北京航空航天大学学报 ›› 2018, Vol. 44 ›› Issue (10): 2238-2246.doi: 10.13700/j.bh.1001-5965.2017.0756

• 论文 • 上一篇    

改进BA优化的MKSVDD航空发动机工作状态识别

何大伟1, 彭靖波1, 胡金海1, 宋志平2   

  1. 1. 空军工程大学 航空工程学院, 西安 710038;
    2. 西安交通大学, 西安 710054
  • 收稿日期:2017-12-06 修回日期:2018-02-25 出版日期:2018-10-20 发布日期:2018-10-29
  • 通讯作者: 彭靖波,E-mail:pjb1209@126.com E-mail:pjb1209@126.com
  • 作者简介:何大伟,男,硕士研究生。主要研究方向:航空发动机控制品质辨识;彭靖波,男,博士,副教授,硕士生导师。主要研究方向:航空发动机分布式控制;胡金海,男,博士,副教授,硕士生导师。主要研究方向:信号处理与故障隔离;宋志平,男,博士,高级研究员。主要研究方向:航空发动机及控制系统建模、发动机故障诊断技术。
  • 基金资助:
    国家自然科学基金(51506221);陕西省自然科学基础研究计划(2015JQ5179)

Aero-engine working condition recognition based on MKSVDD optimized by improved BA

HE Dawei1, PENG Jingbo1, HU Jinhai1, SONG Zhiping2   

  1. 1. Air Force Engineering University, Faculty of Aeronautical Engineering, Xi'an 710038, China;
    2. Xi'an Jiaotong University, Xi'an 710054, China
  • Received:2017-12-06 Revised:2018-02-25 Online:2018-10-20 Published:2018-10-29

摘要: 为了提高航空发动机工作状态识别准确率和效率,避免人工识别中存在的误判和耗时耗力问题,提出了基于混沌脉冲蝙蝠算法(CRBA)优化的多核支持向量数据描述(CRBA-MKSVDD)智能识别方法。研究了多核支持向量数据描述(MKSVDD)改进策略,引入混沌脉冲发射率提高了蝙蝠算法(BA)的收敛速度和收敛精度,得到了CRBA;通过CRBA优化MKSVDD的惩罚因子和核参数,同时对飞参数据进行了特征提取;基于特征飞参数据训练了CRBA-MKSVDD分类器,并对某型发动机一个飞行架次的工作状态进行了识别。结果表明,该方法识别准确率达到97.547 9%,可用于与发动机工作状态的相关研究和应用。

关键词: 多核支持向量数据描述(MKSVDD), 改进蝙蝠算法, 航空发动机, 工作状态识别, 飞参数据

Abstract: In order to ameliorate the accuracy and efficiency of aero-engine working condition identification, and to avoid the misjudgment and time-consuming problems in manual identification of aero-engine working condition, an intelligent recognition method, multi-kernel support vector data description based on chaotic rate bat algorithm (CRBA-MKSVDD), is proposed. The improved strategy of multi-kernel support vector data description (MKSVDD) is researched. The chaotic rate method is introduced to improve the convergence speed and convergence accuracy of the bat algorithm (BA), and the chaotic rate bat algorithm (CRBA) is obtained with this method. The penalty factor and kernel parameter of MKSVDD are optimized by CRBA and the characteristics of the flight parameters have been extracted. The CRBA-MKSVDD classifiers are trained based on the characteristics of flight parameters, and the working condition of a certain type of aero-engine in one sortie is identified by the proposed method. The results show that the accuracy of aero-engine working condition identified by the proposed method is 97.547 9%, which means that the method can be used in the research and application related to aero-engine working condition.

Key words: multi-kernel support vector data description (MKSVDD), improved bat algorithm, aero-engine, working condition recognition, flight parameters

中图分类号: 


版权所有 © 《北京航空航天大学学报》编辑部
通讯地址:北京市海淀区学院路37号 北京航空航天大学学报编辑部 邮编:100191 E-mail:jbuaa@buaa.edu.cn
本系统由北京玛格泰克科技发展有限公司设计开发