大型回转机械故障的神经网络分类器

建立故障类型的自动识别系统是机械设备诊断学的发展方向。神经网络理论的兴起和发展为故障类型的自动识别开辟了一条崭新的道路。神经网络通过对故障样本的学习后，对未知故障的样本具有较高的正确识别率。从神经网络对故障的识别检验结果中发现，神经网络对单一故障的分类与对组合故障的分类效果相差较大。本文分析了产生这一现象的原因，并利用组合网络来克服单一网络对组合故障分类精度不够高的缺陷，取得了令人满意的结果。

The Method of Automatic Recognition of Large-sized Rotating Machinery Faults

It is the future tendency of mechanical equipment diagnostics to build up automatic faults recognition system.The arising and developing of neural networks opens a new way for automatic faults recognition.After it has been trained with a group of faults examples, neural networks can correctly recognize an unknown fault. From the test results of faults recongnition, it is found out that classification effect of nerual networks has apparent difference between single fault and combinating fault.Theoretical analysis to the phenomena above is given and a type of network canlled multi - network is built up to improve chassifiction results of neural networks to combinating fault in this paper.