【作者】 卢卫萍；

【导师】 张文志；

【作者基本信息】 燕山大学 ， 机械设计及理论， 2009， 硕士

【摘要】 随着连轧技术的出现,轧机越来越朝着大型化、连续化、自动化的方向发展。轧机的传动系统一旦出现故障,就会造成整台轧机甚至整条生产线的停产,由此将造成巨大的经济损失。在轧机传动系统出现的故障中,齿轮箱的故障占有很大的比例。齿轮箱的故障诊断方法很多,有振动诊断、声学诊断、油样分析等。其中,振动信号分析是对齿轮箱进行状态监测与故障诊断的重要手段,也是最有效的诊断方法。本文主要对齿轮箱中的齿轮和轴承的振动信号进行分析。本文对齿轮和轴承进行振动机理研究,找出故障发生的原因。并根据小波去噪理论对测得的信号进行去噪处理,从而有效地去除了混杂在其中的高频噪声信号,提取了故障特征信号频带内的信号成分。小波神经网络是建立在小波分析基础上的一种新型神经网络,它综合了小波分析的时频分析特性和神经网络的自学习特性。本文将小波神经网络应用到齿轮箱故障诊断当中。通过试验验证,小波神经网络应用到齿轮箱的故障诊断当中,有利于提高故障的确诊率,有着良好的发展前景。利用MATLAB和Visual C++进行混合编程,并结合nnToolKit神经网络工具包设计出基于小波神经网络的智能诊断程序,该程序界面友好、易于操作、占用空间小、安装方便,并且脱离了MATLAB与VC环境。更多还原

【Abstract】 With the appearance of tandem rolling, the rolling mills become more and more large-sized、sequential and roboticized. Once the gearing of rolling mill breaks down, the whole rolling mill will be unable to work normally, and the economic losses will be enormous. In the fault of rolling mill’s gearing system, gear-box’s fault holds a large proportion.There are many methods about fault diagnosis, such as vibration diagnosis、acoustics diagnosis、oil sample analysis and so on. Viabration analysis is widely used in the condition monitoring and fault diagnosis of the gear-box system, and also is the most effective method. Analysing the viabration signal of gear and bearing is the main work of the dissertation.In the dissertation, the mechanism of gear-box’s vibration is studied, and the reason of fault is found out. And through the theory of de-noise, the high frequency noise is excluded and the characteristic signal frequency band is abstracted.Wavelet neural network which is based on wavelet analysis rationale is a sort of new neural network. It incorporates the time-frequency character of the wavelet analysis and the self-study function of the neural network. In the paper, the wavelet neural network is applied in fault diagnosis of the gear-box. The results of the experimentation show that the wavelet neural network which is applied in fault diagnosis will increase the accuracy and will have a good develop tendency.The intelligentized diagnosis program is designed which combines nnToolKit neural network toolbox with MATLAB and Visual C++. The program , which gets rid of MATLAB and VC environment, has humanization interface , is easy to work and install.更多还原