As the backbone of China’s energy transportation network, train group operation has become the key path for Guoneng Railway to improve transport efficiency and overcome issues such as low efficiency of traditional transport modes and localized line capacity saturation. This study designs a group operation scheme to address the traffic organization bottlenecks of core lines such as the Baoshen South and Shenshuo Lines and optimizes capacity under the 500 million tons transportation volume target. First, it analyzes the adaptability and necessity of train group operation in terms of line layout, traffic flow distribution, and the current status of transportation organization. Second, it clarifies methods for determining key elements of the group operation scheme. Finally, it establishes a group operation scheme formulation model and proposes a corresponding method. Taking the theoretically generated daily transport volume of 1.28 million tons for the Shenshuo Line connecting to the Shuohuang Line under the 500 million-ton target transport volume as an example, results show that the group size of two 5000-ton trains on the Baoshen South Line and two 10000-ton trains on the Shenshuo Line, combined with the stop-based grouping and non-stop de-grouping mode, can reduce line capacity utilization by 10.7
With the rise of intelligent manufacturing, automated quality inspection has become a critical and non-negotiable component of industrial production lines. Hexagonal nuts, as fundamental mechanical parts, require stringent quality control, as even minor surface defects can compromise product safety and long-term performance. Addressing the inefficiency, high cost, and subjectivity of traditional manual inspection, we propose an enhanced object detection model based on YOLOv-5 s for the automated, real-time detection of hexagonal nut surface defects under challenging industrial conditions. Our model introduces three synergistic improvements: (1) the integration of the Convolutional Block Attention Module (CBAM) to refine the feature extraction backbone, enabling the network to focus on salient defect regions amidst noisy backgrounds; (2) the replacement of the standard feature fusion network with the Bidirectional Feature Pyramid Network (BiFPN) for more effective and weighted multi-scale feature aggregation; and (3) the adoption of the Focal-EIoU loss function to address the imbalance between difficult and easy samples and significantly improve bounding box regression accuracy. Through extensive ablation studies on a custom dataset of industrial hexagonal nut images, our proposed model achieves a mean Average Precision (mAP@0.5) of 93.3
The inherent sensitivity to initial conditions and unpredictability of long-term evolution in chaotic systems make them a core theoretical foundation for constructing highly secure confidential communication systems. This study innovatively proposes a cross-coupled chaotic mapping model. Based on this mapping, a pseudorandom number generator is constructed and successfully implemented and verified on the XILINX xc7z100ffg900-2 field-programmable gate array (FPGA) platform. Experimental results demonstrate that the generated pseudorandom sequences exhibit excellent statistical randomness, comprehensively passing the rigorous validation of the NIST 800-22 randomness test suite, with performance significantly surpassing traditional typical pseudorandom number generators. At the application level of communication systems, we further constructed a complete prototype of a synchronous communication system based on this chaotic generator. Through the feedback control matrix, the rapid synchronization of the sending and receiving chaotic systems was achieved, effectively verifying the feasibility and reliability of this scheme in real-time encrypted communication, and demonstrating broad prospects for engineering applications.
Driven by digitalization and intelligentization, indoor broadband has entered the gigabit era, with emerging latency-sensitive services (e.g., IPTV, cloud gaming) demanding higher bandwidth and QoS—making FTTR a key solution under the Fifth Generation Fixed Network (F5G) framework. This study designs a FTTR system based on the P2MP architecture, where master-slave gateways collaborate to simplify networks, cut costs, and support multi-service convergence. Key technologies include: OMCI-based enhanced OAM for full-lifecycle gateway management, reinforcement learning-integrated intelligent DBA for differentiated QoS, and chip-to-system low-power modules for dual-carbon goals. An AI-driven management scheme (LLDP-based topology discovery, ML-enabled fault prediction, SDN-based traffic scheduling) is also implemented to boost automation and reliability. Via a multi-dimensional performance evaluation (bandwidth, latency, energy efficiency, etc.) and optimization, the system achieves “Gigabit to the Room” transmission, reduces O M costs, and ensures stable service quality. Future work will expand FTTR to smart/digital homes, integrate advanced AI, and develop new business models for popularization and monetization.
Differential Evolution (DE) is a population-based evolutionary algorithm for global optimization, widely successful in both competitions and real-world engineering applications, particularly in complex optimization tasks. Recently, DE has been applied to high-complexity problems such as UAV path planning. However, traditional DE often suffers from premature convergence and stagnation due to reduced population diversity in high-dimensional settings. To overcome the aforementioned problems, an adaptive DE algorithm integrating dynamic perturbation and bidirectional dimension crossover is developed in this work. The contributions can be summarized as follows: first, the proposed hybrid adaptive parameter control mechanism separates the evolutionary process into two phases, aiming to generate more effective scaling factors and crossover probabilities; second, a dynamic perturbation mechanism for donor vectors is introduced, which adaptively tunes the perturbation intensity to maintain population diversity; and third, a bidirectional dimension crossover mechanism mitigates stagnation and maintains diversity. Experiments on CEC2017 test suits and a UAV path planning problem demonstrate the superior performance of the proposed algorithm.
