基于RBF神经网络的往复泵泵阀故障诊断研究

摘要

According to wavelet analysis, the energy regularity of each frequency band is found, the characteristic vectors for values of pumps are constructed, the RBF neural network is trained. Through a lot of practices,both the characteristic vectors and the RBF neural network are proved to raise the diagnosis rate for valves of reciprocating pumps.%根据小波包分析,获得了各频带能量的分布规律,构造了泵阀状态特征向量,训练了RBF神经网络.大量的现场试验证明,构造的故障特征向量与RBF神经网络配合使用的方法可以明显提高泵阀故障诊断的准确率.