高分辨率同步压缩变换理论及其在风力发电机传动链故障诊断中的应用

Wind turbine often works in the complex conditions of variable speed and load. Hence, the key parts on its transmission chain are prone to failure and the fault signal often has typical non-stationary characteristics. Moreover, there are problems such as low resolution in time-frequency plane, hardly effective recognition the fault feature under variable speed and poor performance of early weak fault diagnosis. This topic focuses on the theory of high-resolution synchrosqueezing transform. Firstly, the method of parametric high-order synchrosqueezing transform is proposed, which is based on the parameterized demodulation analysis and high-order approximation of the amplitude and phase function. It is aimed to realize the high-resolution expression of the time-frequency curve about complex signal. Secondly, the multi-component signal adaptive decomposition algorithm based on synchrosqueezing reassignment vector (SRV) is proposed. Thus, the fault characteristic components and speed curves are effectively extracted. The problem of fault feature extraction under variable speed is solved by resampling to fault characteristic components in angular domain using speed curves and subsequently order analysis. Finally, the single channel signal processing method is extended to multi-channel by researching the modulated multivariate oscillations model, namely multivariate synchrosqueezing transform. It is designed to solve the problem of weak fault feature recognition. Through the research of this topic, the optimal time-frequency expression scheme for vibration signals at variable speed can be obtained, which is used to achieve the early weak fault diagnosis of key components in wind turbine transmission chain.

风力发电机长期运行在变速变载的复杂工况下，其传动链上的关键零部件容易产生故障，并且故障信号具有典型的非平稳特征，存在着时频分辨率低、变转速下故障特征难以有效识别、早期弱故障诊断的效果不强等问题。本课题重点研究高分辨率同步压缩变换理论，首先提出参数化高阶同步压缩变换方法，通过参数化解调分析和幅值、相位函数的高阶近似，实现复杂信号时频曲线的高分辨率表达。其次，提出基于同步压缩重排向量的多组分信号自适应分解算法，有效提取出故障特征成分和转速曲线，通过转速曲线对故障特征成分进行角域重采样和阶次分析，解决变转速下故障特征提取的问题。最后提出基于多元调制振荡模型的多元同步压缩变换算法，将单通道同步压缩变换拓展到多通道，解决弱故障特征识别的问题。通过本课题的研究，可以获得变转速下振动信号分析的最佳时频表达方案，实现风力发电机传动链关键零部件的早期弱故障诊断。

本项目按照计划完成了既定研究内容，在课题组原有早期微弱故障诊断理论研究的基础上，针对变转速下风力发电机的弱故障非平稳特征难以提取的问题，以同步压缩变换为基本的理论框架，分别从短时傅里叶变换（STFT）、小波变换（WT）和Stockwell变换（ST）入手，有针对性的提出了以同步压缩变换为基础的改进时频分析方法，解决了时频平面受噪声干扰大、强时变特征难以提取的问题。本项目主要的研究内容如下：1）针对STFT-SST进行了深入的稀疏表达、能量重排等理论及应用研究，提出了二阶时频稀疏表示方法、重排的二阶同步压缩变换RFSST2、基于TF域重排向量的信号自适应分解算法等；2）针对WT-SST进行了自适应参数选择、小波母函数优化的理论及应用研究，提出了多重二阶同步压缩小波变换AMWSST2、同步压缩超小波变换（SSLT2）等；3）针对ST-SST进行了自适应算法的理论及应用研究，提出了改进的二阶同步压缩S变换SSMST2等。通过本项目的持续研究，从以上3个不同角度丰富了高分辨率同步压缩变换理论的内涵，实现了复杂振动信号在时频平面的高分辨率表达和时变特征的准确提取，发表了11篇SCI论文，其中TOP期刊论文1篇，获得了湖北省自然科学奖三等奖1项，并将这些方法用于实验室故障模拟试验台和风力发电机工业现场设备的诊断分析，同时其相关理论研究成果在部分校企合作课题上得到了成功地应用，达到了预期的效果。