DPCA和GRNN在燃气轮机故障诊断的方法

摘要

Its key components of nozzle strength regulator as the research object of fault diagnosis, this paper proposes a based on dynamic principal component analysis (DPCA) and generalized regression neural network (GRNN),combination of nozzle strength regulator fault diagnosis methods. First of all,the engine dedicated test platform to test. Collection nozzle torque controller of high pressure rotor speed,low pressure rotor speed,fuel oil,fuel oil consumption of parameters such as original data,the preprocessing,and dynamic principal component analysis,to extract the different health status of the yuan,build the GRNN neural network fault diagnosis model,and through the test data to test the effectiveness of the proposed method. To show the effectiveness of the proposed method,the article is adopted based on the GRNN and DPCA - based RBF method to vent torque controller different health status diagnosis technology research,and the diagnosis results obtained by different methods are analyzed in comparison. Results show that the combination of DPCA and GRNN based fault diagnosis method can effectively identify the nozzle torque controller different state of health,has the very good practical application value.%选取其关键部件—喷口加力调节器作为故障诊断研究对象,提出了一种基于动态主元分析(DPCA)和广义回归神经网络(GRNN)相结合的喷口加力调节器故障诊断方法.在燃气轮机专用试验平台对其进行试验,采集喷口加力调节器的高压转子转速、低压转子转速、燃油油量、燃油耗量等参数原始数据,对其进行预处理,并采用DPCA方法对其进行动态主元分析,提取其不同健康状态的主元,构建特征向量,采用特征向量构建GRNN神经网络故障诊断模型,并通过测试数据对该方法的有效性进行试验验证.为表明该方法的有效性,采用了基于GRNN和基于DPCA-RBF的方法对喷口加力调节器不同健康状态进行了诊断技术研究,并对不同方法所得到的诊断结果进行了对比分析.结果表明,采用DPCA和GRNN相结合的故障诊断方法能有效识别出喷口加力调节器不同的健康状态,具有很好的实际应用价值.