中国机械工程

• 机械基础工程 • 上一篇    下一篇

基于字典相干性优化的稀疏分类在发动机空燃比故障识别中的应用

吴士力1;唐振民1;刘奇2   

  1. 1.南京理工大学计算机科学与工程学院,南京,210094
    2.南京交通职业技术学院长安福特汽车有限公司联合实验室,南京,211188
  • 出版日期:2017-12-10 发布日期:2017-12-08
  • 基金资助:
    国家自然科学基金资助项目(61305134)
    National Natural Science Foundation of China (No. 61305134)

Recognition Applications for Air Fuel Ratio Faults of Gasoline Engines Using Sparse Representation Classification Based on Optimization of Dictionary Coherence

WU Shili1;TANG Zhenmin1;LIU Qi2   

  1. 1.School of Computer Science and Engineering,Nanjing University of Science and Technology,Nanjing,210094
    2.Nanjing Vocational Institute of Transport Technology-Chang'an Ford Joint Research Laboratory,Nanjing,211188
  • Online:2017-12-10 Published:2017-12-08
  • Supported by:
    National Natural Science Foundation of China (No. 61305134)

摘要: 稀疏分类直接把故障样本作为原子会造成分类字典相干性较高,进而影响稀疏分类精度,为此提出一种通过有效降低分类字典相干性来提高稀疏分类效果的优化算法。该方法首先通过传播聚类算法获取分类子字典的代表原子,然后基于极分解和子空间旋转法对子字典进行相干性优化。在某型发动机上的实验结果表明,该算法在低相干性字典上能够对怠速和2000r/min工况下的5种常见空燃比故障进行高精度识别。

关键词: 稀疏分类, 字典相干性, 汽油发动机, 空燃比故障识别

Abstract: Sparse representation classification directly took fault samples as atoms which would result in higher coherence of classification dictionary. Thus, accuracy of sparse classification would be affected. A new optimization algorithm was proposed to improve effectiveness of sparse classification by effectively reducing the coherence of classification dictionary herein. Firstly, the representative atom of each sub-dictionary was obtained by affinity propagation clustering algorithm. Secondly, all the sub dictionaries consisted of representative atoms were optimized based on polar decomposition and subspace rotation methods. The experimental results of an engine show that, the novelty classification algorithm achieves high accuracy of recognition for five common faults in idle and 2000r/min operating conditions using the dictionary with lower coherence.

Key words: sparse representation classification, coherence of dictionary, gasoline engine, air-fuel ratio fault recognition

中图分类号: