基于小波神经网络的车辆构架人工蛇形波重构技术的研究

【作者】 周峰；

【导师】 危韧勇； 李志勇；

【作者基本信息】 中南大学 ， 控制科学与工程， 2008， 硕士

【摘要】 在我国铁路行业,随着高速列车的出现以及铁路的进一步提速,列车脱轨事件也呈上升趋势。脱轨造成人民生命和财产的巨大损失,给铁路安全运输造成了极大的威胁。为了得到列车的安全行驶速度,测量和重构作为列车—桥梁(轨道)系统激励源的车辆构架实测横向振动波(俗称蛇行波)具有重要的理论和工程实际意义。预测车桥系统的振动响应,关键在于要求得与实际构架实测蛇行波接近的构架人工蛇行波,基于Monte-Carlo的人工蛇行波随机模拟方法只保留了实测数据中的方差作为重构的唯一约束条件,而其他一些重要特征参数,如频率、概率等都没有得到充分的利用,造成了重构过程中的频率和相位的机会平均,导致了最后重构的蛇行波与实测蛇行波有一定的差距。本文针对小波良好的时频局部性及神经网络强大的非线性映射能力,用小波基代替了神经网络中的Sigmoid函数,构造了带有轮盘赌遗传选择机制的小波神经网络,并对160公里/小时广深铁路实测蛇行波数据进行了分析、重构,仿真结果表明这种方法能够有效地保留实测蛇行波的特征参数。与传统基于Monte-Carlo方法的三角级数随机重构方法相比,基于小波神经网络的人工蛇行波重构方法能够克服重构过程中的频率、相位机会平均,波形可能会出现突变等缺点,经过重构所得到的波形中带有更多实测蛇行波的信息,过渡、衔接地更加自然。该方法也适用于行驶速度高于160公里/小时的高速列车。更多还原

【Abstract】 In the railroad industry of our country, the appearance of high-speed train and railway system’s speed level increasing induce the derail event increasing. Derailed brings huge losing to human being and society property which menace the safety of railway transportation. In order to get the safety speed of the train, measure and rebuild the train crawl wave which is regard as the actuator of Vehicle-Bridge System is the most important.Firstly, the paper summarize the Vehicle-Bridge System、the factor which results system’s vibration and the effect of system’s actuator to train’s safety in briefly, get the result that the key to predict the response of the Vehicle-Bridge System is to get the artificial crawl wave which is very close to the real crawl wave by measured. Then particular introduce the Monte-Carlo method which is the main means to rebuild artificial crawl wave nowadays. In method research, the random simulation of artificial crawl wave based on Monte-Carlo method only used variance while neglect many other useful information such as frequency and probability, it make chance average of frequency and phase, lead the result is not close the real crawl wave. This paper put out a new method to rebuild the crawl wave, construct the wavelet neural network which contain the roulette wheel select mechanism , process the data which is measured on GuangShen railway at speed 160km/h by wavelet neural networks, utilize the amplitude, frequency and probability adequately, and rebuild the crawl wave by computer. The result show this method can rebuild the crawl wave well.Compare to the Monte-Carlo method, the method based on wavelet neural network conquer the flaws such as chance average and wave saltation, the wave rebuild by wavelet neural network can be close to the real measured crawl wave. Meanwhile this method can also apply in rebuild high-speed train’s crawl wave.更多还原