【作者】 郝茹茹；

【导师】 赵祥模；

【作者基本信息】 长安大学 ， 交通信息工程及控制， 2013， 博士

【摘要】 为了提高汽车主动安全性能，汽车上普遍安装防抱死制动系统（Anti-lock BrakingSystem, ABS），它能够通过控制和调整车轮制动压力，防止制动过程中因车轮抱死引起的制动跑偏、后轴侧滑以及失去转向能力等情况的出现，使车辆能够最大限度地利用地面制动力而减速停车，从而有效改善汽车的制动性能，提高汽车安全性。目前，对于汽车ABS整车工作状况的检测和评价，主要采用道路试验法。但是，道路试验场地占地面积大、造价高、试验准备和试验过程耗时长、危险性大、试验易受环境影响、重复性差。因此，道路试验适合于某种车型的ABS配型试验或部分汽车ABS的抽检，不适合大量汽车及在用汽车的定期检测。针对以上问题，本文提出了一种整车ABS室内试验台检测新方法，该方法通过在试验台上模拟道路试验中的汽车运动惯量及不同路面附着系数，实时采集制动时车轮及车身速度，并计算与ABS性能相关的技术参数，然后利用基于主成分分析及神经网络的判定算法对检测数据进行分析，最终实现整车ABS检测及检测结果的自动判定。该试验台检测方法与道路试验相比，具有占地面积小、成本低、检测速度快、安全性高、重复性好、检测过程不受环境影响等优点。论文主要在以下几个方面展开研究工作：（1）提出了一种汽车ABS室内试验台检测方法，该方法通过滚筒模拟连续移动的路面，并利用飞轮的转动惯量模拟车辆在道路上高速制动时的平动惯量，同时采用扭矩控制器在滚筒上加载与车辆“行驶”方向相反的力矩来模拟车辆行驶阻力，通过独立改变每个扭矩控制器加载扭矩的大小实现不同路面附着系数的模拟。论文在研究扭矩控制器的结构及工作原理的基础上，建立了基于扭矩控制器的多路面附着系数模拟的数学模型，并根据车辆在道路上及试验台上制动时的惯量关系，确定试验台上需要利用飞轮模拟的惯量。（2）进行了ABS试验台检测方法计算机仿真研究。首先根据ABS整车试验台检测理论，建立了试验台上制动时的整车及车轮模型、制动系统模型、滑移率—附着系数模型、试验台制动力模型等，然后由各子模型之间的关系建立试验台上制动时的单轮车辆动力学Matlab/Simulink仿真模型，最后进行各种路面工况下的仿真实验，验证了试验台检测理论的正确性和可行性。（3）提出了ABS试验台及测控系统的技术方案和整体结构。该试验台集轴重测量、常规制动性能检测、ABS性能检测、速度表校验于一体，其中ABS性能检测可前后轴同时测量，并且可根据车辆轴距信息自动调整前后台架滚筒组的中心距，同时为四个车轮提供不同的路面附着系数。提出了基于CAN总线的ABS试验台测控系统结构，并制定测控系统上位机与下位机智能模块的通信协议，同时按照ABS自动检测的要求，设计合理的检测流程，开发测控系统上位机应用软件，以完成整个系统的控制、提供简单清楚的人机交互界面。（4）进行了ABS台架检测与道路试验对比分析。用多个车型对本文提出的ABS试验台检测方法进行实车试验，主要包括单一高附着系数路面、单一低附着系数路面、对开路面以及对接路面试验。完成了同一制动条件下、相同车辆的ABS道路试验，并对台架试验与道路试验结果进行对比分析，结果表明，相同制动工况下的台架试验与道路试验结果符合性较好，说明本文提出的ABS试验台检测方法能够正确模拟道路试验工况，完成汽车ABS性能检测。（5）提出了基于主成分分析与BP神经网络的ABS检测结果自动判定方法。首先利用主成分分析对检测结果中的多个技术参数进行处理，去除数据之间的相关性及冗余信息，同时降低特征向量维数。然后建立了基于BP神经网络的分类器模型，确定了网络结构、输入层与输出层神经元个数、隐含层神经元个数等。最后设计了基于主成分分析与BP神经网络分类器的ABS检测结果判定算法，将经过主成分分析后的具有综合信息的少数主成分输入神经网络分类器完成检测结果判定。本文的研究属于汽车检测技术领域的前沿课题，该研究成果可以有效地解决我国现有的汽车检测试验台无法对整车ABS性能进行检测的问题，对于推动我国汽车检测技术及设备的发展具有重要意义。更多还原

