【作者】 张杰；

【导师】 谢国；

【作者基本信息】 西安理工大学 ， 控制科学与工程， 2023， 硕士

【摘要】 铁路运输作为国家交通运输领域的重要组成部分,维护其运行的安全性及平稳性尤为重要。轨道不平顺激励是导致列车振动的根本原因,不仅会对列车的运行状态造成干扰,也会使列车的平稳度及舒适度大幅下降。因此实现对轨道不平顺状态的监测是国家铁路日常运维和智能化发展的重点研究方向。本文以轨道高低不平顺为研究对象,结合列车车辆动力学模型与轨检车真实数据,研究基于轴箱加速度的轨道高低不平顺估计方法。本文具体研究工作如下:(1)结合仿真环境下不同运行条件对列车垂向加速度的干扰特性,构建基于含噪加速度的高低不平顺检测模型。针对真实列车振动数据在复杂干扰环境下难以定量分析的问题,搭建以实测高低不平顺作为轨道激励的列车动力学仿真模型,设定不同列车运行条件并生成对应的列车垂向振动数据用于后续分析。由于列车在不同运行速度条件下对高低不平顺激扰呈不同响应特性,提出基于列车运行速度的列车垂向位移修正公式。接着研究不同类型不平顺及不同路线对垂向振动数据的干扰特性,以不同类型噪声模拟干扰特性并分析其解决方案。最后将噪声添加到加速度信号中,研究基于含噪加速度的惯性基准法步骤,对不同积分方法和去趋势项方法作对比实验。经实验分析,以时域积分方法及一阶拟合结合无限脉冲响应滤波器(Infinite Impulse Response Filter,IIR)去趋势项方法为基础的高低不平顺检测模型具备较强的适应性。(2)针对真实数据检测高低不平顺过程中存在的噪声干扰及检测缺失问题,在原有模型基础上进行改进。提出自适应噪声完备集合经验模态分解(Complete Ensemble Empirical Mode Decomposition with Adaptive Noise,CEEMDAN)结合粒子群算法(Particle Swarm Optimization,PSO)的去噪方法,完成对加速度有效分量的筛选,在不破坏有效信号的前提下,去除高低频噪声;针对真实数据采集过程中列车速度非恒速问题,提出信号重采样方法将空间域信号转为时域信号稳定采集信号频率,降低滤波误差;针对高速条件下中短波不平顺检测缺失问题,提出基于小波变换(Wavelet Transform,WT)的中高频信息提取方法,有效扩充高低不平顺检测波长覆盖范围。使用轨检车实测加速度数据做模型验证,结果表明不同方法模块均能有效提升不平顺检测精度,证明了所提方法的有效性和准确性。(3)针对单一传感器数据信噪比低问题,基于多源传感器数据融合背景搭建高低不平顺检测模型。由于含噪加速度数据缺少传统数据融合方法所需噪声先验信息,本文提出基于后验方差估计的数据融合方法。使用加速度功率谱密度分布估计方差并利用噪声后验特性修正方差,计算标准差作为信号影响因子做数据融合,将融合结果导入惯性基准法模型实现多传感器融合的高低不平顺检测。最后利用多组含噪加速度数据进行实验,与基础融合方法进行对比,验证了本文所提融合算法的有效性。更多还原

【Abstract】 As an important part of the national transportation field,it is particularly important to maintain the safety and stability of railway transportation.The excitation of track irregularity is the root cause of train vibration,which will not only interfere with the running state of the train,but also lead to a significant decrease in the smoothness and comfort of the train.Therefore,the monitoring of track irregularity is the key research direction of the daily operation and intelligent development of the national railway.In this paper,the track irregularity is taken as the research object,and the track irregularity estimation method based on axle box acceleration is studied by combining the dynamic model of train vehicle and the real data of track inspection vehicle.The specific research work of this paper is as follows:(1)Combined with the interference characteristics of train vertical acceleration under different operating conditions in the simulation environment,longitudinal irregularity detection model based on noisy acceleration is constructed.Aiming at the difficulty of quantitative analysis of real train vibration data in complex interference environment,a train dynamics simulation model with measured irregularity as track excitation is built.Different train operation conditions are set and corresponding train vertical vibration data are generated for subsequent analysis.Due to the different response characteristics of the train to the longitudinal irregularity excitation under different running speeds,the vertical displacement correction formula of the train based on speed is proposed.Then,the interference characteristics of different types of irregularity and different routes on vertical vibration data are studied,and the interference characteristics are simulated with different types of noise and the solutions are analyzed.Finally,the steps of inertial reference method based on noisy acceleration are studied,and comparative experiments are carried out on different integration methods and detrending methods.Through experimental analysis,the longitudinal irregularity detection model based on time domain integration method and first-order fitting combined with Infinite Impulse Response Filter(IIR)detrending method has strong adaptability.(2)Aiming at the problem of noise interference and detection loss in the process of detecting longitudinal irregularity of real data,the original model is improved.A denoising method of Complete Ensemble Empirical Mode Decomposition with Adaptive Noise(CEEMDAN)combined with Particle Swarm Optimization(PSO)is proposed to complete the screening of effective components of acceleration and remove high and low frequency noise without destroying the effective signal.Aiming at the problem of non-constant speed of train in the process of real data acquisition,a signal resampling method is proposed to convert the spatial domain signal into the time domain signal to stably collect the signal frequency and reduce the filtering error.Aiming at the problem of missing detection of medium and short wave irregularity under high speed conditions,a medium and high frequency information extraction method based on wavelet transform(WT)is proposed to effectively expand the coverage of longitudinal irregularity detection wavelengths.The model is verified by the measured acceleration data of the track inspection vehicle.The results show that different method modules can effectively improve the accuracy of irregularity detection,which proves the effectiveness and accuracy of the proposed method.(3)Aiming at the problem of low signal-to-noise ratio of single sensor data,longitudinal irregularity detection model is built based on multi-source sensor data fusion background.Because the noisy acceleration data lacks the noise prior information required by the traditional data fusion method,this paper proposes a data fusion method based on posterior variance estimation.The acceleration power spectral density distribution is used to estimate the variance and the noise posterior characteristic is used to correct the variance,which is used as the signal influence factor for data fusion,and further combined with the inertial reference method model to realize the longitudinal irregularity detection of multi-sensor fusion.Finally,multiple sets of noisy acceleration data are used for experiments,and compared with the basic fusion method,the effectiveness of the fusion algorithm proposed in this paper is verified.更多还原