Abstract: In order to solve the problems of inaccurate population initial and complicated fitness function in feature selection, this paper proposes a novel feature selection algorithm—Improved Quantum Genetic Algorithm-Partial Least Square（IQGAPLS）. In IQGAPLS algorithm, according to the fact of problem, population is given initial value. Meanwhile, a new fitness function is designed for reducing computation amount. Based on previous fitness function, the formula for updating rotation angle is proposed. IQGAPLS is applied to feature selection for fault signal of axial piston pump. The experimental results indicate that IQGAPLS has less computation amount and shorter execution time. The selected features contain more information of fault states, which can enhance classification accuracy.

Key words: quantum computation, fitness function, Quantum Genetic Algorithm Partial Least Square（QGAPLS）, feature selection, axial piston pump

摘要： 针对量子遗传偏最小二乘法在特征选择过程中，存在初始化种群粗糙和适应度函数复杂等问题，提出了一种新的特征选择方法——改进的量子遗传偏最小二乘法（Improved Quantum Genetic Algorithm Partial Least Square，IQGAPLS）算法。该算法根据求解问题的实际情况，赋予种群初始值。同时，设计了一种新的适应度函数，以减少计算量，并基于此适应度函数，提出了一种新的旋转角度更新公式，解决了其方向和大小确定困难的问题。将该算法应用于轴向柱塞泵故障信号的特征选择中。实验结果表明，IQGAPLS算法具有较少的计算量和较短的执行时间，选择出的特征包含更多的工作状态信息，从而提高了分类准确率。

关键词: 量子计算, 适应度函数, 量子遗传偏最小二乘法, 特征选择, 轴向柱塞泵