Research on Evaluation Method of Hydraulic Pump Degradation Degree Based on DBN
Aiming at the problem that the central spring failure of axial piston pump is difficult to be evaluated effectively,a hy-draulic pump deterioration degree evaluation method based on Mel frequency cepstrum coefficient(MFCC)and deep belief network(DBN)was proposed.The normal data and three different degrees of center spring failure data collected in the field were subjected to signal pre-processing,including pre-emphasis,framing and windowing.Fast Fourier transform(FFT)was applied to the pre-processed signal,and their frequency spectrum and power spectrum were obtained.Then they were passed through the Mel filter bank to get the logarithmic energy of the signal.Finally,discrete cosine transform was performed for logarithmic energy,the Mel frequency cepstrum co-efficients and the first-order difference coefficients of the signal were obtained,forming the feature vector.A deep learning model was built based on the DBN method to learn the feature vectors,and the test samples were imported into the deep learning model to evaluate the central spring failure degree.The recognition results of the Mel frequency cepstrum coefficients and the first-order difference coeffi-cients were compared.The results show that when the Mel frequency cepstrum coefficients are selected as the feature vectors,they have a high recognition accuracy and can effectively identify the degree of performance deterioration of the center spring of the axial piston pump.
Mel frequency cepstral coefficient(MFCC)deep belief network(DBN)axial piston pumpdeterioration evaluating
NETL
NSTL
万方数据
