Fault diagnosis for centrifugal pumps based on complementary ensemble empirical mode decomposition, sample entropy and random forest

Fault diagnosis of rotary machine is becoming extremely important because of the increasing complexity of modern industrial systems and the rising demands for quality, cost efficiency, reliability, and safety. In this paper, a fault diagnosis method based on complementary ensemble empirical mode decomposition (CEEMD)-sample entropy (SampEn) combined with random forest (RF) is presented and applied to practical fault diagnosis for centrifugal pumps. CEEMD, a novel noise assisted method based on empirical mode decomposition (EMD), is applied to decompose the nonlinear and non-stationary vibration signals into a series of intrinsic mode functions (IMFs). Then, the SampEn is utilized to characterize the complexity of the IMF components in different time scales. After the feature extraction, an RF classifier is introduced for identification and classification of fault modes of the centrifugal pumps. The analysis results demonstrate that the proposed method has desirable diagnostic performance for different fault types of centrifugal pumps.