A Novel Deep Learning Model for the Detection and Identification of Rolling Element-Bearing Faults

Abstract

Real-time acquisition of large amounts of machine operating data is now increasingly common due to recent advances in Industry 4.0 technologies. A key benefit to factory operators of this large scale data acquisition is in the ability to perform real-time condition monitoring and early-stage fault detection and diagnosis on industrial machinery-with the potential to reduce machine down-time and thus operating costs. The main contribution of this work is the development of an intelligent fault diagnosis method capable of operating on these real-time data streams to provide early detection of developing problems under variable operating conditions. We propose a novel dual-path recurrent neural network with a wide first kernel and deep convolutional neural network pathway (RNN-WDCNN) capable of operating on raw temporal signals such as vibration data to diagnose rolling element bearing faults in data acquired from electromechanical drive systems. RNN-WDCNN combines elements of recurrent neural networks (RNNs) and convolutional neural networks (CNNs) to capture distant dependencies in time series data and suppress high-frequency noise in the input signals. Experimental results on the benchmark Case Western Reserve University (CWRU) bearing fault dataset show RNN-WDCNN outperforms current state-of-the-art methods in both domain adaptation and noise rejection tasks.

Keywords: artificial intelligence; condition monitoring; deep learning; fault diagnosis.

Figure 1

Simple 1D convolutional neural network (CNN) architecture with two convolutional layers.

Figure 2

Basic Long-Short Term Memory (LSTM) architecture.

Figure 3

The proposed recurrent neural network with wide first kernel and deep convolutional path (RNN-WDCNN).

Figure 4

Finding the optimal learning rate range to use in the cyclical learning rate strategy.

Figure 5

The Case Western Reserve University Bearing Center test apparatus (original image taken from [46]).

Figure 6

Sliding window data augmentation.

Figure 7

Domain adaptation accuracy comparison between the different recurrent pathways in recurrent neural network with a wide first kernel and deep convolutional neural network pathway (RNN-WDCNN) for scenario 1 (best values are highlighted in bold).

Figure 8

Domain adaptation accuracy comparison between RNN-WDCNN and WDCNN, state-of-the-art deep learning-based models (SRDCNN), and FFT-based methods for scenario 1 (best values are highlighted in bold).

Figure 9

Confusion plots for the RNN-WDCNN model on the 6 different domain adaptation cases considered in scenario 1.

Figure 10

Domain adaptation accuracy comparison between RNN-WDCNN and WDCNN, SRDCNN, and FFT-based methods for scenario 2 (best values are highlighted in bold).

Figure 11

Confusion plots for the RNN-WDCNN model on the two of the domain adaptation cases considered in scenario 2.

Figure 12

Noise rejection accuracy comparison between the different recurrent pathways in RNN-WDCNN (best values are highlighted in bold).

Figure 13

Noise rejection accuracy comparison between RNN-WDCNN and WDCNN, SRDCNN, and FFT based methods (best values are highlighted in bold).

Figure 14

Confusion plot for the RNN-WDCNN model in the case of severe additive noise (SNR −4 dB).