基于类内类间判据与遗传算法的故障特征选择方法

发布时间：2018-03-26 16:57

本文选题：特征选择　切入点：类内类间距离判据　出处：《兰州理工大学学报》2017年02期

【摘要】：针对原始故障数据集因"高维"和"海量"引发的"维数灾难"问题,提出一种基于类内类间距离判据和遗传算法相结合的故障特征选择方法.在提取出时域、频域、小波包频带能量作为描述系统状态的原始故障特征集基础上,经类内类间距离判据初次选择剔除不相关特征之后,引入遗传算法二次选择去除冗余特征,得到一种近似最优特征子集.结果表明:基于类内类间距离判据和遗传算法的故障特征选择方法可以剔除不相关和冗余特征,最终得到精简特征子集,并且筛选出的特征子集对故障类型的判别有很高的识别能力.
[Abstract]:Aiming at the problem of "dimensionality disaster" caused by "high dimension" and "magnanimity" in original fault data sets, a fault feature selection method based on inter-class distance criterion and genetic algorithm is proposed. On the basis of the wavelet packet band energy as the original fault feature set describing the system state, the genetic algorithm quadratic selection is introduced to remove redundant features after the inter-class distance criterion is selected for the first time to remove the irrelevant features, and the wavelet packet frequency band energy is used as the original fault feature set to describe the state of the system. An approximate optimal feature subset is obtained. The results show that the fault feature selection method based on inter-class distance criterion and genetic algorithm can eliminate irrelevant and redundant features, and finally obtain a reduced feature subset. And the selected feature subset has a high ability to identify fault types.
【作者单位】：兰州理工大学数字制造技术与应用省部共建教育部重点实验室;兰州理工大学机电工程学院;
【基金】：国家自然科学基金(51675253)
【分类号】：TH17;TP18

【相似文献】