Bearing defect identification by swarm decomposition considering permutation entropy measure and opposition-based slime mould algorithm. (June 2021)

• Record Type: Journal Article
• Authors: Vashishtha, Govind; Chauhan, Sumika; Singh, Manpreet; Kumar, Rajesh
• Is Part Of: Measurement. Volume 178(2021)
• Publication Date: 2021-06
• Subjects: Swarm decomposition (SWD) -- Permutation entropy (PE) -- Slime mould algorithm (SMA) -- Opposition-based learning -- Extreme learning machine (ELM); Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2021.109389
• Languages: English
• ISSNs: 0263-2241

Abstract

Highlights: A reliable method with 100% accuracy is proposed for taper roller bearing defect identification. Swarm decomposition method is applied to decompose the vibration signal into different modes. Permutation entropy is considered as a measurement index to select the prominent mode. Filter-based relief algorithm is applied to reduce the redundancy of the obtained feature matrix. ELM classifier parameters are chosen adaptively by a novel opposition-based slime mould algorithm. Abstract: An intelligent defect identification scheme has been proposed to identify the taper roller bearing defects through the extreme learning machine (ELM) model. The raw vibration signal from the bearing test rig is decomposed into different modes by Swarm decomposition (SWD) method to remove noise. The permutation entropy (PE) is taken as a measurement index to select the prominent mode. Features sensitive to defect conditions are extracted from prominent mode using a filter-based relief algorithm. The ranking of fault features is done on the score values. An opposition-based slime mould algorithm is investigated for finding the optimal parameters (weight connecting the input layer with output layer; and biases in hidden neurons) of ELM. Using optimized ELM parameters, the classification model is built. Test data evaluate the fitness of the built ELM model. Both the training and testing accuracy are 100%, with a computation time of 0.0023 s.