基于改进V-detector算法的转子故障识别方法研究

为了提高转子故障诊断识别准确率, 提出一种基于改进V-detector算法的转子故障辨识方法。首先对V-detector算法进行了改进, 该算法通过改变拒绝和接受假设检验的条件来减少无效检测器的产生进而提高算法的检测准确率; 然后将信号的谱熵值作为特征向量, 并根据转子故障类型将其划分为多个自体样本集, 用改进后V-detector算法训练出多个检测器集; 最后利用其设计出能够识别转子故障的分类器。仿真结果表明, 改进的V-detector算法能产生较少的检测器, 覆盖率由95%升高至99%时检测器数目无明显增加, 与原算法相比提高了故障的辨识精度。

Research on fault data classification based on improved V-detector algorithm

In order to improve the rotor fault diagnosis accuracy, this paper proposed a rotor fault identification method based on improved V-detector algorithm. Firstly, it improved the V-detector algorithm, reduced the generation of the null detector to thereby improve the detection accuracy of the algorithm by changing the rejected and accepted conditions of hypothesis testing. Secondly, it used entropy value of the signal as a feature vector and divided it into different self sample set according to fault types, used the improved V-detector algorithm to train it for detector sets. Finally utilized it to design classifier which used to identify rotor fault. Simulation results show that the improved V-detector algorithm can produce fewer detectors, and the number of detectors do not improved when the estimated coverage increased from 95% to 99%. Compare with the traditional V-detector algorithm, the proposed method improve the identification accuracy of the failure.