论文部分内容阅读
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 establishment of fault type of automatic identification system is the development of mechanical equipment diagnostic direction. The rise and development of neural network theory opens up a brand new road for the automatic identification of fault types. After the neural network learns the fault samples, it has a higher correct recognition rate for the samples with unknown faults. It is found from the neural network’s test results of the fault recognition that the classification of a single fault by the neural network is quite different from the classification of the combined fault. This paper analyzes the causes of this phenomenon, and uses a combination of networks to overcome the single network defects in the combination of fault classification accuracy is not high, and achieved satisfactory results.