基于蚁群优化模糊神经网络的金融押运车故障预警问题的研究

发布时间：2018-03-11 07:04

本文选题：故障预警　切入点：模糊系统　出处：《东北大学》2012年硕士论文　论文类型：学位论文

【摘要】：近年来我国金融行业稳定发展,越来越多的人也意识到金融安防工作的重要性。因此逐渐衍生出了一些专业的金融押运公司,并发展成为一种新兴的行业。专业押运公司所采用的押运车不同于普通车辆,其安全性与可靠性要求相对较高,并且由于其自身不易拆解等特点出现故障不能去普通4S店进行维修,通常要返厂维修。如果一部押运车因本身故障返厂维修,那么较长的维修周期可能导致整个押运调度的调整,这无论从经济方面还是安全性角度考虑都是不利的。发动机是押运车的核心部位,也是押运车中故障频发的设备系统。随着汽车技术的进步,押运车电子化程度越来越高,发动机电控系统的功能不断强大,同时也带来故障节点数的节节攀升,传统的故障诊断系统与人工经验的方法已经不能满足押运车快速准确的诊断要求。此外其只能确定故障的存在与否,而对故障的趋势判定无能为力,已难以满足押运车的安全的需要。因此,有必要对传统的故障诊断系统加以改进,以适应押运车的要求。论文的主要研究成果包括： (1)从押运车目前存在的实际问题出发,综述了国内外现有的主流的故障诊断设备及方法； (2)根据押运车发动机的故障信号、故障原因、状态信号之间的关系具有模糊性、非线性,以及传感器采集数据具有周期性,并不能完全做到“实时”的问题。提出将状态信息模糊化的方法,即将模糊系统和BP神经网络串联结合起来,组成模糊神经网络的故障预警方法； (3)针对押运车故障预警快速准确的要求,详细设计了BP网络拓扑结构,利用蚁群算法的全局搜索性优化网络连接权值,解决了训练过程易陷入局部极小点的问题,并提高了收敛速度； (4)针对押运车电控系统复杂,参数多的问题,提出将整个系统切分的办法。然后以切分后的电子燃油喷射控制系统为例完成了基于蚁群算法的模糊神经网络结构设计。主要包括：故障征兆信息的采集与模糊化处理,BP网络结构与参数的优化设计。并实现了此网络在押运车故障预警中的应用。论文的研究结果可以为押运车的智能故障预警系统的进一步开发提供依据。论文所得模型和算法对其他优化系统的建设具有一定的参考价值。
[Abstract]:In recent years, with the steady development of the financial industry in our country, more and more people are also aware of the importance of financial security work. As a result, some professional financial escort companies have gradually emerged. And it has developed into a new industry. The transportation vehicles used by professional escort companies are different from ordinary vehicles, and their safety and reliability requirements are relatively high. And because it is not easy to disassemble and other characteristics such as failure can not go to the ordinary 4S shop for maintenance, usually return to the factory maintenance. If a transport vehicle due to its own failure to return to the factory maintenance, So the long maintenance period may lead to the adjustment of the entire escort scheduling, which is unfavorable both from the economic aspect and the security point of view. The engine is the core part of the truck, and it is also the equipment system with frequent faults. With the development of the automobile technology, the electronic degree of the escort truck is becoming higher and higher, and the function of the engine electronic control system is constantly powerful. At the same time, the number of fault nodes has been rising. The traditional fault diagnosis system and manual experience can not meet the requirements of rapid and accurate diagnosis of the transport vehicle. In addition, it can only determine whether the fault exists or not. Therefore, it is necessary to improve the traditional fault diagnosis system to meet the requirements of the escort vehicle. The main research results include:. 1) based on the practical problems existing in the transport vehicle at present, the existing mainstream fault diagnosis equipment and methods at home and abroad are summarized. (2) according to the fuzziness and nonlinearity of the relationship between the fault signals, the causes and the state signals of the engine, and the periodicity of the data collected by the sensor, The method of fuzzifying state information is put forward, which combines the fuzzy system and BP neural network in series to form the fault early warning method of fuzzy neural network. 3) aiming at the requirement of fast and accurate fault warning for escort vehicle, the topology structure of BP network is designed in detail, and the network connection weight is optimized by using the global search of ant colony algorithm, which solves the problem that the training process is easy to fall into local minima. The convergence rate is improved. 4) aiming at the problem of complex electric control system and many parameters of the transport vehicle, The method of dividing the whole system is put forward. Then taking the electronic fuel injection control system as an example, the design of fuzzy neural network structure based on ant colony algorithm is completed, including: fault symptom information collection and fuzziness. The optimization design of BP network structure and parameters is realized, and the application of this network in the early warning of the vehicle faults is realized. The results of this paper can provide the basis for the further development of intelligent fault early warning system. The models and algorithms obtained in this paper have some reference value for the construction of other optimization systems.
【学位授予单位】：东北大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP18;F830.91

【参考文献】