【Abstract】 In order to improve the active safety of automobile, Anti-lock Braking System (ABS) isgenerally installed. It can prevent the automobile from deviation brake, rear axle sideslip andsteering failure caused for wheels lock by controlling and adjusting the wheel brake pressure.And enable the automobile to make the best use of the ground braking force to slow down andstop, therefore, the braking performance and active safety of automobiles are effectivelyimproved. Currently, the main approach to detect the performance of auto ABS is the roadexperiment. However, the disadvantages of large area occupation, high construction cost of thetest site, long test time, dangerous, susceptible to environment influences and poorreproducibility make it difficult to implement. So, the road experiment is only suitable for ABStype-match experiment of special automobiles or ABS sampling inspection and not suitablefor the periodic testing of a large number of automobiles.To solve the problems above, a novel ABS indoor bench detection approach is proposedin this paper. Translational inertia of automobile and different adhesion coefficients aresimulated on the bench. The real-time speeds of the wheels and the automobile are collectedduring braking process, and technical parameters that can reflect the ABS performance arecalculated. Then, data analysis is performed based on Principal Component Analysis (PCA)and neural network, and ultimately the realization of entire automobile ABS detection andautomatic determination of the ABS detection results are achieved. Compared with the roadexperiment, the ABS bench detection approach has the advantages of small area occupation,low cost, high efficiency, high safety, good repeatability and insusceptible to environmentinfluences, etc.The main research contents are described as follows:(1) An indoor bench detection approach for auto ABS is proposed. The rolling drums areused to simulate the continuously moving road. Translational inertia of the automobilebraking on the road is simulated by the moment of inertia of flywheel on the bench. Differentadhesion coefficients of road surface are dynamically simulated through loading different torques to the drums by the torque controllers. The mathematical model of different adhesioncoefficient simulation is built based on study of the structure and working principles of thetorque controller. The value of flywheel moment of inertia needed on the test bench isdetermined according to the translational inertia of braking on the road.(2) Computer simulation for ABS bench detection approach is studied. According to theABS bench detection theory, the vehicle model, wheel model, brake system model, slip ratio-adhesion coefficient model, braking force model, ect. on the test bench are established. AndMatlab/Simulink simulation model of a quarter-vehicle braking on the test bench is builtaccording to the relationship of each sub-model. Simulations under a variety of simulatedroad conditions are conducted in order to verify the correctness and feasibility of the benchdetection theory.(3) The technical solutions and the overall structure of the ABS test bench andmeasurement&control system are studied. The test bench has the functions of axle loadmeasurement, normal brake performance test, ABS performance test and speedometer test.The front and rear axle of an automobile can be detected at the same time for the ABSperformance, and the drum center distance between the the front and the rear bench can beautomatically adjusted according to the automobile wheelbase information. The system canoffer different road adhesion coefficient for each wheel at the same time. The measurementand control system is proposed based on CAN bus structure, and the communication protocolamong host computer and the intelligent modules is designed and realized. Proper detectionprocess and application software are designed and realized according to the ABS auto-testrequirement.(4) Comparative analysis between ABS bench test and road experiment is performed.Experiments with several types of automobile are conducted on the ABS detection benchproposed in this paper, including the experiments on single high adhesion coefficient road,single low adhesion coefficient road, split-μ surface road and connection-μ surface road. ABSroad experiments with the same automobile under the same braking conditions are performed,and the results are compared and analyzed with those from bench test. The results show that the bench tests achieve good compliance with road experiments under the same brakingconditions. And the proposed ABS bench detection approach can correctly simulate the roadexperiment conditions and realize the entire automobile ABS performance detection.(5) A judgment method for ABS test results based on principal component analysis andBP neural network is proposed. Firstly, the principal component analysis is adopted to processthe technical parameters of the test results to remove the correlation and redundantinformation among them, and the dimension of feature vectors is reduced at the same time.Then a classifier model based on BP neural network is built and the structure of network,neuron number of the input layers, output layers and hidden layer are determined. At last,ABS test results judgment algorithm based on principal component analysis and BP neuralnetwork classifier is put forward, and a small number of principal components withcomprehensive information is input into neural network classifier to complete the test resultsjudgment.The research of this paper is a frontier research topic in automobile performance testfield. The research results can effectively solve the problem of testing the performance of autoABS by the existing automobile test platform and have a great significance for promoting thedevelopment of automobile detection technology and equipment of China.更多还原