基于广义模糊熵的故障特征参数选择

Feature Parameter Selection for Fault Diagnosis Based on the General Fuzzy Entropy

提出了基于广义模糊熵的特征参数优选准则，定义并分析了高维数据分布中心离散度与样本总体可分性的单调关系。基于某液体火箭发动机１６ 种典型故障的仿真数据集，利用改进的遗传算法对其故障特征参数进行优选。数值实验结果表明该算法具有理想的效果。

This paper presents a general fuzzy entropy based-selection criterion for optimal parameters, and the divergence measure of geometrical distribution for multivariate data is defined and its monotonic relation with the classification capability of samples is analyzed. The selection criterion is used to select the minimal and optimal parameters set for a liquid rocket engine system using modified genetic algorithm. The numerical experiments show that this selection algorithm is highly effective and the constructed fault classifier with the selected feature parameters is more robust.