摘要
特征提取是齿轮多重状态或故障分类中一个很重要的问题。为解决该问题,提出了利用频域特征和遗传编程对齿轮箱盖多类状态振动信号进行特征提取的方法。为了使诊断结果更好地可视化,利用遗传编程结果的固有随机性,提取两个新特征指标。结果表明,该方法对振动数据进行了准确的多重故障分类,也可以应用到其它机械故障诊断或分类中。
An approach is proposed in an attempt to solve a problem in gear fault diagnosis, i.e. classification feature extraction of multiple gear faults, for which the procedure based on genetic programming and power spectral density is embloyed. The power spectral density of gearbox vibration signals is estimated by the periodogram from which original feature sets are derived. A new feature index is constructed and selected from the feature sets by genetic programming, which successfully classifies all multiple gear conditions. To visualize the classificat;on results, the second new feature index is constructed fol- lowing the same procedure. Results show that two new feature indices can classify 6 patterns of conditions. The approach can also be applied to other mechanical fault diagnosis and classification.
出处
《振动工程学报》
EI
CSCD
北大核心
2006年第1期70-74,共5页
Journal of Vibration Engineering
基金
国家重点基础研究发展计划资助项目(2005CB724100)
国家自然科学基金(50405033)
武汉科技大学机械传动与制造工程省重点实验室开放基金(2003A15)资助
关键词
故障诊断
齿轮
遗传编程
特征提取
分类指标
fault diagnosis
gear
genetic programming
feature extraction
classification index
