【作者】 侯正清；

【导师】 王祁；

【作者基本信息】 哈尔滨工业大学 ， 仪器科学与技术， 2009， 硕士

【摘要】 公路作为一种人造结构物,其承受重复荷载的能力和次数(使用寿命)是有一定限量的。随着我国经济的发展,和现代化高速运输工具的兴起,现代公路路面也不断呈现新的状况,面对这些新情况,特别是重载超限车辆对公路路面的破坏性影响,旧的道路设计、管理规范需要修正和更新。并且,交通荷载基础数据作为战略储备需求的反映,只有具备了长期而准确的测量信息,决策者才能够以此为依据做出符合实际情况的决策,所以相关的路面信息测试和研究更显重要。而此研究的重点,就在于对路面动态车辆信息的测量。本文研究了一种以类神经网络机器算法支持向量机的扩展应用——最小二乘支持向量机为算法基础,DSP(TMS320F2812)和简单应变式传感器为硬件支持的路面行驶车辆动态信息测量方法。论文首先介绍了动态车辆信息测量的意义,技术重点和研究现状,描述了基于应变传感器的测试系统的工作原理。接着介绍了用以进行车辆分类的支持向量分类机和用以进行车辆动态称重的回归机算法的原理。详细研究了针对该测试系统的基于知识和经验的数据特征提取方法,建立了包含参数预测,验证,传感器融合定量分析等算法的车辆分类识别和回归称重的软件系统。最后在DSP(TMS320F2812)平台上实现了算法系统,结合相关数据进行了相关实验,论证结果。更多还原

【Abstract】 As man-made structures, Highway’s load enduring capacity and durability (useful life) are certainly finite. With China’s economic development, and the construction of modern transport, nowadays, more and more new situations of road surface state has been presented. In the face of these new situations, especially the devastating effects on road surface brought by overrunning heavy-duty vehicles, old road design and management need to be amended and the norms need to be updated. In addition, the traffic load based on data is a reflection of the demand for strategic reserves, so only with a long-term scaled, accurate data and information being evidence, decision-makers will be able to give out the correct decision-making according with the actual situation, then road information tests and researches look more important. And the focus of this study lies on the information measurement for dynamic vehicle on road surface.In this paper, a DSP (TMS320F2812) and simple strain sensors hardware supporting information measurement method for dynamic vehicle which utilizes the Least Squares Support Vector Machine algorithm that it is an extension of Support Vector Machine being similar with neural network was researched. At first, the paper introduces the significance, technology focus and research actuality of information measurement for dynamic vehicle, describes the working principle of testing system based on strain sensors. Then the paper introduces the elements of vehicles classification using Support Vector Classification algorithm and dynamic weighing system which uses Support Vector Regression algorithm. After that, the paper studies in detail for the data feature extraction methods based on human knowledge and experience, and constructs a software system for vehicle classification and regressive dynamic weighing which contains parameters establishing prediction, verification, quantitative analysis by sensors fusion. Finally, we use DSP (TMS320F2812) platform to achieve the algorithm system, run experiments with the related data, and verify the results.更多还原