Abstract
Aiming at the influence of mixed noise of bearing vibration signal on the extraction of useful information, a fault diagnosis optimize classifier based on multi-scale permutation entropy (MPE) and cuckoo search algorithm (CS) is proposed. Firstly, the MPE threshold method is adopted to select the appropriate variational mode decomposition algorithm (VMD) parameters, and then the signal is reconstructed by adding neutral white noise, and the reconstructed signal is decomposed by MPE-OVMD algorithm to obtain the optimal IMF component. Finally, the cuckoo search algorithm is used to optimize the global optimal solution of the support vector machine, thereby achieving the classification model of support vector machine with the best parameters. The analysis results of motor signals show that the method can eliminate the phenomena of mode aliasing and signal over-decomposition. An analytical comparison of the CSSVM classifier is carried out with the performance of the learners such as recall rate, ROC curve, AUC. The contrast experiment shows that the classification model can avoid misrecognition of the fault sample as the normal condition and maximum the optimal maintenance time of the equipment under the premise of ensuring the accuracy. The classifier model of the cuckoo optimization algorithm has better fitting accuracy than others such as the Grid Search algorithm (GS), Particle Swarm Optimization (PSO), Genetic Algorithm search (GA), and the ensemble fault recognition rate is as high as 90%.
Highlights
Nowadays, with the development of the industrial field scale, DC motors are increasingly required to be increasingly larger scale, with better integration and higher speed
The main conclusions are as follows: (1) In this paper, the positive and negative Gauss white noise signal reconstruction and addition method based on signal denoising preprocessing is adopted
(2) In view of the shortcomings of empirical mode decomposition (EMD), ensemble empirical mode decomposition (EEMD) and variational mode decomposition (VMD) methods, a method based on multi-scale permutation VMD algorithm for decomposing the measured signals is proposed
Summary
With the development of the industrial field scale, DC motors are increasingly required to be increasingly larger scale, with better integration and higher speed. Compared with other decomposition methods, it can accurately separate signals with a strong mathematical foundation and higher computing efficiency It has the characteristics of wiener filtering, which can effectively remove noise. In reference [26], BP neural network is used to train and identify the cavitation state of centrifugal pump This method has achieved certain effects, it requires hundreds of thousands of trainings, and the amount of data calculation is too large and the time is too long. Reference [27] utilizes particle optimization algorithm to optimize the SVM to classify rolling bearings This method can get the classification results more accurately, PSO algorithm cannot be widely applied in fault diagnosis of rolling bearings because it is easy to fall into local minimum. Compared with the models PSOSVM, GSSVM and GASVM, it is found that the MPE-OVMDCSSVM model has better overall performance in fault diagnosis of DC motor
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
