Cold Start Problem of Vehicle Model Recognition under Cross-Scenario Based on Transfer Learning

As a major function of smart transportation in smart cities,vehicle model recognition plays an important role in intelligent transportation.Due to the difference among different vehicle models recognition datasets,the accuracy of network model training in one scene will be greatly reduced in another one.However,if you don’t have a lot of vehicle model datasets for the current scene,you cannot properly train a model.To address this problem,we study the problem of cold start of vehicle model recognition under cross-scenario.Under the condition of small amount of datasets,combined with the method of transfer learning,load the DAN(Deep Adaptation Networks)and JAN(Joint Adaptation Networks)domain adaptation modules into the convolutional neural network AlexNet and ResNet,and get four models:AlexNet-JAN,AlexNet-DAN,ResNet-JAN,and ResNet-DAN which can achieve a higher accuracy at the beginning.Through experiments,transfer the vehicle model recognition from the network image dataset(source domain)to the surveillance-nature dataset(target domain),both Top-1 and Top-5 accuracy have been improved by at least 20%.

Alleviating the Cold Start Problem in Recommender Systems Based on Modularity Maximization Community Detection Algorithm

Recommender system (RS) has become a very important factor in many eCommerce sites. In our daily life, we rely on the recommendation from other persons either by word of mouth, recommendation letters, movie, item and book reviews printed in newspapers, etc. The typical Recommender Systems are software tools and techniques that provide support to people by identifying interesting products and services in online store. It also provides a recommendation for certain users who search for the recommendations. The most important open challenge in Collaborative filtering recommender system is the cold start problem. If the adequate or sufficient information is not available for a new item or users, the recommender system runs into the cold start problem. To increase the usefulness of collaborative recommender systems, it could be desirable to eliminate the challenge such as cold start problem. Revealing the community structures is crucial to understand and more important with the increasing popularity of online social networks. The community detection is a key issue in social network analysis in which nodes of the communities are tightly connected each other and loosely connected between other communities. Many algorithms like Givan-Newman algorithm, modularity maximization, leading eigenvector, walk trap, etc., are used to detect the communities in the networks. To test the community division is meaningful we define a quality function called modularity. Modularity is that the links within a community are higher than the expected links in those communities. In this paper, we try to give a solution to the cold-start problem based on community detection algorithm that extracts the community from the social networks and identifies the similar users on that network. Hence, within the proposed work several intrinsic details are taken as a rule of thumb to boost the results higher. Moreover, the simulation experiment was taken to solve the cold start problem.

Effect of catalyst layer mesoscopic pore-morphology on cold start process of PEM fuel cells

Water transport is of paramount importance to the cold start of proton exchange membrane fuel cells(PEMFCs).Analysis of water transport in cathode catalyst layer(CCL)during cold start reveals the distinct characteristics from the normal temperature operation.This work studies the effect of CCL mesoscopic pore-morphology on PEMFC cold start.The CCL mesoscale morphology is characterized by two tortuosity factors of the ionomer network and pore structure,respectively.The simulation results demonstrate that the mesoscale morphology of CCL has a significant influence on the performance of PEMFC cold start.It was found that cold-starting of a cell with a CCL of less tortuous mesoscale morphology can succeed,whereas starting up a cell with a CCL of more tortuous mesoscale morphology may fail.The CCL of less tortuous pore structure reduces the water back diffusion resistance from the CCL to proton exchange membrane(PEM),thus enhancing the water storage in PEM,while reducing the tortuosity in ionomer network of CCL is found to enhance the water transport in and the water removal from CCL.For the sake of better cold start performance,novel preparation methods,which can create catalyst layers of larger size primary pores and less tortuous pore structure and ionomer network,are desirable.

Experimental Study on the Combustion and Energy Flows of Vehicle Engine Under NEDC of Cold Start

To enhance the fuel economy of a vehicle powered by a gasoline engine under road conditions,an energy flow test of a vehicle was performed experimentally under the New European Driving Cycle of cold start.The energy distributions and related influencing factors were analyzed using the test data.Results show that the effective power and thermal efficiency are mainly affected by the engine load except in the early stage of the New European Driving Cycle.Because of the retarded CA50 and longer CA10-90,the effective thermal efficiency is lower in the early phase of driving conditions.Initially,the heat transfer loss mainly comprises the loss of the heating,ventilation,and air conditioning system.The radiator then plays the major role,with its percentage affected by the engine load and decreasing under the extra-urban driving cycle.The exhaust gas loss is decided by the temperature and flow rate of the exhaust gas,while its percentage is mainly affected by the temperature of the exhaust gas.In the early phase of driving conditions,the retarded spark advance angle leads to a higher temperature of the exhaust gas and a greater exhaust gas loss.The pumping loss and its percentage are mainly determined by the engine speed under the urban driving cycle,and both decrease under the extra-urban driving cycle except at maximum vehicle speed.

