摘要
针对大型复杂机械振动建模困难的问题,提出了基于混沌理论及小波神经网络的建模思路。以某大型平整轧机剧烈振动为研究对象,研究了其振动时间序列信号的非线性特征及相空间重构技术。基于混沌理论,应用小波神经网络技术反演了轧机动力系统的振动模型。试验对比了此模型与BP网络模型的预测性能,结果表明小波神经网络模型具有收敛速度快、预测精度高的特点。
In view of the difficulty to model a large-complex machine vibration, a new modeling approach based on chaos and wavelet neural networks was presented. Aimed at the problem of abnormal vibration of a large temper rolling mill, the non-linear properties and phase space reconstruction techniques for time series vibration signals were analyzed. Based on chaos theory, the non-linear vibration model of wavelet neural networks was set up through inversion method. The properties of the model prediction were tested and compared with a BP neural networks. The results show that the wavelet neural networks have an advantage over the BP neural networks in rapid convergence and high accuracy.
出处
《机床与液压》
北大核心
2008年第7期158-160,共3页
Machine Tool & Hydraulics
关键词
轧机振动
相空间重构
小波神经网络
建模
预测
Rolhng mill vibration
Phase space reconstruction
Wavelet neural networks
Modeling
Prediction
