基于时变噪声结构的自适应加权稀疏模型及航空轴承故障诊断  

Time-varying Noise Structure Inspired Adaptive Weighted Sparse Model for Aero-engine Bearing Fault Diagnosis

作　　者：张晗 田毅 杜朝辉 ZHANG Han;TIAN Yi;DU Zhaohui(School of Construction Machinery,Chang'an University,Xi'an 710064;Key Laboratory of Road Construction Technology and Equipment of Ministry of Education,Chang'an University,Xi'an 710064;School of Marine Science and Technology,Northwestern Polytechnical University,Xi'an 710072)

机构地区：[1]长安大学工程机械学院,西安710064 [2]长安大学道路施工技术与装备教育部重点实验室,西安710064 [3]西北工业大学航海学院,西安710072

出　　处：《机械工程学报》2023年第17期162-174,共13页Journal of Mechanical Engineering

基　　金：国家自然科学基金(52275085,51805040);陕西省自然科学基金(2020JQ-131);中央高校基本科研业务费专项资金(300102252201)资助项目。

摘　　要：航空发动机的非平稳运行工况、时变气流环境等导致传感器采集到的振动噪声信号具有时变特性,同时,主轴轴承高转速运行工况导致其故障波形混叠变异,难以有效识别诊断。针对该问题,揭示了航空轴承噪声信号方差的时变特性,提出了基于时变噪声结构的自适应加权稀疏诊断模型。首先针对混叠变异故障波形难以构造匹配的稀疏表示字典的问题,构造了与特征周期先验匹配的分割算子,分析了故障特征矩阵的低秩先验,建立了混叠变异特征的自适应奇异值分解(Singularvalue decomposition,SVD)字典;然后,揭示了航空轴承噪声方差的时变特性,构造了与噪声时变结构和特征奇异值分布模式相匹配的自适应权重矩阵;进而,构建了自适应加权稀疏诊断模型(Adaptiveweighted sparsediagnosismodel,AWSM),开发了AWSM的优化求解算法。仿真分析表明,提出算法可有效实现航空轴承时变噪声的滤除,从而可靠提取轴承微弱的故障特征。最后,航空轴承实验验证了该算法可以实现运行转速高达25000r/min、剥落面积为1.0mm^(2)的航空轴承故障诊断。Due to non-stationary operating conditions and time-varying air flows,the background noises embedded into aero-engine vibration signals have time-varying structures or different statistical variances in different temporal parts.Meanwhile,the fault waveforms of aero-engine bearing present an overlapping distortion morphology.Therefore,it is a challenging task to extract the latent and distorted fault patterns for popular fault diagnosis methods.Exploring the time-varying statistical variance structures,an adaptive weighted sparse diagnosis model is proposed.Based on the characteristic period of the fault signal,a segmentation operator is firstly designed to construct two-dimensional matrix with low-rank patterns,and then an singular value decomposition(SVD)Dictionary is established adaptively to obtain the sparse representation of overlapping distortion fault waveforms.Secondly,the time-varying variance structure of background noises is revealed by sufficient statistical analysis,and then a weighted matrix with different column values is designed to describe that time-varying priors.Furthermore,the low-rank patterns of data matrix are incorporated into the weighted matrix through designating different weights to all rows based on the singular values.Based on the sparse representing model and elaborated weighted-matrix,an adaptive weighted sparse diagnosis model(AWSM)is constructed and meanwhile an optimization algorithm is developed for it.Simulation analysis shows that the proposed algorithm can effectively filter out the time-varying noise of aero-engine bearings,and reliably extract the weak fault features of bearings.Finally,experiments indicate that the proposed algorithm can perform the fault diagnosis of aero-engine bearing with spalling area of 1.0 mm^(2)at speed up to25000r/min.

关 键 词：航空轴承 故障诊断 时变高斯噪声 稀疏表示 自适应加权矩阵 

分 类 号：TH133[机械工程—机械制造及自动化] TP206[自动化与计算机技术—检测技术与自动化装置]