航空发动机气路突变故障表征与稀疏诊断方法研究

The aero-engine is seen as the pillars of a great power. The inadequate basic scientific research on aero-engine fault diagnosis have become one of the bottleneck problems and restricted the development of aero-engine. Gas path key components faults, especially abrupt faults with characteristics of randomness, abruptness and nonlinearity, have threatened the reliability and safety of aero-engine and were hard to deal with through the existing methods , which makes the abrupt faults diagnosis become the basic scientific problem to be broken through urgently. In this attribution, the crucial common scientific problems presented in gas path abrupt faults diagnosis are discussed, and two aspects will be explored based on analysis and understanding of abrupt faults signal characteristics. By combining physical mechanisms and data driven approach, the first is to research on modeling and representation for gas path abrupt faults based on catastrophe theory considering the confluence of multiple factors. Then, the mechanism of sparse representation and quantitative diagnosis for gas path abrupt faults is explored, and the tracking and fault diagnosis methods which adapted to multimode signals of abrupt faults are proposed. Through the research of this project, we intend to solve the core scientific problem of nonlinear and multimode signal representation and sparse diagnosis of gas path abrupt faults. The key theories and the novel technologies mentioned above will be broken through, and provide new theoretical and technical support for ensuring the safety and reliability of aero-engine and driving the development of fault diagnosis technologies.

航空发动机作为“国之重器”，其故障诊断的基础科学研究不足成为了制约我国航空发动机发展的瓶颈问题之一。航空发动机气路关键部件故障，特别是具有随机性、突发性、和非线性特点的突变故障对航空发动机安全可靠性产生了极大的威胁，而现有方法难以对其实施跟踪诊断，使其成为了亟待突破的瓶颈基础科学问题。为此，项目围绕气路突变故障诊断所面临的关键共性科学问题，探索性地从突变故障信号特性分析与理解出发，采用物理机理和数据驱动相结合的方式，研究考虑多重因素融合的基于突变理论的航空发动机气路突变故障建模理论，探索气路突变故障信号稀疏表征与定量诊断的机制机理，提出适应于航空发动机多尺度多模态的气路部件突变故障跟踪与诊断方法，来解决航空发动机气路非线性、多模态突变故障信号特性表征与稀疏诊断等关键科学问题。项目预期的理论突破和技术创新，将为保障航空发动机安全可靠性和推动航空发动机故障诊断技术发展提供新理论和新方法。

航空发动机表征气路突变故障的参数呈现显著的稀疏特性，为强突变性、随机性和非线性的突变故障诊断难题解决提供了新思路。主要工作和创新点如下：.以某型涡扇发动机部件模型为基础，建立了气路部件突变故障表征模型，研究了不同故障模式下的气路参数变化规律，获得了气路突变故障状态的数据库，为后续故障诊断方法研究提供数据基础。围绕气路传感器限制难以准确表征气路部件状态的问题，结合改进遗传算法与禁忌搜索-蜂群算法提出了一种基于禁忌蜂群改进遗传算法的优化方法，设计了一种发动机气路故障诊断传感器优化方案，获得气路故障的最优传感器组合。考虑到气路传感器本身容易发生故障，研究了一种GA-RRLSSVR算法，设计并实现了基于GA-RRLSSVR的发动机气路传感器故障诊断、隔离与重构系统，有效实现了对发动机单个和双重传感器故障的检测、隔离与信号重构。.基于阻尼法和频率采样滤波研究了参数预处理方法，有效滤掉了气路热力学参数强突变性带来的尖峰现象。由于传统稀疏算法字典训练过程仅考虑编码的总体稀疏性，使得诊断结果缺乏实际物理意义，难以适应气路突变故障诊断的要求。项目通过对发动机部件特征原子组进行分类，结合正交匹配追踪算法，提出了改进K-SVD字典训练的稀疏诊断方法，该方法对部分部件突变故障的诊断精度较高，满足在线故障诊断基本要求。.考虑到正交匹配追踪方法存在误差局部积累现象，导致对个别部件的诊断效果较差，且固定稀疏度使得正交匹配追踪算法难以适应不同故障模式。基于StOMP和SWOMP算法分别对气路突变故障进行跟踪诊断，结果表明：StOMP算法诊断时效性较好，但算法精度提升不明显；SWOMP算法诊断精度较高，诊断精度较高，但计算效率有所下降。为此，结合StOMP算法的累计误差消除优势和SWOMP算法的求解精度不均调和策略，提出了基于多原子稀疏度自适应匹配追踪算法（MaSAMP）的气路故障诊断方法，该方法在保持较高匹配速度的同时能够保证了较高的匹配精度。