Abstract:An feature extraction of the key step in prognostic of hydraulic pump. Since vibration signals of hydraulic pump are complex and degradation features are hard to extract, a novel method based upon DCS and relation entropy is proposed. First of all, in order to make reasonable use of feature information, earlier CS is modified by DCT and the DCS algorithm is presented to make fusion of multi-channel vibration signals. And DCS power entropy and singular entropy, which are relatively defined in Shannon entropy and Tsallis entropy, are extracted as features. On this basement, the feature fusion method based on relation entropy is proposed to remain original features performances and improve conciseness. According to max relation entropy criterion and gradual fusion strategy, the four extracted features are fused into a new one, which is considered as degradation feature. Finally, the proposed method is verified by vibration signals sampled from hydraulic pump degradation experiment. Key words:degradation feature extraction;information fusion;composite spectrum;relative entropy 0 前言* 液压泵是液压系统的关键部件之一,其性能好 坏直接影响着整个液压系统的可靠性[1] 。由于自身 结构影响,液压泵在性能退化过程中振动信号的复 杂度和随机性也会发生相应变化 [2][3]