论文部分内容阅读
一般对特定的基于多层感知器的故障诊断问题 ,很难确定神经网络的结构。在分析了多层感知器对故障的识别和诊断能力后 ,采用由小到大和由大到小的方法确定神经网络隐层数与隐层单元数。研究了基于神经网络和振动频谱的旋转机械故障诊断方法 ;研制了一个基于该方法的智能故障诊断系统 ,该系统集网络学习、故障诊断、数据库管理和数据查询为一体。在该智能诊断系统中采用了知识子块的概念 ,系统界面友好 ,交互性强。将其应用于某个大型风机的故障诊断中 ,结果表明该系统操作方便 ,诊断结果准确可靠 ,且具有很强的鲁棒性 ,对某些情况可以实现自动诊断 ,进而证明了该诊断方法的有效性。
In general, it is difficult to determine the structure of a neural network for a particular fault diagnosis problem based on multi-layer perceptrons. After analyzing the multi-layer perceptron’s ability to identify and diagnose faults, the number of hidden layers and hidden layer units of neural network is determined by the method of ascending and descending. A fault diagnosis method of rotating machinery based on neural network and vibration spectrum is studied. An intelligent fault diagnosis system based on this method is developed. It integrates network learning, fault diagnosis, database management and data query. In this intelligent diagnosis system, the concept of knowledge sub-block is adopted, and the system interface is friendly and interactive. It is applied to the fault diagnosis of a large fan. The results show that the system is easy to operate, the diagnosis results are accurate and reliable, and has strong robustness. In some cases, automatic diagnosis can be realized, Effectiveness.