Blow-by and tribological performance of piston ring pack during cold start and warm idle operations

To reduce the fuel consumption of internal combustion engines, more attention has been paid to the tribological performance of the piston ring pack during the cold start and idle operations. In this research, a numerical model considering the cylinder liner deformation and the piston ring conformability is developed to predict the blow-by, lubrication, friction and wear of the piston ring pack under different operating conditions. The gas flow rate, inter-ring gas pressures, minimum oil film thickness, frictional force and wear load during cold start are calculated and compared with those during warm idle operating conditions. The results show that cylinder liner deformation and piston ring conformability together obviously affect blow-by and other tribological performance. Meanwhile, it is found that friction loss is larger during cold start than during warm idle operating conditions. However, the wear process is more severe during warm idle operation than during cold start. From this research, the blow-by and tribological performance of the piston ring pack during cold start and warm idle operations are understood more deeply.

Numerical Simulation of Cold-Start Emission for the Three-Way Catalytic Converter: Mathematical Model and Result Analysis

This paper integrated a two-dimensional axisymmetrical transient model applicable to cold-start emission applications. The model can be used to simulate and explain effects of the flow and temperature distribution on performance of a converter. The evolutions of distribution of the temperature and concentration in the monolith during the cold-start period and the effects of flow distribution in the monolith on the cold-start performance are simulated in terms of the integrated model. The investigation indicates that the axial and radial gradients of temperature of the solid become steeper as the inlet gas temperature ramp increases; this furthermore results in the movement of reaction region in the monolith, and the flow distribution in the monolith affects the radial distribution of temperature of the solid;the radial gradients of temperature of the solid become greater as the flow uniformity index decreases, whereas the light-off time doesn't always increase as the flow uniformity index decreases. The analyses on the distribution of temperature and concentration in the monolith show that the catalytic reaction zone concentrates in central area near the front face. The predicted curves of the velocity distribution have a good agreement with the experimental data.

Electronically Controlling the System of Preheating Intake Air by Flame for Diesel Engine Cold-Start

In order to improve the cold start performance of heavy duty diesel engine, electronically controlling the preheating of intake air by flame was researched. According to simulation and thermodynamic analysis about the partial working processes of the diesel engine, the amount of heat energy, enough to make the fuel self ignite at the end of compression process at different temperatures of coolant and intake air, was calculated. Several HY20 preheating plugs were used to heat up the intake air. Meanwhile, an electronic control system based on 8 bit micro controller unit (MCS 8031) was designed to automatically control the process of heating intake air. According to the various temperatures of coolant and ambient air, one plug or two plugs can automatically be selected to heat intake air. The demo experiment validated that the total system could operate successfully and achieve the scheduled function.

临界温度下循环喷油量对柴油机冷起动和排放特性的影响

为探明循环喷油量对小型农用柴油机冷起动特性和排放性能的影响规律,同时寻求其临界温度下的最佳循环喷油量,该研究以一款直列双缸高压共轨柴油发动机为研究对象,利用环境模拟舱整机试验平台,通过多次预试验测得海拔2000 m下,该机不开启预热装置能够成功起动的临界温度为-5℃。在此工况下,采用3次喷射方式,通过改变主喷油量研究循环喷油量对该机冷起动特性和排放性能的影响。结果表明:较大或较小的循环喷油量均会导致柴油机冷起动时间增加,甚至起动失败;随着循环喷油量增大,冷起动时间先减小后增加,循环喷油量为20 mg时冷起动时间最短,且HC、CO排放峰值均最低,分别为12984和3008 mg/L;增大或减少循环喷油量均会导致HC和CO排放峰值增加。循环喷油量较大时,CO排放峰值显著增加,并且峰值出现时刻提前。改变循环喷油量对NOx排放峰值有较大影响,循环喷油量为30 mg时NOx排放峰值最小,为368 mg/L。综合考虑,海拔高度2000 m,环境温度-5℃工况下该柴油机起动的最佳循环喷油量为20 mg。研究结果可为柴油机冷起动特性和排放优化提供支撑。

