Bearing Fault Detection by One‐Dimensional Convolutional Neural Networks

Abstract

Bearing faults are the biggest single source of motor failures. Artificial Neural Networks (ANNs) and other decision support systems are widely used for early detection of bearing faults. The typical decision support systems require feature extraction and classification as two distinct phases. Extracting fixed features each time may require a significant computational cost preventing their use in realtime applications. Furthermore, the selected features for the classification phase may not represent the most optimal choice. In this paper, the use of 1D Convolutional Neural Networks (CNNs) is proposed for a fast and accurate bearing fault detection system. The feature extraction and classification phases of the bearing fault detection are combined into a single learning body with the implementation of 1D CNN. The raw vibration data (signal) is fed into the proposed system as input eliminating the need for running a separate feature extraction algorithm each time vibration data is analyzed for classification. Implementation of 1D CNNs results in more efficient systems in terms of computational complexity. The classification performance of the proposed system with real bearing data demonstrates that the reduced computational complexity is achieved without a compromise in fault detection accuracy.

Fault Detection and Severity Identification of Ball Bearings by Online Condition Monitoring

“…2) Damage in the inner ring: 9 scenarios corresponding to 9 levels of damage in the inner ring: 0. 35 Fig. 6.…”

“…3) The third dataset 3 consists of vibration signals corresponding to a severe outer ring defect. It is important to note that unlike the previous CNN-based bearing fault detection methods in [35], [36] which require the corresponding set of data for each damage level to train the classifier(s), the proposed method requires signals only from the severe inner-and outer-ring damage scenarios. The choice of the defect size in the severe inner and outer ring damage scenarios is up to the designer.…”

“…They have displayed a fast and accurate performance in several real-time monitoring applications such as classification of electrocardiogram (ECG) beats [19], structural health monitoring [18], [31]- [33], and motor fault detection [34]. Additionally, two recent studies have utilized 1D CNNs for damage detection in bearings [35], [36]. In the first study [35], a 1D CNN was trained to classify the measured vibration signals into two classes, healthy or damaged.…”

“…Additionally, two recent studies have utilized 1D CNNs for damage detection in bearings [35], [36]. In the first study [35], a 1D CNN was trained to classify the measured vibration signals into two classes, healthy or damaged. While this 1D CNN displayed a reasonable damage detection performance, it was not able to identify the location and severity of the detected bearing damage.…”

A Novel Image Feature for the Remaining Useful Lifetime Prediction of Bearings Based on Continuous Wavelet Transform and Convolutional Neural Network

“…The features are inputted into long short-term memory (LSTM) neural networks to construct the HI. Eren [23] also directly inputted the raw signal to a one-dimensional (1-D) CNN, which incorporates feature extraction and fault detection into a single algorithm. Although the CNN was originally designed to process two-dimensional (2-D) visual signals, most of the research in the prognostics field has been used to process 1-D vibration signals.…”