改进BP算法和模糊逻辑在双轴转台故障诊断中的应用

发布时间：2018-04-10 19:56

本文选题：转台 + 故障诊断　；参考：《哈尔滨工业大学》2011年硕士论文

【摘要】：转台是一种复杂的集机、电于一体的惯导测试与仿真设备,在航天、航空和航海等领域中发挥着巨大的作用。为了使转台能够长期稳定可靠地工作,对其开展故障诊断技术的研究非常的有必要。本文针对转台故障与征兆之间存在着复杂的非线性映射关系,利用模糊神经网络强大的模糊信息处理能力和自学习能力,提出了一种基于模糊神经网络的双轴转台故障诊断系统,并实现了诊断系统软件的编制。首先,通过对转台系统结构的深入分析,将转台常见的故障分为五类:机械故障、控制器故障、执行器故障、测角系统故障和通讯故障。根据故障底事件与特征信号之间的关系,建立了转台故障的决策表。为了消除信号中高频噪声,设计了一个二阶低通滤波器,同时还运用了小波分析去噪理论,并对阈值函数进行了改进,取得了一定的效果。其次,系统地介绍了诊断系统的模糊神经网络模型和其所用的BP算法,针对BP算法的不足,把遗传算法、引入惯性项和动态调整学习率三种方法相结合,对普通的BP算法进行优化改进,并将经过改进的BP算法用于转台的故障诊断中。采用一种抗噪声训练方法训练网络,使诊断系统可以克服一定幅值噪声的干扰。通过和标准网络进行对比,引入改进BP算法的模糊神经网络,具有较好的全局逼近能力,能够准确地实现故障的识别。最后,用Delphi和Matlab混合编程技术完成了诊断系统的编制。诊断系统分为神经网络训练、小波去噪、数据库管理和故障诊断四个功能模块,具备实时监测和诊断的能力。
[Abstract]:Turntable is a kind of complex inertial navigation test and simulation equipment, which is integrated with electricity and plays a great role in aerospace, aviation, navigation and other fields.In order to make the turntable work stably and reliably for a long time, it is necessary to study the fault diagnosis technology.Aiming at the complex nonlinear mapping relationship between fault and symptom of turntable, this paper makes use of the powerful fuzzy information processing ability and self-learning ability of fuzzy neural network.A fault diagnosis system for two axis turntable based on fuzzy neural network is proposed, and the software of diagnosis system is developed.Firstly, the common faults of turntable are divided into five categories: mechanical fault, controller fault, actuator fault, angle measuring system fault and communication fault.According to the relationship between the fault bottom event and the characteristic signal, the fault decision table of the turntable is established.In order to eliminate the high frequency noise in the signal, a second-order low-pass filter is designed. At the same time, the wavelet analysis theory is used to remove the noise, and the threshold function is improved, and a certain effect is obtained.Secondly, the fuzzy neural network model of the diagnosis system and the BP algorithm are introduced systematically. Aiming at the shortcomings of the BP algorithm, the genetic algorithm, the introduction of inertia term and the dynamic adjustment of learning rate are combined.The common BP algorithm is optimized and improved, and the improved BP algorithm is applied to the fault diagnosis of turntable.An anti-noise training method is used to train the network so that the diagnostic system can overcome the interference of certain amplitude noise.Compared with the standard neural network, the fuzzy neural network with improved BP algorithm has good global approximation ability and can accurately realize fault identification.Finally, the programming of the diagnosis system is completed with the mixed programming technology of Delphi and Matlab.The diagnosis system is divided into four functional modules: neural network training, wavelet denoising, database management and fault diagnosis. It has the ability of real-time monitoring and diagnosis.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2011
【分类号】：TH165.3;TP18

【相似文献】