基于元学习个性化推荐研究综述

推荐系统作为缓解“信息过载”的工具,为用户过滤冗余信息并提供个性化服务,近年来得到了广泛应用。然而,实际推荐场景中,通常存在冷启动与不同推荐算法难以根据实际环境自适应选择等问题。元学习因其具有基于少量训练样本快速学会新知识和技能的优点,被越来越多地应用于推荐系统相关研究中。对现有基于元学习技术缓解推荐系统冷启动问题以及自适应推荐问题的主要研究进行探讨。首先,分析了基于元学习推荐在上述2个方面已取得的相关研究进展;然后,指出了现有元学习推荐研究存在难以适应复杂任务分布、计算代价高和容易陷入局部最优等问题;最后,对元学习在推荐系统领域的一些最新研究方向进行了展望。

极寒条件下含氧燃料对柴油机冷起动特性的影响

针对柴油机存在低温冷起动困难甚至无法起动的问题,基于起动/发电机集成一体化技术柴油机,在高原低温发动机冷起动试验舱,研究-50号柴油A0(氧质量分数为0%)、A1(氧质量分数为2.05%)和A2(氧质量分数为4.40%)燃料在-43℃下对柴油机冷起动过程的影响和燃烧特性分析.结果表明:相比A0纯柴油,燃用A1、A2含氧燃料起动时间分别缩短了36.64%和42.71%;起动累计油量分别降低47.8%和60.6%;怠速运行前60 s内转速波动率分别降低25.3%和43.8%.燃用燃料氧质量分数越高,起动燃烧首循环的缸内压力、燃烧放热率、缸内燃烧温度和缸内压力升高率峰值越高,且燃烧重心前移,燃烧持续期越短;起动转速上升过程循环数越少,平均最大缸内压力越大,起动过程燃烧稳定性越好.

不同尾气热管理策略对当量天然气发动机WHTC循环排放的影响

国六天然气发动机采用等当量燃烧结合外部冷却EGR及TWC的排放控制技术路线,TWC的转化效率是保证发动机排放达标的重要因素,而催化器的入口温度是决定TWC转化效率高低的最重要影响参数之一。然而,在冷起动阶段及排气管路温降较大的情况下,存在TWC入口尾气温度难以满足发动机及整车排放高效转化需求的问题。针对该问题,在一台4.5 L排量的发动机和一辆搭载该发动机的城市环卫车上探究不同尾气热管理策略对发动机WHTC循环排放及整车作业循环排放的影响。研究结果表明:采用起动快速温升的尾气热管理策略,能有效提高冷起动阶段发动机尾气温度,WHTC循环的TWC入口温度达到300℃所需时间缩短了300 s,显著降低了NO_(x),CH 4和CO循环排放,降幅分别达到38.5%,38.5%和5.8%;采取在TWC上增加金属载体前级方案的尾气控制策略,WHTC循环的NO_(x),CH 4和CO循环排放分别降低了62.2%,42.8%和15.7%;采用对整车排气尾管包裹的尾气热管理策略,在整车作业循环工况下,TWC入口温度相比采用普通石棉包裹的方案提高了近200℃,使得NO_(x)瞬时排放值30 s平均值的最大值降低了60.9%,能更好地满足北京第六阶段的排放法规要求。

