基于循环一致性对抗网络的室内火焰图像场景迁移

基于深度学习的视频火灾探测模型的训练依赖于大量的正负样本数据,即火灾视频和带有干扰的场景视频。由于很多室内场合禁止点火,导致该场景下的火灾视频样本不足。本文基于生成对抗网络,将其他相似场景下录制的火焰迁移到指定场景,以此增广限制性场合下的火灾视频数据。文中提出将火焰内核预先植入场景使之具备完整的内容信息,再通过添加烟雾和地面反射等风格信息,完成场景与火焰的融合。该方法克服了现有多模态图像转换方法在图像转换过程中因丢失信息而造成的背景失真问题。同时为减少数据采集工作量,采用循环一致性生成对抗网络以解除训练图像必须严格匹配的限制。实验表明,与现有多模态图像转换相比,本文方法可以保证场景中火焰形态的多样性,迁移后的场景具有较高的视觉真实性,所得结果的FID与LPIPS值最小,分别为119.6和0.134 2。     

一种调度遥控操作双路径安全校核机制的探索

为解决调度端远方操作路径单一情况下可能出现操作错误的问题,文章探索提出一种对远方操作进行双路径安全确认的构建方案,在常规遥控发起路径之外建立独立的操作许可路径,在操作业务通信的出口处部署操作许可安全网关、操作许可工作站,通过操作发起路径的操作员、监护人和操作许可路径的许可员相结合的多重校验,强化对遥控操作的合法性认证,可以有效避免操作人、操作员站、操作网络环境的不可靠因素带来的操作安全风险,提高电网操作的安全性。     

泵站压力管道振动特性分析及减振优化设计

为探究泵站压力管道的不同布置形式对管道振动的影响,选取宁夏盐环定泵站工程二泵站#1压力管道为研究对象,建立基于流固耦合的泵站压力管道水流ALGOR数值模型,并采用DASP振动测试系统获取振动信号对模拟结果进行验证,分析压力管道水流流态对管道振动的激励特征并提出优化设计方案。结果表明,压力管道进水管处、支管总管相接处及总出水管处均有强烈的振动,但进水管的振动较出水总管振动更为强烈;减振优化设计方案中,调整支管距离对降低水流对管道振动的激励效果微弱,增大管径及改变总管走向对其效果较好,其中总管走向在2°～5°之间较佳,不仅能降低管道振动,而且能减少泵站能耗损失。研究结果可为高扬程泵站压力管道的优化设计提供理论依据。     

双层毛细芯对环路热管传热性能的实验分析

环路热管作为一种高效的相变传热装置,其性能与位于蒸发器和储液槽之间的毛细芯结构密切相关。为了更深入研究双层毛细芯对环路热管传热性能的影响,利用不同颗粒直径铜粉制备双层毛细芯,在毛细芯总厚度为5 mm的条件下,通过调整大粒径和小粒径层的相对厚度来改变毛细芯厚度比,对平板型蒸发器环路热管启动和变工况运行进行实验测试。实验结果表明:在同一工况下,不同厚度比的双层毛细芯启动特性存在显著差异,启动过程中出现小粒径层蒸发效率低引起的温度过冲和环路热管中气液两相流变化导致的温度振荡;同时存在一个较优的双铜层毛细芯厚度比,大粒径(180～280μm)铜层厚度为3 mm可提高蒸发效率,小粒径(56～71μm)铜层厚度为2 mm可提供足够毛细抽吸力保证环路热管稳定运行。搭载该厚度毛细芯的环路热管不仅启动速度快(125 s),而且总热阻和蒸发器壁面温度均最低,最大加热功率达到120 W(21.10 W/cm~2),对应热阻为0.17 K/W。     

Viscosity reduction of isotonic solutions of the photosensitizer TPCS2a by cyclodextrin complexation

