一种基于FCOS神经网络的小建筑物目标检测方法北大核心

提出了一种基于FCOS神经网络的小建筑物目标检测算法,针对FCOS模型在特征提取阶段提取到的小建筑物目标特征较少问题,引入多尺度检测和可变形卷积方式,加强网络对小建筑物目标的特征提取能力,并通过改进后的SGE注意力机制降低特征图中的干扰噪声权重。改进后的网络可以提取到更多的小建筑物目标特征,对环境干扰噪声的鲁棒性更强。在自己搭建的数据集上进行了实验测试,结果表明,在相同环境下网络改进后建筑物的整体检测准确率提升了1.7%,其中对小建筑物目标提升了3.6%,减少了小建筑物目标漏检、误检的问题。     

一种基于抽取LMS的距离旁瓣抑制方法

在噪声雷达中，传统相关处理方法的距离旁瓣受到时宽带宽积的限制，在有限相关处理时间内得到的距离旁瓣较高，会造成微弱目标被强目标、杂波旁瓣淹没的现象。提出一种基于抽取最小均方（Least Mean Square,LMS）滤波的噪声雷达旁瓣抑制方法，将LMS滤波器的系数作为距离压缩结果，从而获取较低的距离旁瓣。对该方法的性能进行了理论分析，并通过数字仿真验证了算法的有效性和理论分析的正确性。     

Molecular dynamics study of interfacial load transfer capability in amorphous SiOx films deposited on alumina surfaces

Effective adhesion between AlO_x and SiO_x is important for protective coatings and high-k films under extreme operating conditions. Here, we study the chemo-mechanical behavior of the AlO_x/SiO_x interface and its delamination mechanism using all-atom reactive molecular dynamics simulations. The structure of the interface is examined by the formation of bridge oxygen and the distribution of nanopores. The cleavage of ionic bonds during delamination and the resulting adhesion strength of the system are quantified using pull-out simulations. The results reveal the dependence of the nanopores and ionic bond formation on the oxide structure. The ionic bond density at the interface increases as the oxidation of the aluminum surface proceeds, which directly increases the adhesion strength with SiO_x. In particular, the global coordination distribution in the homogeneously grown oxide inhibits the formation of nanopores inside the aluminum substrate and contributes to extremely high adhesion strength. This reveals a fundamental relationship between physicochemical parameters and engineering mechanics for hetero-oxide structure design.     

Concentration fluctuation caused by reaction-diffusion coupling near catalytic active sites

The coupling of reaction and diffusion between neighboring active sites in the catalyst pore leads to the spatiotemporal fluctuation in component concentration, which is very important to catalyst performance and hence its optimal design. Molecular dynamics simulation with hard-sphere and pseudo-particle modeling has previously revealed the non-stochastic concentration fluctuation of the reactant/product near isolated active site due to such coupling, using a simple model reaction of A → B in 2D pores. The topic is further developed in this work by studying the concentration fluctuation due to such coupling between neighboring active sites in 3D pores. Two 3D pore models containing an isolated active site and two adjacent active sites were constructed, respectively. For the isolated site, the concentration fluctuation intensifies for larger pores, but the product yield decreases, and for a given pore size, the product yield reaches a peak at a certain reactant concentration. For two neighboring sites, their distance (d) is found to have little effect on the reaction, but significant to the diffusion. For the same reaction competing at both sites, larger d leads to more efficient diffusion and better overall performance. However, for sequential reactions at the two sites, higher overall performance presents at a smaller d. The results should be helpful to the catalyst design and reaction control in the relevant processes.     

基于改进灰狼优化算法的服务功能链映射算法

随着工业互联网、车联网、元宇宙等新型互联网应用的兴起，网络的低时延、可靠性、安全性、确定性等方面的需求正面临严峻挑战。采用网络功能虚拟化技术在虚拟网络部署过程中，存在服务功能链映射效率低与部署资源开销大等问题，联合考虑节点激活成本、实例化开销，以最小化平均部署网络成本为优化目标建立了整数线性规划模型，提出基于改进灰狼优化算法的服务功能链映射（improved grey wolf optimization based service function chain mapping，IMGWO-SFCM）算法。该算法在标准灰狼优化算法基础上添加了基于无环K最短路径（K shortest path，KSP）问题算法的映射方案搜索、映射方案编码以及基于反向学习与非线性收敛改进三大策略，较好地平衡了其全局搜索及局部搜索能力，实现服务功能链映射方案的快速确定。仿真结果显示，该算法在保证更高的服务功能链请求接受率下，相较于对比算法降低了11.86%的平均部署网络成本。     

