摘要
为得到更有效的故障识别分类器,文中采集来自多个传感器的数据,并计算它们的无量纲统计特征值,然后运用GP方法在由这些统计特征值和操作符组成的搜索空间中寻优。将优化得到的复合特片作为故障识别的分类器,并根据不同的检测环境,选取不同的终端符集,从而降低检测环境对诊断结果的影响。将这一理论应用到所研制开发的驱动桥性能测试系统中,效果比单一特征及不考虑检测环境的识别更精确可靠,取得了良好的效果。
In order to obtain better classifier of fault diagnosis, the information from multiple sensors is collected and its statistics features is computed. Then the optimal compound feature is searched by GP in the statistics and operators library. The compound feature is taken as classifier of fault diagnosis. According to different testing environment, different terminal libraries are chosen, which can reduce the influence of the testing environment on recognition result. This theory is applied to a driving-axle testing system developed by us. The result of detection is more reliable and accurate than that of the single feature and that no considering testing environment.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2006年第5期441-446,共6页
Chinese Journal of Scientific Instrument
关键词
遗传编程
信息融合
车桥
故障诊断
Genetic programming（GP） Information fusion Fault diagnosis Driving-axle
