【作者】 查国峰；

【导师】 李迟生；

【作者基本信息】 南昌大学 ， 通信与信息系统， 2010， 硕士

【摘要】 公路车辆动态称重系统作为ITS(公路智能运输系统)的重要组成部分,是抑制公路车辆超载超限有效手段。目前,国内动态称重技术正处在快速发展阶段。本文从动态车辆称重系统中的几个关键部分着手,提出采用ATM-AVR单片机进行车辆称重信号的A/D转换和数据通信,拟定了数据通信协议并实现数据通信。设计了整体系统方案。针对现有传感器的使用情况,选择使用石英压电传感器作为系统的称重单元,并对数据采集调理电路做了相应的改进。利用Microsoft Visual C++ 6.0软件完成对车辆数据的录入和显示,包括对上位机串口设置,数据的存储和实时曲线显示。针对现有数据处理方法的优缺点,在分析车辆称重信号特点的基础上,利用小波多分辨率技术进行信号预处理,提出了使用Elman反馈型神经网络来确定传感器输出电压值与车辆静态重量的对应关系,从而更加准确的得到车辆的真实重量。最后,根据神经网络理论,比较了Elman网络内部各要素对数据分析性能的影响,以及反馈型神经网络和其他神经网络在车辆称重数据处理性能上的优劣。仿真结果表明,相对于BP和RBF网络,Elman网络在车辆动态称重系统中具有更优越的数据处理性能。更多还原

【Abstract】 As being the effective tool to forbid overloading, vehicles’weigh-in-motion (WIM) system is the important part of intelligent transport system(ITS). Nowadays, the WIM technology is being in the rapid developing period.In this article, something important as keys of the system were found such as ATM-AVR singlechip, A/D conversion and data communication. At the same time, the communications protocols were designed. Furthermore, we made some describing about the way using Studio software and designed the whole project. Based on the situation of sensor’s development and performance, we chose quartz piezoelectric sensors to be as the weight part of the WIM system, and made some improvements to the data acquisition circuit.We gained the vehicles’data by using the Microsoft Visual C++ 6.0 software and implemented the communication between upper computer and lower computer. And, at the same time, we stored the data in the document and drew relevant real time curve.Focusing on the merits and drawbacks of the being used methods of data processing and basing on the feature of the WIM signal, we completed data pretreatment by using wavelet multi-resolution techniques. Behind pretreatment, the corresponding relationship between the output of sensors and the static weight of vehicle was found by recurrent networks, which was first touched in this article. At the end of the article, aiming to show the advantages of recurrent networks in the data processing of WIM system, we made matlab stimulation and comparisons not only among different performance acting by different impact factors inside networks, but also between recurrent networks and other neural networks, such as BP network and RBF network.更多还原