航空学报 > 2016, Vol. 37 Issue (11): 3455-3465   doi: 10.7527/S1000-6893.2016.0036

特征选择的多准则融合差分遗传算法及其应用

关晓颖, 陈果, 林桐   

  1. 南京航空航天大学 民航学院, 南京 210016
  • 收稿日期:2015-11-19 修回日期:2016-01-29 出版日期:2016-11-15 发布日期:2016-02-02
  • 通讯作者: 陈果,男,博士,教授,博士生导师。主要研究方向:航空发动机智能诊断及专家系统,航空发动机整机振动与转子动力学、故障诊断。Tel.:025-84891850,E-mail:cgzyx@263.net E-mail:cgzyx@263.net
  • 作者简介:关晓颖,女,博士研究生。主要研究方向:智能计算、遗传算法、模式识别及故障诊断。Tel.:025-84891850,E-mail:xiaoying_close@sina.com;陈果,男,博士,教授,博士生导师。主要研究方向:航空发动机智能诊断及专家系统,航空发动机整机振动与转子动力学、故障诊断。Tel.:025-84891850,E-mail:cgzyx@263.net;林桐,男,硕士研究生。主要研究方向:航空发动机状态检测与故障诊断技术。Tel.:025-84891850,E-mail:nuaa_lintong@163.com
  • 基金资助:

    国家自然科学基金(61179057)

Feature selection method based on differential evolution and genetic algorithm with multi-criteria evaluation and its applications

GUAN Xiaoying, CHEN Guo, LIN Tong   

  1. College of Civil Aviation, Nanjing University of Aeronautics and Astronautics, Nanjing 210016, China
  • Received:2015-11-19 Revised:2016-01-29 Online:2016-11-15 Published:2016-02-02
  • Supported by:

    National Natural Science Foundation of China (61179057)

摘要:

为了全面评价特征子集的好坏,提高特征子集作为最佳子集的可靠性,以及更快找到最佳子集,提出了一种用于特征选择的多准则融合差分遗传算法。引入多个评价准则对特征子集进行评价,并对遗传算法的选择算子进行改进,有利于选出适应度高且具有重要特征的个体;同时,引入差分策略改进变异算子,提高种群多样性和算法搜索能力;最后通过仿真实验和滚动轴承实例验证了该方法的有效性。

关键词: 特征选择, 多准则, 差分进化, 遗传算法, 滚动轴承, 故障诊断

Abstract:

In order to make a whole evaluation to the selected feature subset, which improves the reliability of the best subset and the speed of its searching, the paper presents a novel feature method based on differential evolution and genetic algorithm with multi-criteria evaluation. This algorithm is used to evaluate the feature subset by the multi-criteria evaluation. Meanwhile, the improved genetic operators were proposed, which improves the selection operator and the mutation operator. Designing the selection operator with a combination of feature weight values and fitness is beneficial to selecting the individuals which contain the high fitness and important features from the population. In addition, it introduces differential strategy to improve mutation operator, which improves the diversity of evolution population and searching efficiency. Finally, simulation example tests the validity of the proposed algorithm. The validity of the proposed method is also verified with rolling bearing fault diagnosis.

Key words: feature selection, multi-criteria, differential evolution, genetic algorithm, rolling bearing, fault diagnosis

中图分类号: