中文摘要：

当机械设备多故障并发时，在每个测点测得的信号往往是多个故障信号的叠加，傅里叶变换、小波变换等传统方法都难以有效地分离故障特征。为了克服上述方法的缺陷，利用基于峭度的独立成分分析算法RobustICA对复合故障信息进行分离，提取故障特征。对4种不同信号进行随机混叠而生成的混合信号进行分离，仿真验证了RobustICA算法的有效性。最后，对轴向柱塞泵出现滑靴与斜盘磨损时的复合故障振动信号进行了分离实验，达到了良好的分离效果，证明了该方法对于液压泵复合故障振动信号进行分离的有效性。

英文摘要：

With the mechanical equipment fault concurrency, each measured signal is often composed of several fault signals, but the traditional methods such as the Fourier transformation and the Wavelet Analysis can not realize the separation of fault characteristic effectively. In this study, the Robust Indepe~dent component Analysis （ RobustICA） based on kurtosis is researched, which can separate the compound fault information and extract the fault characteristics. The RobustICA is used to separate four different signals, and the results show that the RobustICA is an effective method. Finally, a separation experiment for compound fault vibration signal is carried out when the slipper and the swashplate wear of the axial piston pump appear. The separation efficiency is perfect. The experiment verifies the effectiveness of the RobustICA for separating the compound fault vibration signal of the hydraulic pump.