Table 1 Examples of bearing fault detection methods.
From: An efficient bearing fault detection strategy based on a hybrid machine learning technique
Refs. | Model-based | Signal-based | Data-driven | Hybrid | Discussion |
|---|---|---|---|---|---|
✓ | A quantitative electrical model that uses stator current to estimate mutual inductance variations is proposed. These variations are then utilized to reconstruct the airgap displacement profile. By applying Fourier series analysis, the airgap variation can be explicitly modeled to identify the bearing fault severity | ||||
✓ | A method based on Gaussian mixture model for selecting fault frequency bands is employed to extract frequency bands directly related to bearing fault harmonics. Statistical features are extracted from the refined signals and classified using the weighted K-Nearest Neighbors (KNN) Algorithm | ||||
✓ | A hybrid time–frequency analysis technique to diagnose bearing faults is developed. In this technique, a spectrum analysis is conducted to identify fault characteristic frequencies | ||||
✓ | In this study, a method based on time-domain sparsity analysis is introduced to detect faults and determine their precise locations, where fault signals are represented as periodic pulses | ||||
✓ | A bearing fault detection strategy in high-speed rail systems is presented. Vibration signals collected from wireless data acquisition system are used to diagnose bearing faults through a multiple correlation analysis method | ||||
✓ | The approach presented in this study integrates the hyperparameter-tuned SVM classifiers with advanced feature selection methods based on genetic algorithm and fisher Score for a robust bearing fault detection | ||||
✓ | The study presents a robust CNN-based model for diagnosing faults in three-phase induction motors. In this model the measured motor data are converted into d-q and image data. The accuracy of the obtained model is tested under a wide range of operating conditions | ||||
✓ | A DTL-based framework for to detect bearing faults in induction machines is presented, where the ResNetV2 model is employed for efficient fault diagnosis model | ||||
✓ | This study presents a fault detection framework for milling machines by integrating high-frequency acoustic emission signal analysis with a DL-based approach. The primary objective is to achieve precise and efficient fault diagnosis in critical machine components, including cutting tools, gears, and bearings. The proposed framework incorporates a VGG16-based CNN for spatial feature extraction, coupled with a bidirectional LSTM network to capture temporal dependencies. Furthermore, a genetic algorithm is utilized for optimal feature selection | ||||
✓ | To address the limitations of traditional diagnostic methods, such as accuracy and noise issues, an approach combining a variational mode decomposition, kernel extreme LM, and metaheuristic optimization algorithms is introduced. The validation of the obtained models is evaluated through experimental tests | ||||
✓ | A hybrid DL model is introduced to identify and classify anomalies in rolling bearings through temporal and spatial feature extraction. Temporal features are captured using a two-layer LSTM network, while spatial features are extracted through a dynamic snake convolution mechanism. The combined features are further enhanced using to ResNet. The proposed approach is compared with other intelligent methods through experimental tests |
