摘要
面向齿轮系统20个常见故障,依据采集的齿轮箱震动噪音单一序列数据,采用基于超限学习机的神经网络算法进行深度挖掘,最终形成该研究设计的故障诊断预警体系。与文献中相关系统给出的实测结果平均值相比,该系统的预警时间延迟显著缩短,预警敏感度、特异度、综合准确率显著提升。作者认为,虽然该系统只针对结构简单的三轴齿轮变速箱开展了仿真研究,但该研究的设计理念适用于大部分复杂齿轮系统。
For 20 common faults of gear system,according to the collected single sequence data of gearbox vibration and noise,the neural network algorithm based on overrun learning machine is used for in-depth mining,and finally the fault diagnosis and early warning system designed in this research is formed.Compared with the average of the measured results given by the relevant systems in the literature,the early warning time delay of the system is significantly shortened,and the early warning sensitivity,specificity and comprehensive accuracy are significantly improved.The author believes that although the simulation research of the system is only carried out for the three-axis gear box with simple structure,the design concept of the research is suitable for most complex gear systems.
作者
秦玉霞
Qin Yuxia(Technological Vocational College of Dezhou,Shandong,Yucheng,251200)
出处
《现代科学仪器》
2022年第4期244-248,共5页
Modern Scientific Instruments
关键词
循环自相关
周期信号解调
齿轮故障
故障预警
神经网络
cyclic autocorrelation
Periodic signal demodulation
Gear failure
Fault early warning
Neural network
