中文摘要：

为消除多通道观测信息冗余，压缩高维故障特征，简化特征提取工作，提出了一种基于独立分量分析和互信息分析的信息融合压缩方法。利用独立分量分析的冗余取消特性和互信息的高阶统计特征，实现了对机器多通道传感观测信息的两级融合压缩。实验结果表明：在尽可能保留原始观测信息的前提下，多通道传感观测实现了信息充分融合，维数显著压缩，并保持了对不同故障模式较好的分类。从而，为构建在线实时的机器故障模式分类器奠定了基础。

英文摘要：

In order to reduce redundancy among multi-channel observations by sensors, compress high dimensional fault features, and make feature extraction simple, a new method for fusion and compression of observations by sensors based on independent component analysis （ICA） and mutual information （MI） is proposed. By means of such characteristics as redundancy reduction of ICA and higher statistic than that of second order of MI, two-step fusion and compression of multi-channel observations by sensors are implemented. Results of experiment shows that observations are fused sufficiently, dimension of observations is reduced remarkably, and good performance in classifying different fault patterns is obtained. Thus, it is possible to construct an on-line and real-time fault classifier by the use of this method.