基于极限学习机的汽轮机故障诊断

Fault Diagnosis of Steam Turbine Based on Extreme Learning Machine

针对汽轮机故障诊断中出现的多故障识别问题,为有效提高汽轮机多故障诊断的准确率,提出了基于极限学习机的汽轮机故障诊断方法。极限学习机算法在训练时只需设置隐含层神经元个数,从而解决了如神经网络及支持向量机等多参数选取困难的问题,有效地提高了学习机的训练速度。在确定了最优参数的基础上,将极限学习机应用于汽轮机故障诊断模型中,并将极限学习机的故障诊断结果与支持向量机的诊断结果进行对比。结果表明:基于极限学习机的多故障诊断速度及准确率均明显优于支持向量机的诊断结果,对汽轮机故障诊断的实践有非常显著的指导作用。

Aiming at fault diagnosis of steam turbine,an extreme learning machine-based fault diagnosis method for it was proposed.In the training,this algorithm just asks for the number of hidden neurons so that the difficulty in selecting parameters like neural network and support vector machine can be solved,and the training speed of learning machine can be effectively enhanced.Basing on having optimal parameters determined,the extreme learning machine was applied to steam turbine＇ s failure diagnosis and then having its diagnostic results compared with that from the support vector machine,the results show that both fault diagnosis and recognition rate of the extreme learning machine outperforms the identification accuracy of support vector machine,and it has significant guidance for steam turbine＇ s fault diagnosis.