Early Bearing Fault Diagnosis of Rotating Machinery by 1D Self-Organized Operational Neural Networks

Turker Ince,Serkan Kiranyaz,Onur Avci,Levent Eren,Junaid Malik,Moncef Gabbouj,Ozer Can Devecioglu

doi:10.1109/access.2021.3117603

Turker Ince, Serkan Kiranyaz + Show 5 more

Open Access

https://doi.org/10.1109/access.2021.3117603

Copy DOI

Abstract

Preventive maintenance of modern electric rotating machinery (RM) is critical for ensuring reliable operation, preventing unpredicted breakdowns and avoiding costly repairs. Recently many studies investigated machine learning monitoring methods especially based on Deep Learning networks focusing mostly on detecting bearing faults; however, none of them addressed bearing fault severity classification for early fault diagnosis with high enough accuracy. 1D Convolutional Neural Networks (CNNs) have indeed achieved good performance for detecting RM bearing faults from raw vibration and current signals but did not classify fault severity. Furthermore, recent studies have demonstrated the limitation in terms of learning capability of conventional CNNs attributed to the basic underlying linear neuron model. Recently, Operational Neural Networks (ONNs) were proposed to enhance the learning capability of CNN by introducing non-linear neuron models and further heterogeneity in the network configuration. In this study, we propose 1D Self-organized ONNs (Self-ONNs) with generative neurons for bearing fault severity classification and providing continuous condition monitoring. Experimental results over the benchmark NSF/IMS bearing vibration dataset using both x- and y-axis vibration signals for inner race and rolling element faults demonstrate that the proposed 1D Self-ONNs achieve significant performance gap against the state-of-the-art (1D CNNs) with similar computational complexity.

Highlights

Electrical rotating elements and machines are widely used in various industrial and commercial applications on account of their reliability and efficiency
In our previous study in [18], we proposed for the first time, a compact 1D Convolutional Neural Networks (CNNs) for real-time motor bearing fault classification and achieved the state-of-the-art performance demonstrated over the benchmark bearing fault dataset using raw current signals
For computational complexity analysis of the proposed classifier model, we provide the formulation for calculating the total number of multiply-accumulate operations (MACs) and the total number of parameters (PARs) of a generative neuron inside a 1D Self-Operational Neural Networks (ONNs)

Summary

Introduction

Electrical rotating elements and machines are widely used in various industrial and commercial applications on account of their reliability and efficiency. Effective condition monitoring and early fault detection and diagnosis of RM is critical for maintaining reliable operation, avoiding unpredicted breakdowns, reducing operating costs and improving productivity. The signal-based methods are based on typical signal analysis such as vibration, motor current, speed, and temperature, using signal processing methods such as fast Fourier transform [3], spectral estimation [4], time-frequency [5] and wavelet transformation [6], sequence analysis [7] and scale-invariant feature transform (SIFT) [8]. The signal-based FDD methods, similar to the model-based FDD, require a priori knowledge of signal patterns and often advanced signal processing tools with increased computational complexity need to be applied effective fault

Objectives

Methods

Results

Conclusion