Wang Liangwen

Multifractal characterization of mechanical vibration signals through improved empirical mode decomposition-based detrended fluctuation analysis:

A novel multifractal detrended fluctuation analysis based on improved empirical mode decomposition for the non-linear and non-stationary vibration signal of machinery is proposed. As the intrinsic mode functions selection and Kolmogorov–Smirnov test are utilized in the detrending procedure, the present approach is quite available for contaminated data sets. The intrinsic mode functions selection is employed to deal with the undesired intrinsic mode functions named pseudocomponents, and the two-sample Kolmogorov–Smirnov test works on each intrinsic mode function and Gaussian noise to detect the noise-like intrinsic mode functions. The proposed method is adaptive to the signal and weakens the effect of noise, which makes this approach work well for vibration signals collected from poor working conditions. We assess the performance of the proposed procedure through the classic multiplicative cascading process. For the pure simulation signal, our results agree with the theoretical results, and for the contaminated time series, the proposed method outperforms the traditional multifractal detrended fluctuation analysis methods. In addition, we analyze the vibration signals of rolling bearing with different fault types, and the presence of multifractality is confirmed.

Multifractal characterization of plunger pump vibration signal through improved empirical mode decomposition based detrended fluctuation analysis

Promptly and accurately detecting the plunger pump fault in the hydraulic system is a serious issue in terms of improving reliability and decreasing accidents. A main work is analyzing the character of the collected samples. We used an improved empirical mode decomposition (EMD) based multifractal detrended fluctuation analysis (MFDFA) to extract the multifractal characters. The current method utilizes intrinsic mode functions (IMFs) selection and Kolmogorov - Smirnov test (K-S test) in the detrending procedure. The IMFs selection is used to deal with the undesired IMFs, and the two-sample K-S test works on each IMF and Gaussian noise to detect the noise-like IMFs. The proposed method adaptive to the nature of data and weakening the effect of noise make this approach work well for the non-stationary signal from the real system. We used the proposed method on the plunger pump vibration signal in the hydraulic system to verify the present of multifractal.

Intelligent fault diagnosis of plunger pump in truck crane based on a hybrid fault diagnosis scheme

At the initial stage of the mechanism, the collected samples are always in actual state, and the signals in fault conditions are gathered after a certain running time, so the general fault diagnosis model cannot be trained effectively. In this paper, a hybrid fault diagnosis scheme for pump in truck crane was proposed based on particle swarm optimization (PSO) SVDD and DBI K-Cluster method. Firstly, the SVDD procedure was constructed with the data in actual state, and the model parameters were optimized with PSO algorithm. Secondly, when the total number of novelty samples reached a given threshold, the K-Cluster method was utilized to classify the collected samples and the labels were allocated. In this procedure, the number of the class was determined with the Davies Bouldin index (DBI). Finally, each class data was trained with SVDD, and a whole diagnosis model was constructed with all the two-class classifiers. For the multi-fault mode samples of the pump in truck crane, experiments show that a promising classification performance is achieved.