燃油温度对柴油液相喷雾特性的影响

在低温环境下柴油黏度增大流动性变差,抑制了柴油机柴油喷雾的雾化和蒸发,直接影响柴油机的冷启动可靠性。为解决上述问题,采用背光法试验对比研究高温、低温柴油在不同工况条件下喷雾宏观结构的变化规律,并利用MATLAB对拍摄的喷雾图像进行数字图像处理获取喷雾宏观特性参数。试验结果表明,喷孔直径0.12 mm时,随着喷油压力的变化柴油温度对喷油量的影响比较明显,黏性力的持续作用减慢了靠近壁面的流体层的流动速度,抑制柴油喷出,柴油温度低于0℃时随着柴油温度降低喷油量减少,高压时柴油温度降低导致液滴间增强的黏性力在动能和惯性力的作用下反而促进柴油持续喷出,喷油量随着柴油温度降低而增多。喷油压力为75 MPa时,柴油温度降至-20℃相较38℃柴油喷油量增加了近23.87%。并且小孔径下柴油温度降低液滴间黏性力增大,导致喷雾液滴尺寸增大,液滴所具有的动量越大,削弱柴油喷雾径向发展趋势,轴向运动能力增强,随着柴油温度的降低,喷雾贯穿距增大而喷雾锥角减小,柴油温度从38℃降至-20℃,柴油喷雾在0.70 ms时刻喷雾贯穿距增大了39.89 mm。喷孔直径为0.28 mm时,喷射压力小于75 MPa,喷油量随着柴油温度的降低而减少,相同喷油压力下柴油更多的动量消耗于克服更大的内部摩擦力,使其到达最远距离的能力减弱,降低柴油温度会导致喷雾贯穿距和喷雾锥角均减小,当喷油压力为75 MPa时,0.70 ms时刻-20℃柴油喷雾贯穿距相对38℃柴油喷雾贯穿距缩短了30.86 mm。相同喷射压力下增大喷孔直径,-20℃柴油喷雾贯穿距减小,喷雾锥角增大。随着喷油压力的提升柴油温度对喷雾宏观特性的影响更加突出,喷油压力越高降低柴油温度造成的喷雾贯穿距缩短趋势越明显。研究结果可为柴油机冷起动过程低温柴油喷雾特性提供支撑。

基于知识图谱与邻域感知注意力机制的推荐算法研究

为解决传统推荐算法在面对数据稀疏的推荐任务时产生的冷启动问题,本研究将知识图谱引入推荐算法,结合一种新的邻域感知注意力机制代替传统图注意力机制来挖掘实体间的高阶连通信息,提出了基于知识图谱和邻域感知注意力机制的推荐模型KGNPAN。得益于知识图谱可使推荐具有精准、多样和可解释的特点,该模型能够很好地缓解数据稀疏与冷启动问题。首先利用基于自对抗负采样的图嵌入方法RotatE对原有物品和用户表征的语义信息进行扩充,将实体和关系向量映射成低维嵌入向量;其次,根据协同邻居的不同类型分别应用邻域感知注意力机制聚合邻居节点信息,丰富目标节点语义,并以卷积形式递归挖掘高阶连通信息;最后对用户与项目向量应用内积操作计算交互概率,得到推荐结果。在Amazon-book和Last-FM两个公共基准数据集上进行实验,结果表明,在与CKE,BPRMF,RippleNet,KGAT,KGCN和CAKN 6个基准模型的对比中,KGNPAN相较于基准模型中结果最优的CAKN模型,在召回率(Recall)上分别提升了1.30%和1.37%,在归一化折损累计增益上(NDCG)分别提升了1.26%和1.14%,充分验证了其有效性和可解释性。

融合信息瓶颈与图卷积的跨域推荐算法

基于迁移学习的跨域推荐可以有效地学习连接源域和目标域的映射函数,但其性能仍然受到表征质量不高和负迁移问题的影响,不能有效地为冷启动用户进行推荐,为此提出了一种融合信息瓶颈与图卷积网络的跨域推荐模型(IBGC)。利用图卷积神经网络聚合有关联的用户-用户和项目-项目信息;利用注意力机制学习用户和项目偏好,以提高节点特征表示质量;考虑到两个领域的信息交互,将重叠用户进行嵌入表示的同时限制特定信息的编码,利用信息瓶颈理论设计了三种正则化器,以捕获域内和跨域用户-项目的相关性,并将不同领域的重叠用户表征对齐以解决负迁移问题。在Amazon数据集中的四对公开数据集上进行实验,实验结果表明该模型在MRR、HR@K和NDCG@K三个推荐性能指标上的表现均优于基线模型,在四对数据集上与最优对比基线模型相比,MRR平均提升34.36%,HR@10平均提升34.94%,NDCG@10平均提升36.83%,证明了IBGC模型的有效性。

结构起点、政策选择与中美关系的另一种可能

中美关系的走向决定未来国际秩序。从结构现实主义出发,有观点认为中美势必走向冲突与战争。通过定位中美所处的结构起点并对其进行深入分析,发现在经济、安全、意识形态和应对全球性问题上,中美关系具有不同于“新冷战”的特征。总体上,经济结构是发展导向的,安全环境是和平导向的,意识形态是包容导向的,解决全球性挑战是合作导向的。从特定的结构起点出发,中美关系走向和平共处、合作共赢是优先选项,而走向冲突和战争代价巨大、成本极高。尽管如此,特定的结构起点及结构过程能否持续发挥作用,也取决于中美两国的政策选择。总体上,中美两国政府有意管控分歧,增强合作,这有助于避免中美关系滑入“新冷战”。无论未来美国政府对华政策如何变化,中国要更为积极地塑造中美关系,推进有利于发展、和平和多元的结构进程,提升结构的和平性、互利性和稳定性,借助结构的力量来限制美国对华“新冷战”。

