一种复合KPCA故障诊断模型

A Compound KPCA Fault Diagnosis Model

核函数主元分析(KPCA)故障诊断方法中核函数的具体形式对诊断性能的影响非常大.针对核函数具体形式选取问题,基于径向基高斯核函数和一类特定多项式核函数,构造出一种新的复合核函数模型.对模型的构造方法进行了论述,给出了具体的故障诊断算法的实现步骤.该模型在兼顾全局信息提取的前提下,保证了局部灵敏度,具有很好的拟合能力.通过与其他算法仿真比较表明所提出方法不但可以避免对模型的事先假设,且具有较高的故障诊断效率.

The concrete form of kernel function in the KPCA fault diagnosis model has great impact on the diagnosis performance.A new compound kernel function model is presented based on RBF Gaussian kernels and polynomial kernels.The construction of the model and the implementation steps of the fault diagnosis algorithms are given.In the condition of considering the global information the model guarantees partial sensitivity and good fitting ability.Compared with other KPCA algorithms＇ simulation results,the proposed model not only avoids prior hypothesis but also has higher fault diagnosis efficiency.