基于超复数边际谱时频响应函数的变载条件下钢轨弱损伤识别研究

项目名称： 基于超复数边际谱时频响应函数的变载条件下钢轨弱损伤识别研究

项目编号： No.51305064

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 机械、仪表工业

项目作者： 任同群

作者单位： 大连理工大学

项目金额： 25万元

中文摘要： 高铁钢轨疲劳损伤是影响行车安全的重要因素。而钢轨所受激励复杂多变、内部存在温度力、工况恶劣、损伤微弱等实际问题增加了基于振动分析的钢轨损伤识别难度。对此，本项目提出基于超复数边际谱时频响应函数的钢轨薄弱环节损伤识别方法。跨薄弱环节设置两个三轴加速度传感器，将三通道响应作为矢量信号进行整体处理。实现矢量振动三维EMD分解，建立超复数HHT边际谱；由HHT边际谱求解两测点间的钢轨超复数时频响应函数，构造吸引子轨迹矩阵作为损伤特征矩阵；将损伤识别转化为彩色图像模式识别问题，以目标图像投影系数向量与模板图像奇异值向量之间的欧氏距离为损伤定量识别指标；探明钢轨温度力对时频响应函数的影响规律，认为时频响应函数仅在同温度力段内一致，确定适当的温度力段划分原则，并建立对应的基准时频响应函数样本库。本项目的研究成果有望形成新的钢轨弱损伤识别方法并推动其在工程实际中应用，具有重要的学术参考价值和实际应用价值。

中文关键词： 状态特征；四元数传递率；四元数小波变换；状态识别；轴向力

英文摘要： The fatigue damage of HSR rail is an important influencing factor for train running safety.The identification difficulty is increased for reasons of Variable load, longitudinal temperature force,harsh working condition and weak damage. Aiming at these problem, a damage identification method for weak link of HSR rail based on hypercomplex marginal spectrum Time-frequency response function(TFRF) is proposed. Two three-axis accelerometers are placed bestrideing the weak link of the rail. The response signals of three channels are considered as vector signal and processed integrally. Three dimensional EMD is performed and the HHT marginal spectrum is set up subsequently. Then the TFRF between the two measuring points is obtained and based on which the Hankel matrix is built as the feature matrix of damage. As a result, the damage identification is treated as color image pattern recognition. The euclidean distance between projection coefficient vector of object image and singular value vector of template image is selected as the quantitative index for damage identification. The influence law of TFRF affected by longitudinal temperature force will be ascertained. Only the Time-frequency response functions lying in the same segment of longitudinal temperature force is treated consistent.As a result, the division princi

英文关键词： state characteristic；quaternion-based three-channel joint transmissibil；quaternion wavelet transform；state recognition；longitudinal force