Bearing Defect Detection with Unsupervised Neural Networks

Bearings always suffer from surface defects, such as scratches, black spots, and pits. Those surface defects have great effects on the quality and service life of bearings. Therefore, the defect detection of the bearing has always been the focus of the bearing quality control. Deep learning has been successfully applied to the objection detection due to its excellent performance. However, it is difficult to realize automatic detection of bearing surface defects based on data-driven-based deep learning due to few samples data of bearing defects on the actual production line. Sample preprocessing algorithm based on normalized sample symmetry of bearing is adopted to greatly increase the number of samples. Two different convolutional neural networks, supervised networks and unsupervised networks, are tested separately for the bearing defect detection. The first experiment adopts the supervised networks, and ResNet neural networks are selected as the supervised networks in this experiment. The experiment result shows that the AUC of the model is 0.8567, which is low for the actual use. Also, the positive and negative samples should be labelled manually. To improve the AUC of the model and the flexibility of the samples labelling, a new unsupervised neural network based on autoencoder networks is proposed. Gradients of the unlabeled data are used as labels, and autoencoder networks are created with U-net to predict the output. In the second experiment, positive samples of the supervised experiment are used as the training set. The experiment of the unsupervised neural networks shows that the AUC of the model is 0.9721. In this experiment, the AUC is higher than the first experiment, but the positive samples must be selected. To overcome this shortage, the dataset of the third experiment is the same as the supervised experiment, where all the positive and negative samples are mixed together, which means that there is no need to label the samples. This experiment shows that the AUC of the model is 0.9623. Although the AUC is slightly lower than that of the second experiment, the AUC is high enough for actual use. The experiment results demonstrate the feasibility and superiority of the proposed unsupervised networks.

Ball bearing fault detection via multi-sensor data fusion with accelerometer and microphone

Early detection of defects in bearings is essential to avoid the complete failure of machinery and the associated costs. This study presents a novel method for fault diagnosis of bearings using sensor fusion with a microphone and an accelerometer. The system has five modules, namely data acquisition, signal processing, feature extraction, classification and decision-making. A test-rig is designed to collect acoustic and vibration signals. Then, for each signal, indices are calculated in the time and frequency domains. After using principal component analysis (PCA) for feature extraction, the k-nearest neighbours (kNN) method is used in the classification module. Finally, a decision on the kind of fault and its size is made based on the decision fusion module. The aim of this study is to propose a fusion method to improve the effectiveness and reliability of bearing defect diagnosis compared to what can be achieved with vibration or acoustic measurements alone. The results obtained from this preliminary study show that condition monitoring using the accelerometer is the more effective technique for determining the type of fault, while the microphone is effective for classifying the size of fault. Experimental results also confirm these findings.

Bearing Defect Classification Algorithm Based on Autoencoder Neural Network

The postproduction defect classification and detection of bearings still relies on manual detection, which is time-consuming and tedious. To address this, we propose a bearing defect classification network based on an autoencoder to enhance the efficiency and accuracy of bearing defect detection. An improved autoencoder is used to reduce dimension feature extraction and reduce large-scale images to small-scale images through encoder dimensional reduction. Defect classification is completed by feeding the extracted features into a convolutional classification network. Comparative experiments show that the neural network can effectively complete feature selection and substantially improve classification accuracy while avoiding the laborious algorithm of the conventional method.

Demodulation of Vibration Signal Based on Envelope-Kurtogram for Ball Bearing Fault Detection

Rolling element bearings often suffer damage due to harsh operating and environmental conditions. The method commonly used in detecting faults in a bearing is envelope analysis. However, this method requires setting the central frequency and the correct bandwidth - which corresponds to the resonance frequency of the bearing - for signal demodulation to be effective. This study proposes a kurtogram to determine the correct central frequency and bandwidth to obtain the frequency band with the highest impulse content or the highest kurtosis value. Analysis envelope is applied to the filtered vibration signal using the central frequency and bandwidth parameters obtained from the kurtogram. The results showed that the envelope-kurtogram method is effective for faulty bearing detection as shown in the envelope spectrum where the peaks coincide with the bearing defect characteristic frequency (BPFO) with high accuracy. Likewise, it can be observed several BPFO harmonics which provide information on the level of bearing fault.