基于小波包—混沌支持向量机的液压泵压力信号预测

摘要

For non- linear and non stationary characteristics of pressure signals for a hydraulic pump,this paper presented a kind of prediction method, which combined wavelet packet analysis and chaos theory with SVM. This method can construct model and predict for pressure signals. Firstly,it used wavelet packet to decompose the time series of pressure signals for hydraulic pump and reconstruct every components. Secondly,it builts the corresponding chaos theory and SVM prediction model to predict every component. Finally, it reconstructed all levels of prediction results. Through the experiments, the method can predict the trend of pressure signals effectively and have higher prediction precision,and can be applied in the system of condition monitoring and prediction.%液压泵压力信号呈现的非线性、非平稳的特性,提出一种将小波包分析、相空间重构理论与支持向量机(SVM)相结合的预测方法,实现液压泵压力信号监测数据的建模及预测.首先将采集到的压力信号通过小波包进行分解,将分解得到的各个分量进行重构,其次对重构后的每一个分量通过混沌支持向量机预测模型进行预测,最后对各预测值进行合成.试验数据表明,该方法能够有效地预测液压泵压力信号的变化趋势,具有较高的预测精度,可有效地应用于系统的状态监测和故障预测.