航空发动机主轴承故障稀疏时频分析与运行可靠性评估方法研究

The key bearing is considered to be the lifeblood of aero-engine. However, the key bearing services in extreme environment, such as high temperature, high speed, heavy load, and strong disturbance, which may result in serious accidents. Thus, fault diagnosis and prediction of the key bearing play a significant role in guaranteeing the safety and reliability of aero-engine, and even the whole airplane. Firstly, using advanced signal processing techniques as the foundation, a method using energy compaction band time-frequency accumulation spectrum based on fractional lifting wavelet transform would be studied to analyze the high-oscillated fault signal, and thus to improve the energy concentration and readability of the time-frequency representation for the high-oscillated fault signal. Then, the sparse time-frequency representation dictionary can be established by the proposed fractional lifting wavelet transform, and the object function for sparse time-frequency representation of the high-oscillated fault signal can be constructed and eventually solved fast by advanced optimization algorithms. As a result, the sparsity of the high-oscillated fault signal in time-frequency domain and the fault feature extraction accuracy can be improved. Secondly, using information fusion technology as a bridge, a running reliability evaluation index which can capture the performance degradation information of key bearing would be constructed. Consequently, the physical degradation processes of the bearing can be revealed. Thirdly, using machine learning as the core, a mapping relationship between the key bearing running state information and the reliability prediction model based on multi-variable and multi-resolution relevance vector machine(RVM) would be established, and then the accurate assessment of key bearing health status can be achieved. In conclusion, a fault diagnosis method based on sparse time-frequency diagnosis and operational reliability assessment for aero-engine key bearings would be proposed. The technologies developed from this research project would provide new theoretical basis and technical support for the timely detection of key bearings hidden faults, and the establishment of scientific maintenance strategies.

主轴承是航空发动机的生命线，常常工作于高温、高速、重载、强扰动等复杂服役环境，导致灾难性事故时有发生。因此，主轴承的故障诊断与预示技术对保障航空发动机乃至整个飞机运行安全至关重要。本项目以先进信号处理为基础，研究分数阶提升小波变换聚能带时频累加谱方法，提高故障振动信号时频表示的时频能量聚集性和可读性；构建故障信号稀疏时频表示字典，研究信号稀疏度的逼近和信号的精确重构快速求解方法，提高故障信号在时频域的稀疏性与提取精度。以特征信息融合为桥梁，构建能捕捉主轴承性能退化的运行可靠度评价指标，揭示主轴承的物理退化过程；以机器学习为核心，建立主轴承状态信息与多变量多分辨相关向量机可靠性预测模型之间的映射关联，实现主轴承运行状态健康精确评估。本项目在上述工作的基础上，提出航空发动机主轴承故障稀疏时频分析与运行可靠性评估方法，对及时发现航空发动机故障隐患，科学的制定维修策略提供新的理论基础和技术支持。

主轴承是航空发动机的“生命线”，其质量和性能直接影响到发动机运行安全。本项目瞄准主轴承安全服役、稳定运行中的重大科学问题，深入研究主轴承故障稀疏时频分析与运行可靠性评估方法，具体内容包括： ①设计了自适应平移不变分数阶B-样条小波变换，同时提出了结合非线性算子与一种新的不依赖噪声经验估计且具有对称或非对称阈值的广义硬阈值算法的小波稀疏增强技术，丰富和发展了稀疏表示框架理论。②基于经典的轴承局部故障振动模型，构造了融合轴承结构特征与运行工况的振动模型参数化时频原子字典，物理意义明确。同时利用鲸鱼优化算法提高了正交匹配追踪（OMP）对信号的稀疏表示效率与精度。③构建了轴承损伤的可靠度评价指标，建立了基于多变量第二代小波支持向量机寿命预测模型和性能退化特征之间的直接映射关系，为主轴承运行安全故障定量诊断提供可靠保障。本项目成果为航空发动机等重大装备运行安全科学的制定维修策略提供了新的理论依据和技术支撑，具有一定的实用性和普适性。发表国际知名学术SCI论文13篇、申请发明专利2项（公开）；培养毕业博士生1人，硕士研究生1人。项目申请人于2020年获得国家自然科学基金委面上基金资助(No. 52075080)。