Research On Gear Pump Wear Fault Diagnosis Based On Metaheuristic Optimization Algorithm

Gear pump is widely used in hydraulic system of engineering machinery because of the reliable performance,simple structure and low cost.However,gear pump worked for a long tiame under the harsh conditions that high-pressure heavy loads makes it become more prone to failure and higher cost component.In order to avoid irreparable losses caused by failure and render certain the steady running of the hydraulic system,this paper deeply studies the fault diagnosis method of gear pump based on data driven.Firstly,a denoising method based on arithmetic optimization variational mode decomposition(AOA-VMD)was adopted for eliminating the noise and interference of vibration signal from gear pump.The effectiveness of this method was proved by simulating the sinusoidal signal.The decomposed intrinsic mode components(IMF)with greater correlation were reconstructed to retain the complete primeval information as much as possible.28 characteristics of fault information are extracted based on the reconstructed signal from the time-domain,frequency domain and time-frequency domain.Secondly,the dimension of characteristics from the reconstructed signal was large,and characteristics contained a few irrelevant and redundant features.The infinite latent feature selection(ILFS)was used to select the more useful fault features of gear pumps,and the experimental results were compared with the results of ReliefF and maximum correlation minimum redundancy(mRMR)selection.The experimental results verified the expressiveness and separability of the optimal feature subset selected by ILFS.At the same time,t-distribution similarity network evaluation(t-SNE)was used to decrease the dimension of characteristics and display the results,which equilibrium the reservations of local and global information of features.Experimental results showed that the differentiability of t-SNE was more effective than kernel principal component analysis(KPCA)and linear discriminant analysis(LDA).Finally,two kinds of gear pump wear fault diagnosis methods were proposed.One was a gear pump wear fault diagnosis method based on sand cat swarm optimizes support vector machine(SCSO-VMD),which used SCSO to adaptively optimize the two parameters of SVM according to the characteristic parameters to obtain the best diagnostic effect.The other was synchronous optimization feature selection based on Bald Eagle Search(BES).The BES synchronization was used to optimize SVM parameters and complete feature selection.The advantages of optimization,feature selection and classification are integrated to achieve the optimal feature selection and fault diagnosis quickly and accurately.Experimental results showed that SCSO-SVM and BES synchronous optimization feature selection had higher diagnostic accuracy and computational efficiency than particle swarm optimization random forest(PSO-RF)and slime mold optimization kernel extreme learning machine(SMA-KELM).