基于深度学习的学生课堂注意力评价

针对学生课堂行为状态识别准确率较低的问题, 提出一种基于YOLOv4的改进模型. 通过建立学生课堂行为状态数据集, 调整YOLOv4算法训练模型的参数, 修改卷积块激活函数为ELU函数以优化模型, 同时提出将DIoU-Soft-NMS作为非极大值抑制机制, 识别分析教室中学生课堂行为状态; 根据各状态持续时长及状态变化频率计算学生听课有效时长, 并参考山东高考赋分原则, 建立学生课堂注意力量化评价准则, 同时建立教师课堂授课效果量化评价标准. 实验结果表明, 以同一评价指标衡量模型, 该模型在学生课堂行为检测速率不变的情况下, 平均精度均值(mAP)达到98.8%, 比原YOLOv4模型提升了3.53%, 学生服课堂注意力量化评价准则, 有较高的契合度.     

Highly parallel GEMV with register blocking method on GPU architecture

GPUs can provide powerful computing ability especially for data parallel applications, such as video/image processing applications. However, the complexity of GPU system makes the optimization of even a simple algorithm difficult. Different optimization methods on a GPU often lead to different performances. The matrix–vector multiplication routine for general dense matrices (GEMV) is an important kernel in video/image processing applications. We find that the implementations of GEMV in CUBLAS or MAGMA are not efficient, especially for small or fat matrix. In this paper, we propose a novel register blocking method to optimize GEMV on GPU architecture. This new method has three advantages. First, instead of using only one thread, we use a warp to compute an element of vector y so that the method can exploit the highly parallel GPU architecture. Second, the register blocking method is used to reduce the requirement of off-chip memory bandwidth. At last, the memory access order is elaborately arranged for the threads in one warp so that coalesced memory access is ensured. The proposed optimization methods for GEMV are comprehensively evaluated on different matrix sizes. The performance of the register blocking method with different block sizes is also evaluated in the experiment. Experiment results show that the new method can achieve very high speedup for small square matrices and fat matrices compared to CUBLAS or MAGMA, and can also achieve higher performance for large square matrices.     

低压供配电线路故障电弧检测方法

通过搭建试验平台模拟建筑物低压供配电线路中的故障电弧,分析故障电弧特征,提取出三种表征故障电弧的特征量,包括各周期采样点平均值、波形斜率和小波高频系数。应用MATLAB模糊工具箱,采用模糊逻辑推理的方法将三种判据融合,建立故障概率判定系统。     

Effects of Si3N4 and WC on the oxidation resistance of ZrB2/SiC ceramic tool materials

ZrB₂/SiC, ZrB₂/SiC/Si₃N₄ and ZrB₂/SiC/WC ceramic tool materials were prepared by spark plasma sintering technology, and their oxidation resistance was tested at different oxidation temperatures. When the oxidation temperature is 1300 °C, the oxide layer thickness, oxidation weight gain and flexural strength of ZrB₂/SiC/Si₃N₄ ceramic tool material after oxidation are 8.476 μm, 1.436 mg cm^?2 and 891.0 MPa, respectively. Compared with ZrB₂/SiC ceramic tool materials, the oxide layer thickness and oxidation weight gain are reduced by 8.2% and 11.8%, respectively, and the flexural strength after oxidation is increased by 116.1%. However, the addition of WC significantly reduces the oxidation resistance of the ceramic tool material. A dense oxide film is formed on the surface of ZrB₂/SiC/Si₃N₄ ceramic tool material during oxidation, which effectively prevents oxygen from entering the inside of the material, thereby improving the oxidation resistance of the ceramic tool material.     

Pool boiling heat transfer on the microheater surface with and without nanoparticles by pulse heating

We study the pool boiling heat transfer on the microheater surface with and without nanoparticles by pulse heating. Nanofluids are the mixture of de-ionized water and Al₂O₃ particles with 0.1%, 0.2%, 0.5% and 1.0% weight concentrations. The microheater is a platinum surface by 50 × 20 μm. Three types of bubble dynamics were identified. The first type of bubble dynamics is for the boiling in pure water, referring to a sharp microheater temperature increase once a new pulse cycle begins, followed by a continuous temperature increase during the pulse duration stage. Large bubble is observed on the microheater surface and it does not disappear during the pulse off stage. The second type of bubble dynamics is for the nanofluids with 0.1% and 0.2% weight concentrations. The microheater surface temperature has a sharp increase at the start of a new pulse cycle, followed by a slight decrease during the pulse duration stage. Miniature bubble has oscillation movement along the microheater length direction, and it disappears during the pulse off stage. The third type of bubble dynamics occurs at the nanofluid weight concentration of 0.5% and 1.0%. The bubble behavior is similar to that in pure water, but the microheater temperatures are much lower than that in pure water. A structural disjoining pressure causes the smaller contact area between the dry vapor and the heater surface, decreasing the surface tension effect and resulting in the easy departure of miniature bubbles for the 0.1% and 0.2% nanofluid weight concentrations. For the 0.5% weight concentration of nanofluids, coalescence of nanoparticles to form larger particles is responsible for the large bubble formation on the heater surface. The microlayer evaporation heat transfer and the heat transfer mechanisms during the bubble departure process account for the higher heat transfer coefficients for the 0.1% and 0.2% nanofluid weight concentrations. The shortened dry area between the bubble and the heater surface, and the additional thin nanofluid liquid film evaporation heat transfer, account for the higher heat transfer coefficient for the 0.5% nanofluid weight concentration, compared with the pure water runs.     