Meso-tetraphenyl chlorin disulphonate (TPCS_2a) is a photosensitizer (PS) particularly developed and patented for use in the technology of photochemical internalization (PCI) against cancer. TPCS_2a is known to aggregate in aqueous media even at low concentrations (≥0.1?µM) and to form a high-viscosity network at clinically relevant concentrations (mM). The aim of this work was to evaluate the effect of two hydroxypropylated cyclodextrin derivatives of beta and gamma type, respectively i.e. HPβCD and HPγCD, on the aggregation and solubilization of TPCS_2a in isotonic solutions. Samples containing micromolar concentrations of TPCS_2a were studied spectrophotometrically, while samples containing a clinical relevant concentration (10?mM?=?9?mg/ml) of TPCS_2a were evaluated by dynamic viscosity measurements. HPβCD was determined to be a more suitable solubilizer of TPCS_2a than HPγCD in aqueous media both in the absence and presence of salt. The complexation stoichiometry between TPCS_2a/HPβCD at micromolar to millimolar concentrations of TPCS_2a was determined to be 1:3 and 1:2 in the absence and presence of isotonic NaCl, respectively. The network of TPCS_2a (10?mM) was broken down in the presence of 3% w/v (=?20?mM) HPβCD, i.e. a 1:2 molar ratio between TPCS_2a and the cyclodextrin. Formation of the inclusion complex resulted in low viscosity samples both in water and in the presence of isotonic NaCl or phosphate buffered saline (PBS) at 25?°C and 37?°C.     

Value stream mapping approach and analytical network process to identify and prioritize production system’s Mudas (case study: natural fibre clothing manufacturing company)

The current research has been done to identify and prioritize the Mudas of the production line of a natural fibre clothing manufacturing company; in this way, value stream mapping has been used to identify the Mudas, then analytical hierarchy method has been used to screen the Mudas. In the first step, necessary information to design value stream of production process has been obtained by observation, measurement and interview; afterward, according to the obtained information and by forming a focus group, Mudas in the production line have been identified and prioritized according to seven categories of Mudas. In order to prioritizing the Mudas by the usage of analytical hierarchy method, the Super Decision software has been applied. Then, according to the obtained priorities, suggested map of the future value system has been drawn and some suggestions have been presented to eliminate Mudas and move towards the lean manufacturing.     

Sustainability assessment of Iranian petrochemical companies in stock exchange: A data envelopment analysis-based approach

The assessment of efficiency is always of particular importance according to different indicators from different perspectives. There are various techniques for evaluating petrochemical companies, among which the data envelopment analysis technique is one of the best techniques that can be used to calculate the relative efficiency of a set of decision-making units with network structures. In the present paper, seven petrochemical companies listed in the Iranian stock exchange were analysed. These companies were evaluated in terms of financial performance and sustainable development, and their relative efficiency was calculated during 2015–2016. According to the obtained results, only Marun Petrochemical Co. was found to be efficient in all areas and years. The results also showed that four companies were efficient in financial terms over the period under study. In the general conclusion regarding the companies' performance, Marun was ranked first, Jam was ranked second, and Zagros was ranked third.     

基于卷积神经网络的葡萄叶片氮含量识别方法

为实现对葡萄叶片氮素含量快速、便捷的识别，在卷积神经网络VGG-16网络结构基础上，将数据增广后的图像按不同梯度划分进行模型训练，通过十折交叉验证法探究最佳的训练集与验证集分配比例，并构建4个不同深度的网络模型进行训练对比，采用全局平均池化代替全连接层约简网络参数量。训练结果表明，氮含量梯度设为0.70%、0.35%和0.175%时，室内简单背景识别准确率分别为85.9%、76.2%和71.1%；晴天室外复杂背景下识别准确率分别为44.6%、35.0%和30.4%。研究结果表明利用VGG-16建立的网络学习模型对葡萄叶片氮含量识别提供了一种新的便捷方法，对农业信息化和智能化技术应用具有一定促进作用。     

Densely connected convolutional network block based autoencoder for panorama map compression

As a novel virtual reality (VR) format, panorama maps are attracting increasing attention, while the compression of panorama images is still a concern. In this paper, a densely connected convolutional network block (dense block) based autoencoder is proposed to compress panorama maps. In the proposed autoencoder, dense blocks are specially designed to reuse feature maps and reduce redundancy of features. Meanwhile, a loss function, which imports a position-dependent weight item for each pixel, is proposed to train and adjust network parameters, in order to make the autoencoder fit to properties of panorama maps. Based on the proposed autoencoder and the weighted loss function, a greedy block-wise training scheme is also designed to avoid gradient vanishing problem and speed up training. During training process, the autoencoder is divided into several sub-nets. After each sub-net is trained separately, the whole network is fine-tuned to achieve the best performance. Experimental results demonstrate that the proposed autoencoder, compared with JPEG, saves up to 79.69 % bit rates, and obtains 7.27dB gain in BD-WS-PSNR or 0.0789 gain in BD-WS-SSIM. The proposed autoencoder also outperforms JPEG 2000, HEVC and VVC in both BD-WS-PSNR and BD-WS-SSIM. Meanwhile, subjective results show that the proposed autoencoder can recover details of panorama images, and reconstruct maps with high visual quality.