To address challenges in the detection of wind turbine blade damage images, characterized by complex backgrounds and multiscale feature distribution, we propose a method based on an enhanced YOLOV8 model. Our approach focuses on three key aspects: First, we enhance the extraction of small target features by integrating the CBAM attention mechanism into the backbone network. Second, the feature fusion process is refined using the Weighted Bidirectional Feature Pyramid Network (BiFPN) to replace the path aggregation network (PANet). This modification prioritizes small target features within the deep features and facilitates the fusion of multiscale features. Lastly, we improve the loss function from CIoU to EIoU, enhancing sensitivity to small targets and the perturbation resistance of bounding boxes, thereby reducing the gap between computed predictions and real values. Experimental results demonstrate that compared with the YOLOV8 model, the CBAM-BiFPN-YOLOV8 model exhibits improvements of 1.6%, 1.0%, 1.4%, and 1.1% in precision rate, recall rate, mAP@0.5, and mAP@0.5:.95, respectively. This enhanced model achieves substantial performance improvements comprehensively, demonstrating the feasibility and effectiveness of our proposed enhancements at a lower computational cost.

This article introduces a novel framework for diagnosing faults in rolling bearings. The framework combines digital twin data, transfer learning theory, and an enhanced ConvNext deep learning network model. Its purpose is to address the challenges posed by the limited actual fault data density and inadequate result accuracy in existing research on the detection of rolling bearing faults in rotating mechanical equipment. To begin with, the operational rolling bearing is represented in the digital realm through the utilization of a digital twin model. The simulation data produced by this twin model replace traditional experimental data, effectively creating a substantial volume of well-balanced simulated datasets. Next, improvements are made to the ConvNext network by incorporating an unparameterized attention module called the Similarity Attention Module (SimAM) and an efficient channel attention feature referred to as the Efficient Channel Attention Network (ECA). These enhancements serve to augment the network’s capability for extracting features. Subsequently, the enhanced network model is trained using the source domain dataset. Simultaneously, the trained model is transferred to the target domain bearing using transfer learning techniques. This transfer learning process enables the accurate fault diagnosis of the main bearing to be achieved. Finally, the proposed method’s feasibility is validated, and a comparative analysis is conducted in comparison with similar approaches. The comparative study demonstrates that the proposed method effectively addresses the issue of low mechanical equipment fault data density, leading to improved accuracy in fault detection and classification, along with a certain level of robustness.

During the operation process of a gearbox, the vibration signals can reflect the dynamic states of the gearbox. The feature extraction of the vibration signal will directly influence the accuracy and effectiveness of fault diagnosis. One major challenge associated with the extraction process is the mode mixing, especially under such circumstance of intensive frequency. A novel fault diagnosis method based on frequency-modulated empirical mode decomposition is proposed in this paper. Firstly, several stationary intrinsic mode functions can be obtained after the initial vibration signal is processed using frequency-modulated empirical mode decomposition method. Using the method, the vibration signal feature can be extracted in unworkable region of the empirical mode decomposition. The method has the ability to separate such close frequency components, which overcomes the major drawback of the conventional methods. Numerical simulation results showed the validity of the developed signal processing method. Secondly, energy entropy was calculated to reflect the changes in vibration signals in relation to faults. At last, the energy distribution could serve as eigenvector of support vector machine to recognize the dynamic state and fault type of the gearbox. The analysis results from the gearbox signals demonstrate the effectiveness and veracity of the diagnosis approach.

In this paper, a novel model for fault detection of rolling bearing is proposed. It is based on a high-performance support vector machine (SVM) that is developed with a multifeature fusion and self-regulating particle swarm optimization (SRPSO). The fundamental of multikernel least square support vector machine (MK-LS-SVM) is overviewed to identify a classifier that allows multidimension features from empirical mode decomposition (EMD) to be fused with high generalization property. Then the multidimension parameters of the MK-LS-SVM are configured by the SRPSO for further performance improvement. Finally, the proposed model is evaluated through experiments and comparative studies. The results prove its effectiveness in detecting and classifying bearing faults.

In recent years, the Broad Learning System (BLS) has been acknowledged for its potential to revolutionize traditional artificial intelligence methods due to its short training time, strong interpretability, and simple structure. In the evolution of BLS, Prof. C. L. Philip Chen’s team introduced the Fuzzy Broad Learning System (FBLS) by replacing the feature nodes of BLS with fuzzy subsystems, thereby further reducing the training time. However, the traditional FBLS, with its straightforward structure, falls short in achieving higher fault diagnosis accuracy when handling raw vibration signals. This paper presents a bearing fault diagnosis approach employing multi-domain feature selection and the fuzzy broad learning system (MS-FBLS), aiming to enhance the diagnostic accuracy of FBLS through multi-domain feature selection. Primarily, a set of 49 features spanning time domain, frequency domain, time-frequency domain, and entropy values is extracted from the original vibrational signals. This combination builds a 49-dimensional multidomain feature set that exploits the information behind the input data as much as possible, thus compensating for the lack of feature extraction capability in FBLS. Afterward, the Random Forest algorithm assesses the significance of all features, leading to a reordering of the multidomain feature set based on their respective importance levels. Ultimately, the reorganized multidomain feature set is then fed into the FBLS, enabling the identification of various failure states within the bearing. The experimental validation conducted on the rolling bearing fault simulation test bed showcased that, in comparison to the traditional FBLS, the MS-FBLS method not only elevates diagnostic accuracy by 23.46%, but also substantially enhances diagnostic speed. These results serve as comprehensive evidence affirming the effectiveness of the method.

Due to the numerous disadvantages that come with having anchors in the detection process, a lot of researchers have been concentrating on the design of object detectors that do not rely on anchors. In this work, we use anchor-free object detectors in the field of computer vision for surface defect detection. First, we constructed a surface defect detection dataset about real wind turbine blades, which was supplemented with several methods due to the lack of natural data. Next, we used a number of popular anchor-free detectors (CenterNet, FCOS, YOLOX-S, and YOLOV8-S) to detect surface defects in this blade dataset. After experimental comparison, YOLOV8-S demonstrated the best detection performance, with a high accuracy (79.55%) and a short detection speed (9.52 fps). All the upcoming experiments are predicated on it. Third, we examined how the attention mechanism added to various YOLOV8-S model positions affected the two datasets—our blade dataset and the NEU dataset—and discovered that the insertion methods on the two datasets are the same when focusing on comprehensive performance. Lastly, we carried out a significant amount of experimental comparisons.