Search

Rolling bearings in production practice usually serve in a healthy state. Some fault state labels are scarce or even no labels, resulting in unbalanced data categories. Meanwhile, frequent working condition switching results in significant differences in data distribution among working conditions, and labeled data in some working states cannot be fully utilized. To deal with the challenge of low fault identification accuracy caused by these practical factors, this paper proposed a novel adversarial unsupervised subdomain adaption multi-channel deep convolutional network (ASMDCN). Firstly, a parallel three-channel depth feature extraction module is built, and a multi-scale convolution kernel is used to fully extract the rich features of vibration signals under various working conditions. Secondly, a novel loss function is designed to adequately consider the classification difficulty of samples and the degree of class imbalance. Finally, the adversarial training strategy is used to force the feature extractor to extract the domain invariant features, and the Local Maximum Mean discrepancy (LMMD) is used to align the global and related subdomains of the source and target domains. The experimental results show that the designed feature extraction can fully extract the domain-invariant features of the rolling bearings under different working conditions. Under the proposed objective function optimization, the network model can fully align the features of multi-source and single-target domain under unbalanced data and has strong generalization performance.

This paper studies a class of quasi-cyclic LDPC (QC-LDPC) codes, i.e., Tanner (3, 23)-regular QC-LDPC codes of code length $23p$ with $p$ being a prime and $p \equiv 1 (\mathrm {mod} 69)$ . We first analyze the cycle structure of Tanner (3, 23)-regular QC-LDPC codes, and divide their cycles of lengths 4, 6, 8, and 10 into five equivalent types. We propose the sufficient and necessary condition for the existence of these five types of cycles, i.e., the polynomial equations in a 69th unit root of the prime field $\mathbb {F}_{p}$ . We check the existence of solutions for such polynomial equations by using the Euclidean division algorithm and obtain the candidate girth values of Tanner (3, 23)-regular QC-LDPC codes. We summarize the results and determine the girth distribution of Tanner (3, 23)-regular QC-LDPC codes.

With the advancement of the wood processing industry, the demand for the detection of surface defects in wood has become increasingly urgent. The application of automated production technology has enhanced the efficiency and precision of wood processing, which can significantly impact product quality and competitiveness. However, current methods for detecting surface defects in wood suffer from issues such as low detection accuracy, high computational complexity, and poor real-time performance. In response to these challenges, this paper proposes a high-precision, lightweight, real-time wood surface defect detection method based on YOLO(GBCD-YOLO) model. Firstly, the Ghost Bottleneck is introduced to improve the computational efficiency and inference speed of deep neural networks. Furthermore, the BiFormer is incorporated in the neck to enhance the performance of natural language processing tasks. Simultaneously, CARAFE is utilized as an upsampling replacement to enhance perceptual and capture abilities for details. In addition, the Dynamic Head is introduced to enhance the method’s flexibility and generalization ability, and the loss function is replaced with complete intersection over union (CIoU). The proposed method was evaluated using an optimized dataset and the YOLOv5s model was chosen as the baseline. The experimental results show that compared with the original YOLOv5s, the mAP (0.5) has been improved by 13.45%, reaching 88.72%. The mAP (0.5:0.95) increased by 11.95%, and FPS increased by 6.25%. In addition, the parameter of the improved model has been reduced by 15.49%. These results indicate that the proposed GBCD-YOLO improves the real-time detection performance of wood surface defects.

The increasing integration of distributed household photovoltaics (PVs) and electric vehicles (EVs) may further aggravate voltage violations and unbalance of low-voltage distribution networks (LVDNs). DC distribution networks can increase the accommodation of PVs and EVs and mitigate mutilple power quality problems by the flexible power regulation capability of voltage source converters. This paper proposes schemes to establish hybrid AC/DC LVDNs considering the conversion of the existing three-phase four-wire low-voltage AC systems to DC operation. The characteristics and DC conversion constraints of typical LVDNs are analyzed. In addition, converter configurations for typical LVDNs are proposed based on the three-phase four-wire characteristics and quantitative analysis of various DC configurations. Moreover, an optimal planning method of hybrid AC/DC LVDNs is proposed, which is modeled as a bi-level programming model considering the annual investments and three-phase unbalance. Simulations are conducted to verify the effectiveness of the proposed optimal planning method. Simulation results show that the proposed optimal planning method can increase the integration of PVs while simultaneously reducing issues related to voltage violation and unbalance.

This study explores the consensus of multi-agent systems in the presence of ambient noise using intermittent Brownian noise stabilization. Firstly, the research provides a mathematical explanation of multi-agent systems and intermittent stochastic noise, and establishes a sufficient condition for multi-agent systems to achieve consensus using intermittent stochastic noise control input. Secondly, a consensus criterion is proposed for a class of multi-agent systems that are affected by ambient noise. Finally, the simulation results demonstrate that the intermittent stochastic noise stabilization technique can facilitate the establishment of consensus in multi-agent systems.

Accurately classifying tree species in urban forests remains a major challenge due to the limitations of existing methods. While hyperspectral imaging is capable of capturing detailed spectral information, it struggles with the issue of ‘similar reflectance curves for different species’. On the other hand, the intensity frequency curve also has the problem of ‘similar intensity frequency curves for different species’. To overcome these challenges, a new method is proposed in this study. The method combines UAV LiDAR intensity frequency features (IF) and hyperspectral features (HF) to accurately identify tree species, abbreviated as HICM. The results of the study demonstrate that HICM can accurately identify 16 common urban tree species with an overall accuracy (OA) of 92.0% and kappa of 91.2%. Compared with using only IF or HF for identification, HICM improves OA by 16.9% and 12%, and kappa by 21.6% and 10.1%, respectively. This shows that HF and IF have complementarity, and the combination of these two features effectively addresses the problems of “same species different frequency” and “same species different spectrum”.

NASA's (National Aeronautics and Space Administration) ICESat-2 with a Photon Counting LiDAR (Light Detection And Ranging) Sensor sensitively detects signal photons at high speed with an advanced detection system called the Advanced Topographic Laser Altimeter System (ATLAS). However, the sensor also extracts a large amount of background photon noise coming from the atmosphere, ground, sun, or other radiation. This condition is particularly evident in forest areas. This study proposes an automatic machine learning approach to utilize data for forestry applications to improve data availability compared to NASA's official product. Our method uses only a very limited number (10%) of sample points for training, ensuring operational efficiency and training accuracy. We conclude that the integrated learning performance generally outperforms single models, and the mean F1 score of all tests is approximately 0.9. The mean F1 score of the Stacked Ensembles model is 0.957 ahead of the other models. The top three variables used in training models are kNNdist5, kNNdist10, and h. These three variables could explain 51.6% of the components of the models. Over the regions tested, the proposed method could improve the proportion of signals correctly identified by 6.4%, 12.2%, 2.7%, 9.3%, and 1.4% in five datasets. The model performs better in low signal-to-noise (SNR) datasets less than 7.5. The method would be largely unaffected by differences in topography, noise distribution, and SNR. The classifiers could correct misclassified labels in ATL08 products and show good stability in different conditions.