基于量子遗传算法的轴向柱塞泵故障特征选择

为了进一步减少特征维数、缩短运算时间、提高分类正确率等，提出了一种基于量子遗传算法的轴向柱塞泵故障特征选择方法，该方法采用量子位进行染色体编码，利用量子门更新种群。首先，对轴向柱塞泵振动信号进行小波包变换，提取出原始信号和各个小波包系数的统计特征；然后，利用量子遗传算法从原始特征集中选择出最优特征集；最后，以神经网络为分类器(其输入为最优特征集)，对故障进行诊断与识别。利用该方法对轴向柱塞泵正常、缸体与配流盘磨损和柱塞滑履松动三种状态的特征集进行选择，试验结果表明，与普通遗传算法相比，量子遗传算法可以更有效地减少特征维数，提高分类正确率。

Fault Feature Selection Method for Axial Piston Pump Based on Quantum Genetic Algorithm

In order to reduce feature dimension, shorten calculation time and improve classification accuracy, a fault feature selection method for axial piston pump was proposed based on quantum genetic algorithm. In this method, chromosomes were coded by quantum bits, and population was updated with quantum gate. Firstly, the vibration signals of axial piston pump were decomposed by wavelet transform, and the statistic features were extracted from original signals and each wavelet coefficient. Then, the optimal feature set was selected form original feature set by QGA. Finally, by using neural network as classifier, the optimal feature set was used as input for fault diagnosis. This proposed method was used for distinguishing different operating states of axial piston pump. The experimental results show, compared with common genetic algorithm, QGA can reduce feature dimension more effectively and improve classification accuracy greatly.