基于改进EEMD和WVD联合时频分析的车轮多边形状态识别方法

为了准确识别高速列车车轮多边形状态以及磨耗幅值，提出了一种改进的聚合经验模态分解(EEMD)与魏格纳-威尔分布(WVD)相结合的随机振动信号联合时频分析方法；利用相关系数法和频谱分析来评估筛选轴箱振动加速度信号经EEMD分解后的变量，然后进行WVD计算，在保持WVD高时频分辨率的同时可有效抑制交叉干扰项；应用该方法分析了周期性车轮多边形磨耗与现场实测随机车轮多边形磨耗引起的轴箱振动加速度信号。研究结果表明：利用EEMD-WVD二维时频谱的主频率可识别车轮多边形状态，利用EEMD-WVD三维时频能量谱的能量幅值分布可评估车轮多边形磨耗幅值，最大误差为0.3%；将改进EEMD和WVD联合时频分析方法的识别结果与短时傅里叶变换、小波分解、WVD传统时频分析方法进行对比，表明此方法应用时无需改变任何参数，自适应强，保留了WVD高时频分辨率的特点，而且可有效抑制EEMD产生的模态混叠现象和WVD产生的交叉干扰项，验证了所提出联合时频分析方法的有效性及其优势，为高速动车组车轮多边形识别和评估提供了新的技术途径。

Wheel polygon state recognition method based on improved EEMD-WVD joint time-frequency analysis

To accurately identify wheel polygon state and wheel abrasion amplitude of high-speed EMUs, a random vibration signal's joint time-frequency analysis method combing improved ensemble empirical mode decomposition (EEMD) and Wigner-Ville distribution (WVD) was presented. The correlation coefficient method and spectrum analysis were used to evaluate and filter the EEMD decomposition variables of the axle box vibration acceleration signal. Subsequently, the WVD of each intrinsic mode component was calculated for maintaining a high time-frequency resolution and effectively curbing cross-interference items. The method was applied to analyze the vibration acceleration signals of axle box caused by periodic and measured random wheel polygons. Research results indicate that the wheel polygon type can be recognized using the dominant frequency of the two-dimensional time-frequency spectrum of EEMD-WVD, and the abrasion amplitude can be evaluated using the energy amplitude distribution of the three-dimensional time-frequency-energy spectrum of EEMD-WVD, with the maximum error of 0.3%. Compared with the time-frequency analytical results by the short-time Fourier transform, wavelet transforms, and WVD method, the improved EEMD-WVD joint time-frequency analysis method dose not require parametric variation, has strong adaptability, retains the characteristics of WVD high time-frequency resolution, and effectively curbs both the modal aliasing phenomenon caused by EEMD decomposition and the cross-interference items caused by WVD distribution. The current study verifies the effectiveness and advantages of the proposed joint time-frequency analysis method, providing a novel technical approach for wheel polygon recognition and the evaluation of high-speed EMUs. 1 tab, 11 figs, 30 refs.