Abstract:In view of the frequent failures occurring in rolling bearings, the strong background noise present in signals, weak features, and difficulties associated with extracting fault characteristics, a method of enhancing and diagnosing rolling bearing faults based on coarse-grained lattice features (CGLFs) is proposed. First, the vibrational signals of bearings are subjected to adaptive filtering to eliminate background noise. Second, frequency-domain transformation is performed, and a coarse-grained approach is used to continuously segment the spectrum. Within each segment, amplitude-enhancement operations are executed, transforming the data into a CGLF graph that enhances fault characteristics. This graph is then fed into a Swin Transformer-based pattern-recognition network. Third and finally, a high-precision fault diagnosis model is constructed using fully connected layers and Softmax, enabling the diagnosis of bearing faults. The fault recognition accuracy reaches 98.30% and 98.50% with public datasets and laboratory data, respectively, thereby validating the feasibility and effectiveness of the proposed method. This research offers an efficient and feasible fault diagnosis approach for rolling bearings.

What problem does this paper attempt to address?

This paper attempts to solve several key problems in fault feature extraction and diagnosis of rolling bearings: 1. **Strong background noise**: The vibration signals generated during the operation of rolling bearings often contain strong background noise, which makes it difficult to effectively extract fault features. 2. **Weak fault features**: The fault features of rolling bearings are usually relatively weak. Especially in complex working environments, these features are easily masked by various interference signals. 3. **Difficult feature extraction**: Traditional fault feature extraction methods have limitations in dealing with the above problems. For example, the short - time Fourier transform (STFT) has restrictions in satisfying the Heisenberg uncertainty principle, and there are problems such as the selection of wavelet transform basis functions and the mode - mixing problem of empirical mode decomposition (EMD). To address these problems, the paper proposes a method for enhancing and diagnosing rolling - bearing fault features based on coarse - grained lattice features (CGLFs). The specific steps are as follows: 1. **Adaptive band - pass filtering**: First, perform adaptive band - pass filtering on the vibration signal of the bearing to eliminate background noise. The center - frequency bandwidth of the adaptive filter is determined by the spectral kurtosis (SK) method, which can identify the frequency band of transient impact signals. 2. **Frequency - domain transformation**: Perform a fast Fourier transform (FFT) on the filtered signal to convert it from the time domain to the frequency domain and obtain the frequency spectrum. 3. **Coarse - grained segmentation and feature enhancement**: Continuously segment the frequency spectrum and perform an amplitude - enhancement operation within each segment to convert the data into a CGLF graph, thereby enhancing the fault features. 4. **Pattern - recognition network**: Input the CGLF graph into a pattern - recognition network based on Swin Transformer, and construct a high - precision fault - diagnosis model through fully - connected layers and Softmax to achieve the diagnosis of bearing faults. The experimental results show that the fault - recognition accuracies of this method on public data sets and laboratory data reach 98.30% and 98.50% respectively, verifying the effectiveness and feasibility of this method.

Bearing-Fault-Feature Enhancement and Diagnosis Based on Coarse-Grained Lattice Features

A Method of Bearing Fault Feature Extraction Based on Improved Wavelet Packet and Hilbert Analysis

Optimal Time Frequency Fusion Symmetric Dot Pattern Bearing Fault Feature Enhancement and Diagnosis

Improving rolling bearing online fault diagnostic performance based on multi-dimensional characteristics

A Lightweight Model for Bearing Fault Diagnosis Based on Gramian Angular Field and Coordinate Attention

Fault Diagnosis Method for Rolling Element Bearing Based on Enhanced Cross-Correlation Spectrum

Fuzzy Broad Learning System Combined with Feature-Engineering-Based Fault Diagnosis for Bearings

Research on rolling bearing fault diagnosis based on multi-dimensional feature extraction and evidence fusion theory

A rolling bearing fault diagnosis method based on a convolutional neural network with frequency attention mechanism

A New Fault Diagnosis of Rolling Bearing on FFT Image Coding and L-CNN

An intelligent diagnosis method using fault feature regions for untrained compound faults of rolling bearings

Rolling bearing fault diagnosis method using time-frequency information integration and multi-scale TransFusion network

A novel fault diagnosis approach of rolling bearing using intrinsic feature extraction and CBAM-enhanced InceptionNet

Feature extraction based on hierarchical improved envelope spectrum entropy for rolling bearing fault diagnosis

A sequential feature extraction method based on discrete wavelet transform, phase space reconstruction, and singular value decomposition and an improved extreme learning machine for rolling bearing fault diagnosis

Intelligent fault diagnosis of rolling bearings based on continuous wavelet transform-multiscale feature fusion and improved channel attention mechanism

A bearing fault diagnosis method with improved symplectic geometry mode decomposition and feature selection

A deep feature enhanced reinforcement learning method for rolling bearing fault diagnosis

Bearing Fault Feature Extraction and Fault Diagnosis Method Based on Feature Fusion

An intelligent fault diagnosis method for rolling bearings based on feature transfer with improved DenseNet and joint distribution adaptation

Nonlinear convolutional sparse filtering with Gaussian fitting for rolling bearings compound fault diagnosis