摘要
首先利用蛙跳算法对最佳影响参数组合进行搜索,搜索结束后选择最优的参数,利用优化参数的变分模态分解对故障信号处理,得到本征模态函数;为了验证蛙跳算法得到的参数是否为最优参数,选择最佳的本征模态函数进行包络分析,将包络谱的特征频率与实际故障频率相比较;以得到的模态函数构成矩阵,进行奇异值分解,得到信号的奇异值,以奇异值作为极限学习机的输入,对故障类型进行分类。利用优化参数的变分模态分解对仿真信号和实测信号进行分析,均能提取特征信息,对故障类型进行识别,表明该方法有一定的实际意义和实用价值。
First,the leapfrog algorithm was used to search the optimal combination of influence parameters.The optimal parameters was selected after the search.Variational modal decomposition of optimized parameters was used to process fault signal.The eigenmode function was obtained.In order to verify whether the parameters obtained by the leapfrog algorithm were the optimal parameters,the Best Eigenmode function was selected to analyse envelope.The characteristic frequency of envelope spectrum was compared with the actual fault frequency.The matrix was formed with the obtained modal function.The singular value of the signal was obtained by using singular value decomposition.As input of ultimate learning machine,the singular value was used to Classify fault types.The simulated and measured signals were analysed by the variational modal decomposition of the optimized parameters.The feature information can be extract,and the type of fault was Identified.The result shows that this method has some practical significance and practical value.
出处
《中国机械工程》
EI
CAS
CSCD
北大核心
2018年第4期390-397,共8页
China Mechanical Engineering
基金
国家自然科学基金资助项目(51577007)
新能源电力系统国家重点实验室资助项目(LAPS15019)
中央高校基本科研业务费专项资金资助项目(2014JBZ017)
关键词
蛙跳算法
变分模态分解
极限学习机
滚动轴承
leapfrog algorithm
variational modal decomposition(VMD)
extreme learning machine(ELM)
rolling bearing