An optimal configuration for hybrid SOFC,gas turbine,and Proton Exchange Membrane Electrolyzer using a developed Aquila Optimizer

In this research, a technical, economic and environmental analysis has been proposed to a Hybrid Solid Oxide Fuel Cell (SOFC) system-based hybrid system including biomass, gas turbine, and Proton Exchange Membrane Electrolyzer. A multi-objective optimization technique has been utilized to improve the overall product cost and the exergy effectiveness based on a developed version of Aquila Optimizer (DAO). The main idea of using the developed version is to improve the accuracy and the precision of the original Aquila optimizer. The system is then authenticated in terms of energy/exergy effectiveness, and energy-economic efficiency. The achievements indicate that employing the optimization algorithm for different configurations provided satisfying results for the system.     

Comprehensive numerical investigations on the thermal performance of heat recirculating micro combustor with pin fins for micro-thermal voltaic system applications

Aiming at improving the relatively low energy output and energy conversion efficiency of the micro-thermal voltaic (MTPV) system, an innovative heat recirculating micro combustor with pin fins is designed. The effects of pin fins arrangement, hydrogen/air equivalent ratio on the energy output and performance of CHMC, HMCP and HMCI are compared and investigated. The result shows that when the V_in is 6 m/s and Φ is 1.0, the emitter power of CHMC is 72.76W, and that of HCMP and HCMI micro combustor are 75.99W and 76.35W. and the emitter efficiency of CHMC, HCMP and HCMI is 41.93%, 43.26% and 44.01%. HMCI has better energy output capability compared with CHMC and HMCP. Even though, HMCI brings a higher pressure drop, it is within the acceptable range. When the V_in is 6 m/s, the pressure drop from the pin fins only accounts for 26.4% of the total pressure drop for HMCI. Through the study of equivalent ratio, it is found that HMCI has good adaptability in different equivalent ratio range. This work provides new ideas for the development of MTPV system in the future.     

Study on vapour dispersion and explosion from compressed hydrogen spill: Risk assessment on a hydrogen plant

Hydrogen produced from renewable resources is one of the cleanest fuels and could be used to store intermittent solar, wind and other energies. The main concern about using hydrogen is its hazards, such as high storage pressure, wide-range flammability, low mass density, and high diffusion. This study investigated the hazards of compressed hydrogen storage by developing a CFD model to understand the gas dispersion behaviour. The model was validated using the past experimental data and showed a good agreement, which could demonstrate the diffusion characteristics and gas stratification of a buoyant gas. A case study of an accidental release of compressed hydrogen from a storage tank was investigated to evaluate the risk of a hydrogen plant. A mathematical model of the jet spill was used to account for the choking effect from a high-pressure release to ensure the input velocity in CFD simulation is suitable for modelling gas dispersion using verified spatial and temporal scales, then the simulation results were used as inputs of vapour cloud explosions (VCEs) to investigate the potential overpressure effect. It was found the CFD model could predict a more reasonable flammable gas amount in cloud than using the bulk hydrogen release rate. The safety distance based on the overpressure prediction was reduced by 35%. The method proposed in this study can provide more validity for the consequence analysis as part of risk assessment.     

Automated prediction system of household energy consumption in cities using web crawler and optimized artificial intelligence

The supply of electrical energy is critical to convenient and comfortable living. However, people consume a large amount of energy, contributing to an energy crisis and global warming, and damaging some ecological cycles. Residential electricity consumption has greater elasticity than industrial and business consumption; it therefore has high energy-saving potential. This work establishes an automated platform, which provides information about residential electricity consumption in each city in Taiwan. Machine learning was used to forecast future residential electricity demand. A nature-inspired optimization method was applied to enhance the accuracy of the best machine learner, yielding an even better hybrid ensemble model. Performance measures indicate that the resulting model is accurate and provides effective information for reference. An automatic web-based system based on the model was combined with a web crawler and scheduled to run automatically to provide information on monthly residential electricity consumption in each county and city. By providing energy consumption information across the country, power providers and government can discuss policy and set different goals for energy use. The results of this study can facilitate the early implementation of energy-saving and carbon emission-reducing in cities and aid utility companies in establishing energy conservation guidelines.