Research On Fault Detection Of Rotating Machinery Based On Variational Mode Decomposition

As a vital part of the power system,rotating machinery is widely used in electric power,metallurgy,petrochemical and other fields.When a fault occurs,the normal operation of the entire industrial system will be affected and cause inestimable losses.Based on vibration signal analysis of rotating machinery,this paper starts from three aspects: vibration signal processing,fault feature extraction and fault detection.The main work is as follows:(1)Aiming at the problems of mode mixing in traditional adaptive recursive decomposition method,a variational mode decomposition(VMD)method was proposed to process vibration signals of rotating machinery.For the problem of lack of basis for the selection of the key parameter in variational mode decomposition,the center frequency observation method was adopted to select the appropriate K-value,and the penalty factor was selected according to the comprehensive analysis of decomposition effect and calculation efficiency.The composite index of kurtosis and correlation coefficient was used to select the main components for signal reconstruction.For the selection of reconstruction components,the kurtosis-correlation coefficient composite index is used as the evaluation basis to select the main components for signal reconstruction.Verified by the simulation signal and measured data of bearings,the result showed that the VMD method had a better effect on the retention of the fault information,and it could remove noise.(2)Aiming at the problem that traditional time-frequency fault feature extraction methods are difficult to extract comprehensive fault characteristics,multiscale fluctuation-based dispersion entropy is used for feature extraction in this step.The skewness is selected as the evaluation index,and key parameters of the multiscale fluctuation-based dispersion entropy were discussed to obtain the best fault feature extraction effect.Finally,the effectiveness of the proposed method was verified by the measured data of bearings,and the experimental results showed that the MFDE method with optimized parameters can better distinguish the state of bearings.(3)Traditional classification methods,such as support vector machine,have high requirements for positive and negative sample data,in order to solve this problem,a fault detection method based on support vector data description(SVDD)is proposed.The ant lion optimization(ALO)was selected to optimize SVDD parameters,and ALOSVDD method was proposed,which solved the problem that traditional parameter optimization methods needed to set too many parameters before parameter optimization,and had the advantages of strong robustness and high computational efficiency.Three kinds of fault state and one normal state were identified in the measured data of bearings.The result showed that ALO-SVDD algorithm could reduce the false alarm rate of faults by 10% compared with SVDD algorithm.(4)Complete the design and implementation of rotating machinery fault experiment on the laboratory platform,and build a data acquisition system based on LabVIEW to collect vibration data of centrifugal pumps in two fault states and one normal state.Then the VMD method was used to decompose and reconstruct the vibration data to remove the noise in the signal,the MFDE of the denoised signal was calculated as the fault feature,and input the fault feature data into ALO-SVDD model to complete the fault detection.Experimental result showed that the proposed method could greatly reduce the false alarm rate of the fault.