Sparsity-based fault diagnosis methods have achieved great success. However, fault classification is still challenging because of neglected potential knowledge. This paper proposes a combined sparse representation deep learning (SR-DEEP) method for rolling bearing fault diagnosis. Firstly, the SR-DEEP method utilizes prior domain knowledge to establish a sparsity-based fault model. Then, based on this model, the corresponding regularization parameter regression networks are trained for different running states, whose core is to explore the latent relationship between the regularization parameters and running states. Subsequently, the performance of the fault classification is improved by embedding the trained regularization parameter regression networks into the sparse representation classification method. This strategy improves the adaptability of the sparse regularization parameter, further improving the performance of the fault classification method. Finally, the applicability of the SR-DEEP method for rolling bearing fault diagnosis is validated with the CWRU platform and QPZZ-II platform, demonstrating that SR-DEEP yields superior accuracies of 100% and 99.20% for diagnosing four and five running states, respectively. Comparative studies show that the SR-DEEP method outperforms four sparse representation methods and seven classical deep learning classification methods in terms of the classification performance.

To solve the problems of leakage and low detection accuracy in the detection of different types of diseases in rice leaves, this paper proposes the SPW YOLOv8 (SPD-Conv-RPSA-WIoU-YOLOv8) detection algorithm, which is built based on the YOLOv8 structure, and synergistically optimizes the feature extraction, attention, and loss design to improve the detection effect and generalization ability. First, a Sparse Depthwise Convolution (SPD-Conv) is used to replace the original convolution to enhance the extraction of low-resolution and small-target features, better adapting to the needs of small-scale and overlapping scene detection of leaves; Then, the RPSA attention module is proposed to optimize the design for the potential correlation between the channel and the spatial features, and the ReLU activation function is employed to introduce a learnable scaling factor to improve the efficiency of the resource allocation and significantly enhance the ability to detect complex backgrounds and tiny targets. Finally, the WIoU v3 loss function is adopted to optimize the detection convergence process and enhance detection accuracy, thereby improving the detection effect. Experiments demonstrate that SPW YOLOv8 achieves 90.2% mAP, 88.0% recall, and 90.2% detection accuracy in the task of detecting rice leaf diseases, which are 2.0%, 5.9%, and 3.7% higher than YOLOv8, respectively. It performs better relative to YOLOv5, YOLO-World, YOLOv11, RT-DETR, and SSD detection algorithms, showing higher detection capability and effectiveness, highlighting the significance of the SPW-YOLOv8 model in improving the detection precision and effectiveness in real agricultural scenarios.

Aiming at the uncertainty problem caused by the time-varying modeling parameters associated with ship speed in the course tracking control of underactuated surface vessels (USVs), this paper proposes a control algorithm based on the dynamic neural fuzzy model (DNFM). The DNFM simultaneously adjusts the structure and parameters during learning and fully approximates the inverse dynamics of ships. Online identification and modeling lays the model foundation for ship motion control. The trained DNFM, serving as an inverse controller, is connected in parallel with the fractional-order PI^λD^μ controller to be used for the tracking control of the ship’s course. Moreover, the weights of the model can be further adjusted during the course tracking. Taking the actual ship data of a 5446 TEU large container ship, simulation experiments are conducted, respectively, for course tracking, course tracking under wind and wave interferences, and comparison with five different controllers. This proposed controller can overcome the influence of the uncertainty of modeling parameters, tracking the desired course quickly and effectively.

Only using the point clouds collected by a Kinect sensor to perform 3D object detection is a very challenging task for indoor sense understanding. In this paper, we propose a novel point clouds based 3D object detection framework named Weighted VoteNet (W-VoteNet), which systematically enhances seed points, voting points, and proposal objects to varying degree, thereby enhancing the detection accuracy of indoor scene point clouds. Firstly, a new neighborhood feature enhancement module was designed to enhance the features of the seed point, which can better utilize the neighborhood feature of the seed point. Secondly, a weighted voting module is implemented to increase the accuracy of voting, allowing more prospect seed points to participate in voting. Finally, a new semantic relation reasoning module is proposed to get the semantic relationship features of the proposal object, which can further decrease false alarm rate. The three modules we proposed all contribute to more accurate voting and more effective proposals. The experimental results on two benchmark indoor 3D object detection datasets, SUN RGB-D and ScanNet V2, demonstrate the effectiveness of our approach. The source code is available at: https://github.com/Guo-JQ/W-VoteNet.