燃气轮机关键部件故障诊断方法研究

Research on the Intelligence Fault Diagnosis Methods of Gas Turbine Key Components

燃气轮机的工作环境恶劣,突发情况和出现故障的模式多、几率大。为此研究有效地燃气轮机故障诊断方法尤为重要。提出了一种EMD小波阈值降噪和KPCA-GRNN相结合的方法,对燃气轮机喷口加力调节器故障诊断进行了深入研究。针对某型真实燃气轮机进行测试试验采集的喷口加力调节器高压转子转速、低压转子转速等八个参数数据,首先采用经验模态分解(EMD)方法对8个参量信号进行EMD分解,然后采用软阈值函数对其进行小波降噪,并进行信号重构,从而可得到燃气轮机喷调工作状态有效数据。在此基础上采用核主元分析法提取喷口加力调节器样本集的不同主元,构建特征向量,并由特征向量建立GRNN神经网络故障诊断模型,通过测试数据进行试验验证,验证了该方法的有效性。此外,尚采用基于KPCA-GRNN的方法对传感器感知的喷口加力调节器的八个参数原始数据进行了诊断方法研究。结果表明,采用EMD小波阈值降噪和KPCA-GRNN相结合的方法,能有效识别出喷口加力调节器不同的状态,具有很好的实际应用价值。

Gas Turbine because of the bad work environment,mode of failure,more likely.For this study effectively gas turbine fault diagnosis method is particularly important.It proposes a EMD wavelet threshold de-noising and the method of combining KPCA-GRNN,fault diagnosis of gas turbine nozzle torque regulator were studied.For a certain type of gas turbine test experiment was carried out to collect real nozzle torque regulator high-pressure rotor speed,low pressure rotor speed eight parameters,first using empirical mode decomposition(EMD)method of each parameter signal EMD decomposition,and then the soft threshold function is adopted to carry out noise reduction,and reconstruct the signal,thus can get nozzle torque regulator work status valid data.On the basis of using kernel principal component analysis method to extract nozzle torque regulator different principal component of the sample set,construct feature vectors,and the eigenvector GRNN neural network fault diagnosis model is established,through the test data,test and verify the effectiveness of the proposed method.In addition,it is based on KPCA-GRNN method to nozzle torque regulator for the study of diagnostic technique.Results show that the EMD wavelet threshold de-noising and combining KPCA-GRNN method,can effectively identify the nozzle strength regulator of different state,has the very good practical application value.