【作者】 孙从玺；

【导师】 潘洪达；

【作者基本信息】 吉林大学 ， 载运工具运用工程， 2004， 硕士

【摘要】 汽车已成为当今人们生产和生活中重要的交通工具，在使用过程中各总成、部件故障不仅会影响和困扰运输生产过程，造成巨大的损失，而且会带来一系列安全性和经济性等潜在问题，严重时甚至产生不可预料的后果。因此，对总成部件尤其是关键部件故障进行早期预测、监测及诊断是保证系统安全运行的重要环节，也是近年来人们关注并致力研究的问题。然而由于汽车结构和装备的复杂化，要及时、准确地对各个设备的状态、可能的故障以及已有的故障进行监测、预测和诊断分析，并迅速地做出避免和排除故障的措施，一直是一个难题。人工神经网络技术为解决这些问题提供了一条有效的途径。人工神经网络是在现代神经学基础上提出的，通过模拟大脑神经网络处理、记忆信息的方式而发展起来的自适应动力系统。五十多年来，人们对人工神经网络进行了大量的研究，提出了近四十种神经网络模型，如前向网络、反馈网络、随机网络、自组织神经网络等。人工神经网络以其大规模的并行计算能力、自适应性和容错性，在工业过程中解决诸如建模、智能控制、预测、分类、模式识别、组合优化、数据和信号处理等方面发挥了重要作用，并迅速向各个应用领域渗透。近年来，人工神经网络技术在汽车故障预测、监测和诊断领域的应用研究日趋活跃。人工神经网络与人工智能的另一重要分支—专家系统相结合，在汽车故障监测和诊断领域具有广阔的应用前景。由于汽车复杂程度的增加，为了保证汽车的各种性能和使用寿 82<WP=88>摘 要命，对汽车所要检测的项目也随之增多，仅靠一种检测设备检测一个项目已经不能满足汽车在线检测的要求。在这种情况下，研究和开发多功能的汽车检测诊断设备具有十分重要的意义。本论文对多功能汽车检测设备以及神经网络及其在汽车轮、轴松旷间隙及车轮定位参数检测中的应用进行了研究。在理论上，分析了汽车轮、轴松旷间隙产生和检测的机理、汽车制动性能理论、平板式制动性能试验台的检测机理和车轮侧滑产生的机理和检测方法。提出了涉及到轮、轴松旷间隙检测、制动性能检测、侧滑以及称重等多种功能的检测设备，该设备在构思方案、具体结构和技术上都有新的思路。在应用上重点研究了神经元网络在轮、轴松旷间隙识别和车轮外倾角和前束角的检测中的应用。论文研究的主要内容安排为：一、回顾了汽车底盘检测的国内外发展状况，概括介绍了汽车底盘检测的意义、基本内容以及神经网络的定义和基本类型。论述了神经网络在汽车故障监测和诊断中的应用和最新进展状况。二、研究了汽车轮、轴松旷间隙产生和检测的机理、汽车制动性能理论、平板式制动性能试验台的检测机理和车轮侧滑产生的机理和检测方法。提出了包括轮、轴松旷间隙检测、制动性能检测、侧滑以及称重等多种功能的检测设备，给出其结构形式。三、研究了多功能检测设备的测控系统。它分为两个部分，即信号采集系统和控制系统。在信号采集系统中包括传感器的选择、放大电路的设计和 A/D 转换卡的选择；在控制系统的研究中，详细叙述了控制方案的具体内容。接着介绍了电路的抗干扰设计，包括直流电源回路的抗干扰措施；信息检测中的抗干扰设计和信息传输中的抗干扰措施等。四、研究了神经元网络、模式识别基本理论和方法。在此基础上讨论了自组织竞争神经元网络。结合检测特点将模糊聚类算法和神经元网络竞争学习算法相结合，将模糊等价关系和类距离阈限引入自组织神经网络模型中，提出了模糊自组织神经元网络模型和 BP网络模型并将其应用于汽车底盘轮、轴松旷间隙识别和车轮前束角 83<WP=89>吉林大学硕士研究生毕业论文与外倾角检测中。五、介绍了整个实验的流程。实验结果表明，多功能汽车检测设备能够很好地应用于检测现场，缩短了检测时间，降低了检测费用。模糊自组织神经元网络很好地解决了间隙识别的问题，为维修人员快速诊断故障起到很好的指导作用。本论文通过探寻检测的新方法、新技术，在完善相应检测理论模型的基础上，研究开发了多功能汽车底盘综合性能检测设备。并将神经元网络技术应用在汽车轮、轴松旷间隙识别和车轮外倾角和前束角的检测中，从理论上和应用上丰富和发展了汽车检测方法，建立起了较为完备的理论体系，为神经网络在汽车故障监测和诊断这一领域中的实际应用提供了基础。更多还原

