神经网络对水电机组振动故障的应用研究

摘要

研究水电机组振动故障问题.由于引起水电机组故障原因十分复杂,且监测数据量大、冗余信息严重,采用传统的RBF神经网络对水电机组故障进行诊断,存在RBF网络结构复杂、训练时间长和诊断准确率低等缺陷.为了提高故障诊断的准确性,提出一种主成分分析和RBF神经网络相结合的水电机组故障诊断方法.首先用主成分分析方法对故障信息进行预处理,消除无用和冗余的信息,然后采用RBF神经网络对预处理后的故障信息进行训练和诊断,从而减少RBF神经网络的规模,简化网络结构,加快故障诊断速度.借用某电站实测机组数据进行仿真,结果表明,本文算法不仅很好地克服了传统RBF神经网络的不足,而且有效地提高了水电机组故障诊断准确性和效率,结果证明本文提出的水电机组故障诊断方法的有效性和优越性.%The hydroelectric units fault diagnosis problem is researched. Due to the excessive monitoring data and the complexity of fault reasons for hydroelectric units, the problems exist in neural network when it is used for the fault diagnosis of hydroelectric units, such as the complex structure, the long training time and the low diagnosis rate. In order to improve the accuracy of the fault diagnosis, a fault diagnosis method for hydroelectric units is put forward based on RBF neural network and principal component analysis. The fault information of hydroelectric units is reduced by the principal component analysis. Then the information is diagnosed by RBF neural network, which not only decreases the number of the network input nerve cells effectively, but also predigests the network structure. The application of an example proves that the proposed method can improve the accuracy and the efficiency of fault diagnosis of hydroelectric units.