate early fault warning of key bearings. In the end, the adaptive matching of time-frequency representation for the weak signal is studied. Using machine learning as the core, a strategy of combination of resonance-based sparse signal decomposition and hierarchical SDE is researched and the early fault type identification of key bearings can be realized. In conclusion, the technologies developed from this research project would enrich the entropy theory, which would further provide new theoretical basis and technical support for the timely detection and health management of key bearings. Consequently, the project has important scientific significance and engineering application value.

主轴承是航空发动机转子系统的关键部件，常常工作于高温、高速和重载的复杂服役环境，导致主轴承故障时有发生。因此，对主轴承进行早期异常检测和故障诊断对保障航空发动机的安全运行至关重要。本项目以符号动力学熵的复杂度表征为基础，建立主轴承不同故障模式与复杂度间的映射关联，研究符号化降噪方法与参数优选策略，提高符号动力学熵对故障信号的特征提取能力；以主轴承正常与故障状态振动信号的差异性为桥梁，研究基于多尺度符号动力学熵的主轴承性能衰退监测指标，揭示其物理退化规律，实现主轴承早期异常的准确报警；研究弱故障信号稀疏特征自适应匹配方法，以机器学习为核心，构建基于共振稀疏分解和层次符号动力学熵的故障特征提取框架，实现主轴承早期损伤位置的准确定位。本项目的研究可以丰富熵值理论方法，并为航空发动机主轴承的状态监测与健康管理提供新的理论基础与技术支持，具有重要的研究价值和工程应用前景。

本项目旨在针对航空发动机关键部件—主轴承的故障诊断难题，从基于符号动力学熵复杂度表征的故障微弱特征提取技术、基于共振稀疏分解和符号动力学熵的早期故障特征提取、航空发动机主轴承早期异常检测方法研究等三个方面，对故障早期异常检测和故障诊断两个层次的理论、方法、关键技术以及试验展开多学科交叉融合的研究。经过三年的研究，取得了如下成果：（1）针对已有熵值方法难以在低信噪比下进行微弱故障信号特征提取的难题，通过引入符号动力学滤波理论，提出了符号动力学熵，极大地提高了熵值对噪声的抑制能力和计算效率；（2）结合主轴承振动信号特点，提出了基于符号动力学熵的轴承故障特征提取技术，有效地检测到滚动轴承系统动力学特性的突变行为，实现滚动轴承不同损伤位置的准确区分；（3）在符号动力学熵的基础上，提出了层次符号动力学熵、广义复合多尺度符号动力学熵，进一步丰富熵值理论体系；（4）在符号动力学熵的基础上，结合基于特征的迁移学习思想提出一种新的特征提取方法—多尺度迁移符号动力学熵；与符号动力学熵相比，迁移符号动力学熵能够有效提升数据驱动模型的泛化能力，将熵值理论拓展到全新的领域。在实际场景中工况不同的情况下也能准确诊断旋转机械的故障类型，拓展了熵值理论的应用范围；（5）针对航空发动机主轴承振动信号慢变尺度和快变尺度的特点，结合多尺度分析优势，提出多尺度符号动力学熵用于描述主轴承振动信号在不同尺度下的动力学特性，构建用于描述主轴承性能衰退的监测指标；(6)设计和完成了轴承不同故障模式的实验以及加速全寿命试验，收集了振动等大量的实验数据，对提出的理论与技术进行了验证。基于以上研究成果，共发表论文25篇，包括SCI期刊论文18篇、EI会议论文7篇，出版英文专著1部，完成硕士学位论文5篇,申请国家发明专利8项。