一种基于小波神经网络的车辆构架人工蛇行波重构方法研究  

Research of Rebuilding Artifical Crawl Waves of Vehicle Frames by Wavelet Neural Networks

作　　者：周峰[1] 危韧勇[1] 李志勇[1] 

机构地区：[1]中南大学信息科学与工程学院,湖南长沙410075

出　　处：《铁道学报》2009年第3期50-53,共4页Journal of the China Railway Society

基　　金：国家自然科学基金项目(50405034);湖南省自然科学基金项目(03JJY3094)

摘　　要：预测车桥系统的振动响应,减少工作人员去轨道实测列车蛇行波的工作量,关键在于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,造成重构过程中的频率和相位的机会均等,导致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,用小波基代替神经网络中的Sigmoid函数,构造带有轮盘赌遗传选择机制的小波神经网络,并对列车运行速度为160km/h的广深铁路实测蛇行波数据进行分析、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,重构的蛇行波过渡、衔接更加自然。同时,该方法也适用于高速列车的蛇行波重构。The key to predict the vibration responses of the vehicle-bridge system is to get the artificial crawl waves which are very close to the real measured crawl waves. The random simulation of artificial crawl waves based on the Monte-Carlo method maintains the measured variances as the sole constraint condition for regener- ation neglects some other important characteristic parameters such as the frequency and probability, thus creating chance average of frequencies and phases in the process of regeneration and leading to some degree of difference between the finally regenerated crawl waves and real crawl waves. This paper puts forward a new method to rebuild the crawl waves. By the new method, the wavelet neural network containing the roulette wheel selection mechanism is constructed, the data measured on the Guang-Shen Railway at the speed of 160km/h are processed by wavelet neural networks, the parameters of the amplitude, frequency and probability are utilized adequately, and the crawl waves are rebuilt by computer. The results show this method can rebuild the crawl waves well. Meanwhile this method proves applicable in rebuilding crawl waves for high-speed trains.

关 键 词：列车 蛇行波 小波神经网络 轮盘赌 

分 类 号：U270[机械工程—车辆工程]