Combined classification models for bearing fault diagnosis with improved ICA and MFCC feature set. (November 2022)

• Record Type: Journal Article
• Authors: M, Azim Naz; R, Sarath
• Is Part Of: Advances in engineering software. Volume 173(2022)
• Publication Date: 2022-11
• Subjects: Fault diagnosis -- Improved ICA features -- Improved MFCC features -- LSTM -- SI-SSO algorithm; Computer-aided engineering -- Periodicals; Engineering -- Computer programs -- Periodicals; Engineering -- Software -- Periodicals; Periodicals; 620.0028553
• DOI: 10.1016/j.advengsoft.2022.103249
• Languages: English
• ISSNs: 0965-9978

Abstract

Highlights: Proposes a novel framework for fault diagnosis, where the LPSE, improved ICA and improved MFCC based features are extracted. Deploys hybrid classifiers including LSTM and ANN for fine diagnosis of bearing faults. Proposes a novel algorithm termed as self improved SSA for fine tuning the weights of LSTM and ANN. Abstract: Earlier analysis of faults in machinery is widely analyzed, for maintenance and cost savings. Bearing faults and motorized imbalances are the larger faults in machinery, particularly for the machinery of smaller sized -medium. As a result, these analyses are widely analyzed in the field of research. Rolling component bearings are extensively deployed in rotating machinery and, simultaneously, they are simply scratched, which are owing to harsh working conditions and environments. Consequently, rolling component bearings are significant to the safer function of motorized devices. The main aim of this research is to diagnose bearing faults using deep learning algorithms. The bearing fault of machinery is identified based on the features extracted and represented by using several monitoring techniques. Framing a new fault diagnosis model includes 2 major phases. Initially, LPSE improved ICA and improved MFCC-based features are extracted. These features are then subjected to LSTM and ANN for the fine diagnosis of bearing faults. For exact diagnosis, this work aims to optimize the weights of LSTM and ANN using the Self Improved Salp Swarm Optimization (SI-SSO) model. In the end, the development of the deployed method is confirmed regarding miscellaneous metrics.