航空发动机中介主轴承剥落故障的数据驱动预测方法研究

Failures of the rear intershaft main bearing is one of the leading causes of the inflight shutdowns, unscheduled engine removals, and mishaps of some single engine powered fighter, which often appear as the surface initiated spall faults. In order to develop effective, robust and reliable monitoring device with sufficient leading alarm time, it is urgent to study the spall fault detection and prognostics to know how to detect the incipient fault as early as possible and when to remove the bearing and issue the warning or alarm. We propose to study the spall prognostics of the intershaft bearing with spalled outer running race based on theoretical study, failure analysis, and test validation via a novel intershaft bearing test rig with rear turbine support and outer case parts from some turbofan engine. As for the complicated vibration transfer path and fluctuations of the spools, we propose a new weak signal separation and enhancement method based on synthesized speed difference tachometer synchronous sampling, minimum entropy deconvolution, spectral Kurtosis and the squared envelope spectrum analysis to filter non-bearing related vibration responses or noises, and enhance the signal-to-noise ratio. As for the incipient fault detection, we propose to use the derived vibration features to build a support vector machine fault detector and use the oil debris features to build an oil debris fault detector separately, and then fuse the two detectors’ decisions via fuzzy inference. As for the spall prognostics, sensitive features with good monotonicity and trend are firstly selected based on the correlation with the spall size, and then Gaussian process regression is used to predict the remaining useful life of the bearing respectively, finally fuzzy inference is used to fuse the prediction results. The alarm threshold is set by letting the spall size be equal to the length between two rolling elements, and a novel low false alarm fault detection threshold determination method is proposed. All these achievements can provide the science basis and technology support to the development of the intershaft main bearing monitoring device with sufficient leading alarm time.

后中介主轴承失效是导致某型单发飞机空中停车、非计划换发以及等级事故多发的主导原因之一，为研制有效、鲁棒可靠的大提前量预警装置，迫切需要研究其剥落故障早期检测和剥落传播建模问题。本项目以某型发动机后中介主轴承外滚道剥落故障为对象，依托带涡轮支承和外机匣的中介主轴承试验器，通过理论研究、失效分析、试验验证等，针对复杂振动传输路径以及高低压转子振动耦合等特点，提出基于转差域合成同步采样、最小熵去卷积、谱峭度值、平方包络谱等的微弱振动信号分离与增强方法；提出支持向量机振动检测和滑油碎屑检测方法及综合振动和滑油碎屑的融合检测方法；基于剥落相关性确定振动敏感特征，提出基于振动敏感特征和滑油碎屑总累计颗粒的高斯过程回归预测方法和综合振动、滑油碎屑的模糊推理融合预测方法；将剥落尺寸等于滚子间距作为报警阈值，提出低虚警率检测和报警阈值设定方法。为研制后中介主轴承大提前量预警装置提供科学依据和技术支撑。

针对导致某型单发飞机空中停车、二等事故的某系列发动机主轴承失效问题，研究了某系列发动机中介主轴承故障的高频振动监测方法体系，揭示了不同损伤程度不同故障模式下的振动敏感特征和规律，建立了基于高频振动敏感特征和滑油碎屑累积颗粒数特征的主轴承故障检测阈值设定方法。主要研究内容和重要成果包括：.1.建立了某系列航空发动机中介主轴承剥落故障的微弱信号分离和增强方法体系.（1）建立了某型发动机中介主轴承剥落故障信号的传递模型、特征频率计算模型，提出了适用于航空发动机中介主轴承的振动监测和滑油碎屑监测方案。.（2）为解决转速波动问题，提出了一种航空发动机专用的转差重采样阶次跟踪预处理新方法。.（3）为滤除确定性频率成分、考虑传输路径的影响，提出了在转差阶次域中基于线性预测和基于峭度最大的最小熵去卷积微弱信号分离和增强方法。.（4）提出了一种基于转差重采样阶次分析和快速谱关联分析的主轴承故障检测新方法，在发动机服役环境下验证了某中介主轴承滑蹭故障检测的有效性。.2.建立了综合高频振动故障检测和滑油碎屑故障检测的后中介主轴承剥落故障融合检测方法.（1）在单体轴承检测仪上，完成了内外圈滚动体等部件不同损伤程度故障下的振动特征研究，验证了单体轴承环境下特征频率检测的有效性。.（2）考虑挤压油膜阻尼器和复杂振动传输路径的影响，完成了外圈预置不同损伤程度故障下的典型振动敏感特征研究，揭示了试验器环境下的主轴承振动特征变化规律。.（3）提取了某系列发动机中介主轴承故障的振动时域和频域敏感特征，建立了中介主轴承的振动敏感特征检测阈值设定方法以及剥落故障综合振动检测方法。.3.建立了综合高频振动故障检测和滑油碎屑故障检测的后中介主轴承剥落传播模糊推理融合预测方法.（1）依托小轴承试验器，完成了剥落至失效的全过程试验，建立了发动机某附件机匣小轴承基于剥落总累积磨损碎屑数量的滑油碎屑检测方法和初始设定阈值。.（2）基于高频振动和总累积滑油碎屑颗粒计数特征，揭示了某型轴承失效全过程的铁磁性金属碎屑颗粒数特征和振动敏感特征的变化规律。.（3）利用模糊推理方法，建立了融合振动故障检测和滑油碎屑颗粒特征检测的决策级融合预测方法。.项目研究成果对于破解某型发动机主轴承大提前量预警难题、研制发动机主轴承健康诊断系统具有重要参考价值。