【作者】 石晶晶；

【导师】 金炜东；

【作者基本信息】 西南交通大学 ， 检测技术与自动化装置， 2013， 硕士

【摘要】 随着我国高铁技术的飞速发展,列车车辆的速度也在不断提高,而在列车速度提高的同时也恶化了列车的动态环境,使得列车的轮轨作用力增加、各个部件的振动和蛇行运动加剧。同时随着列车服役时间的增加,列车零部件磨损加快导致其性能参数快速蜕变,严重影响了列车的运行品质。因此如何提取高速列车监测数据的有效特征,并快速准确的估计出高速列车安全服役性能,已成为目前高速列车安全预警领域亟待解决的问题。高速列车监测数据具有非线性和非平稳性,分数阶傅里叶变换(Fractional Fourier Transform, FRFT)是傅里叶变换的一种推广形式,能够同时表征信号的时域和频域信息,在非平稳信号处理领域得到广泛的应用。因此本文采用分数阶傅里叶变换提取信号的分数域特征,并运用分数域特征实现列车运行状态的评估,主要研究工作如下：分析了短时傅里叶变换、小波变换、魏格纳-威利分布(WVD)和Randon-Wigner分布等经典时频分析方法的优缺点,并深入研究了新时频分析方法分数阶傅里叶变换的基本原理和相关理论知识。提取高速列车监测数据的分数域特征,首先对高速列车监测数据进行分数阶傅坐叶变换,将数据变换到分数阶傅里叶空间,然后对变换后的三维数据进行侧面投影,得到不同分数阶下的信号峰值曲线,最后计算峰值曲线的统计特征。研究了高速列车转向架的结构和相关动力学分析。运用分数阶傅里叶变换提取高速列车监测数据特征,研究速度对仿真数据各个工况特征分布情况的影响。对所有通道单工况的仿真数据进行仿真,并运用支持向量机对四种单一工况进行分类,通过分析不同通道不同速度下四种工况的识别率情况,统计出对各个工况具有较好分类效果的通道。对试验监测数据进行仿真和分类统计,进一步验证通道和特征的有效性。通过比较两种单一工况及其混合工况在横向和垂向上的特征分布情况,来探讨了多故障工况与单一工况之间的关联。对参数渐变工况进行仿真,研究列车从原车正常变化到三种完全故障过程中的特征分布情况。综上所述,本文运用分数阶傅里叶变换对高速列车不同工况的数据进行仿真分析,实验结果证实了分数域特征对高速列车故障识别的可行性与有效性。更多还原

【Abstract】 With the rapid development of China high-speed train technology, the speed of the train vehicles are constantly improved, and the increase of train speed also worsen the train dynamic environments, increases the wheel-rail forces of train, the the vibration of various components and hunting oscilations. At the same time as the train service time increases, the train components wear out more rapidly resulting in the rapid disintegration of its performance parameters, seriously affected the quality of train running. How to extract effective features of the high-speed trains monitoring data, and quickly and accurately estimate the performance of high-speed train security service, has become a serious problem in the field of security early warning of high-speed trains.High-speed train monitoring data are nonlinear and non-stationary,fractional Fourier transform (Fractional Fourier Transform, FRFT) is an extended form of the Fourier transform, simultaneously characterize signals in time domain and frequency domain, widely used in the field of non-stationary signal processing field. This paper uses the fractional Fourier transform to extract signal fractional-domain features, and estimate the status of the train running by using fractional-domain features, the main research work are as follows:This paper analyze the advantages and disadvantages of the short-time Fourier transform, wavelet transform, Wigner-Willie distribution (WVD) and Randon-Wigner distribution and other classic time-frequency analysis, and study the basic principles and theoretical knowledge of the new time-frequency analysis method fractional Fourier transform.Extraction of the fractional-domain features of the high-speed train monitoring data, firstly transform high-speed train monitoring data to fractional Fourier space, then using the fractional Fourier space data to get lateral projection, which shows signal maximum peak in different fractional space, finally calculat the statistical features of the peak curve.Study the structure of the high-speed train bogie and related dynamics analysis. Using fractional Fourier transform to extract the features of high-speed train monitoring data, to study the distribution of different speed of simulation data each working condition. Simulate the data of all channels a single working condition and the use the support vector machines to classify working condition of four kinds of single. Compile the statistics of channels of a more good classification effect by analyzing the recognition rate of the four conditions of different channels of different speeds.Simulate and get the statistics of test data, further verify the effectiveness of the channel and features. By comparing the two kinds single working condition and the mixed condition on the lateral and vertical features distribution, exploring the association between multiple fault condition and single working condition. The simulation of parameters gradient conditions shows the features distribution of train working condition from the original train to the three complete failure processing.In summary, this paper uses the fractional Fourier transform to simulate high-speed train different running status data, the simulation and analysis result proves the feasibility and effectiveness of fractional domain features on high-speed train failure recognition.更多还原