High-temperature oxidation response of Mo–Si–B composites with TiO2W/SiCW addition

In this study, TiO_2W addition improved the oxidation resistance of the Mo–Si–B composite at 1300 °C. The TiO₂ partially dissolved in SiO₂ modified the network structure of the SiO₂ glass and improved its fluidity at the initial oxidation stage. This favored to a continuous scale cover on the surface of the Mo–Si–B composite rapidly. The residual TiO_2W promoted the formation of a passivated multilayer borosilicate scale at current temperature, which could impede the MoO₃ volatilisation and O diffusion at the stable oxidation stage. The SiC_W addition, compared to the TiO_2W, especially could ensure the Mo–Si–B–SiC composite withstand a higher temperature such as 1400 °C. Its oxidation and the more intermetallics in the composite could increase the number of active sites of the SiO₂ glass, thereby supplying the borosilicate scale with a relatively sufficient Si element. Thus, the transient oxidation stage was minimised and the initial mass loss was reduced, which indicated a continuous borosilicate scale had formed quickly at the initial stage. Finally, the improved viscosity of the borosilicate due to a lower B/Si ratio, could obviously decreased the oxygen diffusion and enabled the formation of a protective borosilicate layer at or above 1400 °C.     

基于机器视觉的螺纹钢表面缺陷检测方法

螺纹钢是一种广泛应用的建筑材料,在轧制过程中如果不能及时发现其尺寸和表面缺陷,就会生产出大量废品,给企业带来损失.本文设计了一种基于视觉的螺纹钢表面缺陷检测方法.先利用仿射变换对图像中歪斜的螺纹钢进行校正,然后基于霍夫变换检测纵肋边缘直线位置的方法对螺纹钢正面、侧面图像进行区分.最后针对正面、侧面图像分别进行缺陷检测,快速准确地判别表面是否存在缺陷.实验表明所设计的方法具有较好的稳定性和实用性,能有效地解决人工检测过程中效率低、误检率高等问题.     

一种低压电弧故障检测的新方法

利用二进小波变换法分解低压配电线路电弧故障时的电流波形, 获得电流波形各尺度小波变换的高频系数. 电弧故障时的电流波形小波变换高频系数与负载正常启动和运行状态时相比, 有其明显的特征, 可有效地应用于电弧故障检测. 利用LabVIEW和MATLAB软件平台实现了电流在线采集和分析, 实验结果表明该方法能准确地辨识低压配电线路中的电弧故障.     

Modelling high-temperature co-implantation of N+ and Al+ ions in silicon carbide: the effect of stress on the implant and damage distributions

This work is an initial attempt to model the fundamental processes that occur when SiC is implanted at elevated substrate temperatures T_i (200°–800°) with high doses of N⁺ and Al⁺ ions to synthesise buried layers of (SiC)_1 − x(AlN)_x. The theoretical treatment has involved ballistic calculation of the implant and damage profiles by means of computer codes (TRIRS and DYTRIRS) specifically developed for modelling complex, multi-elemental targets. The influence of the mechanical stress induced the by implanted ions has been taken into account by adding a special term to the differential equations describing the evolution of the implant and damage distributions. Results from the simulations have been correlated with data obtained by Rutherford backscattering spectrometry/ion channelling (RBS/C). The theoretical approach described has enabled one to determine the interaction energies of the interstitials with the internal stress field as well as the role of stress on the defect distribution.     

Fast and scalable lock methods for video coding on many-core architecture

Many-core processors are good candidates for speeding up video coding because the parallelism of these applications can be exploited more efficiently by the many-core architecture. Lock methods are important for many-core architecture to ensure correct execution of the program and communication between threads on chip. The efficiency of lock method is critical to overall performance of chipped many-core processor. In this paper, we propose two types of hardware locks for on-chip many-core architecture, a centralized lock and a distributed lock. First, we design the architectures of centralized lock and distributed lock to implement the two hardware lock methods. Then, we evaluate the performance of the two hardware locks and a software lock by quantitative evaluation micro-benchmarks on a many-core processor simulator Godson-T. The experimental results show that the locks with dedicated hardware support have higher performance than the software lock, and the distributed hardware lock is more scalable than the centralized hardware lock.