【Abstract】 Automobile has become the important traffic tool in people’severyday life. In the course of using, every fault of every part of the carcannot only affect transportation production, cause huge loss, but canbring a series of latent problems such as safety and economy, seriously,they can even produce the consequence that cannot expect. Therefore,it’s very important to foresee and diagnose the fault early, which canguarantee the system operate safely, and it is also the problem whichpeople have paid attention to and worked for in recent years. Butautomobile is a complex system, it is hard to monitor the condition,foresee and diagnose the fault of every equipment and put forward goodmeasures to avoid and get rid off the faults in time and accurately. It’salways a baffling problem. Artificial neural network technology hasoffered effective method to solve these problems. Artificial neural network is put forward on basis of modernneurology studies. It’s a self-adapting dynamical system based onsimulating human’s brain. In recent more than 50 years, people havecarried out plenty of researches on artificial neural network, and putforward about 40 kinds of neural network models, such as feedforwardnetwork, feedback network and random network, self-organizing neuralnetwork. Artificial neural network has the ability of large-scale parallelcalculation, self-adapting and fault tolerance, so it has played important 85<WP=91>吉林大学硕士研究生毕业论文role in solving problems such as modeling, intelligent control, forecast,classification, pattern recognition, combination optimization, data andsignal processing, and now widely used in every field. In recent years,the artificial neural network technical has applied more frequently in thefield of inspection, diagnosis and forecast of vehicle failure. Thecombination of artificial neural network and the other important branchof artificial intelligence, expert system, has wide application prospect inthe field of diagnosis and inspection of vehicle failure. Because of the increasing of automobile complexity, you mustdetect plenty of items to guarantee the service life and variousperformances of automobile. The equipment with only one functioncannot satisfy automobile online detection. On this condition, it hasgreat significance to research and develop the automobile detectionequipment with multifunction. This paper studied for multifunctional automobile detectionequipment as well as artificial neural network and its application inautomobile chassis performance detection. In theory, we analyzed themechanism that conjugate of wheel and Kingpin produce and how todetect, the theory of automobile brake performance and the testingmethod on plain plate and the mechanism of wheel sideslips. On basis ofthe theory above, we have put forward the multifunctional detectionequipment with functions such as conjugate detection, brakeperformance detection, sideslip as well as weight. On application wehave studied fuzzy self-organizing neural network based on competitivestudy algorithm. The major content of the paper as follows: 1．Has reviewed domestic and international development conditionof automobile chassis detection, generally introduced basic type and thedefinition of neural network, basic content of automobile chassisdetection. Has discussed neural network newest progress condition andthe application in the diagnosis and inspection of vehicle failure, 86<WP=92>Abstractanalyzed some basically theoretical problems that this field has notsolved. 2．Introduced the mechanism of conjugate of wheel and Kingpin anddetection method, the theory of automobile brake performance, the brakeperformance testing method on plain plate and wheel sideslip theory.With the theory put forward the multifunctional detection equipment thatincludes the various functions such as suspension cleara更多还原