基于差分进化和核主元分析的燃气轮机故障检测

燃气轮机气路部件的状态检测参数具有极强的非线性,其故障特征难以提取,而利用传统核主成分分析(KPCA)进行故障检测难以对核参数进行科学取值,从而降低故障检测的准确性.针对该问题,论文提出了基于优化混合核的核主元分析故障检测算法(DE-KPCA).首先建立动态权值混合核函数,通过调节核函数的权重比实现全局映射和局部映射优化组合.以样本检测精度作为优化目标,对混合核参数进行逐次优化.最后构造了基于优化混合核函数的主元异常状态检测方法,实现对燃气轮机气路故障的在线检测.本文通过对双轴涡喷发动机气路故障仿真的验证,证明了该方法相较传统KPCA检测,能够实现核参数的科学取值且对燃气轮机气路故障检测具有更高的准确性和实用性.

Fault detection of gas turbine air path system based on KPCA and DE

The state detection parameters of gas turbine gas-path components are extremely nonlinear and their fault characteristics are difficult to be extracted,using traditional KPCA for fault detection is difficult to scientifically value nuclear parameters, thus reducing the accuracy of fault detection.To solve this problem, this paper proposes a fault detection algorithm for kernel principal component analysis based on optimized hybrid kernel(DE-KPCA).Firstly, the dynamic weight hybrid kernel function is established, and the global and local mappings are optimized by adjusting the weight ratio of the kernel function. With the sample detection accuracy as the optimization target, the mixed core parameters were optimized successively.Finally, a principal component abnormal state detection method based on optimized hybrid kernel function is constructed to realize on-line detection of gas turbine gas-path faults.In this paper, the fault simulation of turbojet turbojet engine is verified, which proves that this method can realize the scientific value of nuclear parameters and is more accurate and practical for gas turbine gas-path fault detection than traditional KPCA detection.