【作者】 汪斌；

【导师】 过学迅；

【作者基本信息】 武汉理工大学 ， 车辆工程， 2010， 博士

【摘要】 近年来,中国的汽车工业发展迅速,新产品层出不穷,开发周期越来越快。同时,随着人民生活水平的提高,对汽车质量、安全和舒适性的要求也不断提高。利用道路模拟试验台,在试验室内对汽车零部件和整车进行道路模拟试验是加速新车型开发、提高产品质量的有效手段。而传统的道路模拟试验多是重现被试车辆在道路上行驶时的载荷或加速度,这就存在如下问题：①需要事先进行专门的数据测量,当开发新车型缺乏相应数据时便无法试验。②试验信号与被试车辆和道路密切相关,不同车辆之间难以进行对比试验。针对这些问题,本文研究了利用道路模拟试验台来再现车辆最直接的路面输入——路面不平度信号的方法和流程,其主要创新点有：1)基于MATLAB开发了路面不平度重构软件,集成了现有主要再现方法；2)建立了道路模拟试验台自回归滑动平均动态神经网络模型；3)利用迭代学习控制在道路模拟试验台上再现出了期望信号。研究了路面不平度的数学模型和描述方法,利用回归分析得到了不同描述方法间的相互关系。在综合分析现有路面不平度重构方法特点的基础上,开发出路面不平度重构软件,便于实际使用。四通道道路模拟试验台的组成很复杂,用一般的非线性模型难以描述。本文采用非线性自回归滑动平均动态神经网络模型来对系统进行辨识,利用试验数据建立的试验台非线性模型,其结果具有较高的精度。研究了传统的远程参数控制法的迭代方法,结合路面不平度再现的实际特点,将这一方法成功的用于再现路面不平度信号。详细论述了频率响应函数的主要辨识方法,相干函数的估计方法,迭代过程等,并应用于单输入单输出和多输入多输出系统再现路面不平度的试验,取得了较好的效果。讨论了此方法在试验中应注意的问题。在实现上述控制方法后,针对远程参数控制在再现路面不平度信号中存在的不足,本文提出,利用迭代学习控制的特点和对试验台的先验知识,在无需辨识频率响应函数的条件下,直接在时域内设计了基于平滑滤波器的迭代学习控制器,试验结果表明这一方法是有效和可行的。这种方法和前述方法相比,易于实现,计算速度快,迭代过程无需试验人员的干预。迭代学习律的设计是迭代学习控制的关键,为进一步研究设计迭代学习律,本文采用了仿真分析的方法。以已经建立的试验台非线性模型为控制对象,针对它设计了P型开环迭代学习律,但收敛速度慢,并不实用。为提高收敛速度,提出了基于离散PID控制器的开闭环迭代学习控制器,仿真结果表明所设计的控制器达到了预期的控制效果。更多还原

【Abstract】 In recent years, due to the rapid development of Chinese automobile industry, new products pile up one after another and development cycle becomes faster and faster. At the same time, with the people’s living standard growing, requirements for automobile quality, security and amenity are also continuously increasing. Road simulation test for components and vehicles in laboratory is an active means to accelerate new models development and to improve the quality of products. While in traditional road simulation test, the load or acceleration from a running vehicle is commonly recurred, thus two problems follow:①It needs precede special data measurement beforehand so that when lack of pertinent data at developing new models, the test is hardly to run.②The test signals have closely connections with the tested vehicle and the road, so it is difficult to compare the test results with different vehicles. Aiming at these issues, this text researches the methods and processes that are applied to reproducing the most direct road surface input of vehicle-road roughness. The main innovation points are:1) developing a software to reconstruct road roughness with MATLAB, which integrating the current main reconstruction methods; 2) building a nonlinear autoregressive moving average (NARMA) dynamic neural network model for road simulator; 3) by using iterative learning control (ILC), the desired signals are represented successfully in test rigs.This paper looks into the road roughness mathematical model and description method; utilizes regression analysis to find correlation between the different descriptions. Based on the comprehensive analysis of some existing typical methods’ features, the software, facilitating practical use to reconstruct road unevenness, is developed.For a 4-poster road simulator is a very complex system, it is difficult to describe it by the general nonlinear model. Thus the author presents a NARMA dynamic neural network model to identify the system. Because of this nonlinear test bed model which is built through test data, the results show high degree of precisionThis paper researches the iterative method of the traditional remote parameter control (RPC) and combining physical characteristics of road roughness, it is successfully used to reproduce this unevenness signal. Dissert on prime identification method of frequency response function, estimation method of coherence function and iteration process, etc. By applying these to single input-single output and multiple input-multiple output system to recur road roughness, the tests achieve preferable results. After realizing the above controlling means, against the insufficiency existing in recurring road unevenness using RPC, this article further presents a method that uses ILC and the prior knowledge of test bed directly devise a smooth filter based ILC controller in time domain, while without identifying system FRF. Compared with RPC, this technique is easy to implement, computing speed fast, and the test person need not intervene the iteration process. Devising iterative learning law is the key to ILC, for further study on it, this article adopts simulation technique. With having been built non-linear model as the controlled object, a P-type open-loop iterative learning law has been designed, however, its convergence rate is too slow to use in practical application. To increase the convergence rate, a discrete PID controller based open-closed loop ILC controller has been presented and the simulation proves the designed controller comes to the expected control purpose.更多还原