基于小波包变换和RBF网络的液压系统泄漏故障诊断∗

摘要

Conventional method was difficult to acquire good fault diagnosis effects for there were various fault mechanisms and types in hydraulic system leakage. A new method was presented which based on wavelet packet transform and RBF neural networks, the fault features extraction with wavelet packet decom-position and RBF neural network learning algorithm was proposed. The vibration signal was collected with experiment, with three layer wavelet packet decomposition, the 8 bands energy of different frequency was get and serve as the input of RBF neural networks, with the training of RBF neural networks, the fault was recognized. This method combined the high time-frequency resolution feature of wavelet packet transform with the learning ability of RBF neural networks, so the higher diagnosis precision and efficiency was got.%液压系统泄漏故障原因众多，故障机理复杂，常规手段难以取得较好的故障诊断效果。提出基于小波包变换和RBF神经网络相结合的故障诊断方法。给出了基于小波包的故障特征提取方法和RBF神经网络训练算法。通过试验获取液压系统的振动信号，通过三层小波包分解，获取8个频段的能量信号，并以此作为神经网络的输入，通过网络训练，进行故障特征识别。该方法将小波包的时频分解能力和RBF神经网络的学习能力有机结合，取得了较高的故障诊断效率。