An improved ensemble empirical mode decomposition(EEMD) algorithm is described in this work, in which the sifting and ensemble number are self-adaptive. In particular, the new algorithm can effectively avoid the mode ...An improved ensemble empirical mode decomposition(EEMD) algorithm is described in this work, in which the sifting and ensemble number are self-adaptive. In particular, the new algorithm can effectively avoid the mode mixing problem. The algorithm has been validated with a simulation signal and locomotive bearing vibration signal. The results show that the proposed self-adaptive EEMD algorithm has a better filtering performance compared with the conventional EEMD. The filter results further show that the feature of the signal can be distinguished clearly with the proposed algorithm, which implies that the fault characteristics of the locomotive bearing can be detected successfully.展开更多
针对现有的卷积、循环模型预测滚动轴承剩余使用寿命(Residual Life,RL)精度低的问题,提出一种基于改进自注意力机制的RL预测模型。首先,针对Transformer模型中自注意力机制内存占用高、信号存在噪声信息的问题,在窗口自注意力机制(Wind...针对现有的卷积、循环模型预测滚动轴承剩余使用寿命(Residual Life,RL)精度低的问题,提出一种基于改进自注意力机制的RL预测模型。首先,针对Transformer模型中自注意力机制内存占用高、信号存在噪声信息的问题,在窗口自注意力机制(Window Based Multi-head Self-attention,W-MSA)的基础上,提出概率窗口自注意力机制(Probwindow Based Multi-head Self-attention,PW-MSA);然后,针对多头信息不匹配和缺少局部信息的问题,采用Talking Head方法实现多头信息融合,并在前馈神经网络层加入深度可分离卷积提取局部信息,从而提升模型的预测精度。采用PHM2012轴承数据集将改进前后的自注意力机制模型进行比较,并和现有的先进预测模型对比,结果表明,改进自注意力机制模型可使预测精度提升13.04%。展开更多
基金Project(61573381)supported by the National Natural Science Foundation of ChinaProject(2012AA051601)supported by the National High-tech Research and Development Program of China
文摘An improved ensemble empirical mode decomposition(EEMD) algorithm is described in this work, in which the sifting and ensemble number are self-adaptive. In particular, the new algorithm can effectively avoid the mode mixing problem. The algorithm has been validated with a simulation signal and locomotive bearing vibration signal. The results show that the proposed self-adaptive EEMD algorithm has a better filtering performance compared with the conventional EEMD. The filter results further show that the feature of the signal can be distinguished clearly with the proposed algorithm, which implies that the fault characteristics of the locomotive bearing can be detected successfully.
文摘针对现有的卷积、循环模型预测滚动轴承剩余使用寿命(Residual Life,RL)精度低的问题,提出一种基于改进自注意力机制的RL预测模型。首先,针对Transformer模型中自注意力机制内存占用高、信号存在噪声信息的问题,在窗口自注意力机制(Window Based Multi-head Self-attention,W-MSA)的基础上,提出概率窗口自注意力机制(Probwindow Based Multi-head Self-attention,PW-MSA);然后,针对多头信息不匹配和缺少局部信息的问题,采用Talking Head方法实现多头信息融合,并在前馈神经网络层加入深度可分离卷积提取局部信息,从而提升模型的预测精度。采用PHM2012轴承数据集将改进前后的自注意力机制模型进行比较,并和现有的先进预测模型对比,结果表明,改进自注意力机制模型可使预测精度提升13.04%。