融合协同过滤的神经Bandits推荐算法

针对数据稀疏性和“冷启动”对协同过滤的限制以及现有的协同多臂老虎机算法不适用于非线性奖励函数的问题,提出一种融合协同过滤的神经Ba ndits推荐算法COEENet.首先,采用双神经网络结构学习预期奖励及潜在增益;其次,考虑邻居协同作用;最后,构造决策器进行最终决策.实验结果表明,该方法在累积遗憾上优于4种基线算法,推荐效果较好.

基于单片阻抗一致性吹扫的燃料电池低温冷启动策略研究

车用石墨板燃料电池低温冷启动能力弱,是影响燃料电池技术在北方寒冷地区大规模推广的重要瓶颈。饥饿自升温是一种常见的低温冷启动策略,其基本原理是通过降低反应物供应速率来增加过电势,短时间在电池内部产生大量热量从而实现快速升温。该方法原理简单,但对电堆单体初始含水量一致性要求高、且易出现单片反极和尾排氢浓度超标的情况,影响系统安全性和电堆耐久性。针对上述问题,课题组研制了单体多通道交流阻抗在线测试装置,提出了面向单片阻抗一致性的电堆优化吹扫策略,建立了基于定电压变流量泵氢控制的低温冷启动方法,实现了低温启动瞬态过程的高产热、高安全、高动态的电压、电流、进堆/旁通空气流量的多目标多参数的耦合协调控制。台架实验结果表明,采用优化吹扫策略后,单片阻抗间最大差值由0.7降低至0.2 mΩ以下;燃料电池发动机系统可实现124 s内-40℃下快速启动,且重复性好。相关技术在北京冬奥会燃料电池示范中获得了应用,验证了其有效性。

基于蓄热装置辅助的燃料电池电动车供暖系统动态仿真

为了辅助燃料电池车辆预热,提升燃料电池在低温下的冷启动速率以及降低电动车能耗,该文提出了一种基于热泵和蓄热装置联合的燃料电池电动车供暖方案,并开展了燃料电池车辆系统动态仿真研究,对比了有无蓄热装置辅助的系统对车辆运行过程的影响。结果表明:在寒冷运行环境下,蓄热装置可确保驾驶循环工况下循环冷却液热量来源的连续性,相较于无蓄热装置的热管理系统,热泵和蓄热联合供暖方案可使空调系统能耗下降26%;利用蓄热装置辅助的供暖系统能够改善燃料电池电动车的冬季冷启动特性,优化车辆电池系统的热管理。

基于边缘计算的融合多因素的个性化推荐算法

针对传统推荐算法使用单一上下文信息不能有效地解决信息过载、数据稀疏、冷启动的问题,提出了一种基于边缘计算的融合多因素的个性化推荐算法F-SVD和新的用户相似度计算方法F-PEARSON(改进后的PEARSON相关系数).在边缘服务器处理个性化用户数据以分散云服务器的压力,目前大多采用的集中式处理方式无法在爆炸性式增长的数据下提供准确的推荐,在云服务器融合多因素挖掘用户之间的潜在关系,从而构建预测F-SVD算法.实验结果表明:在公开数据集MovieLens上,与传统算法相比,所提出的算法在RMSE和MAE上的误差更小,精确度提升了2.2%.

基于金属有机框架的汽油机冷启动碳氢化合物吸附和扩散性能的研究

为研究汽油机冷启动过程产生的碳氢化合物在金属有机框架上的吸附和扩散行为特性,建立金属有机框架(MOFs)的计算模型,并验证模型的准确性。然后,对比分析C2H4和C6H6组分在沸石与MOFs上的吸附性能。最后,采用蒙特卡罗和分子动力学方法,分析C2H4、C6H6及等物质的量比混合组分在ZIF-20、ZIF-67、ZIF-318这3种MOFs结构的吸附性能和扩散性能。研究结果表明:与沸石材料相比,MOFs具有更优异的碳氢吸附效果。3种MOFs结构中,ZIF-20对C2H4和C6H6具有最大的吸附量,其吸附量分别为202.1和187.8个晶胞。在ZIF-20中C2H4的吸附位点主要集中在β-笼,C6H6的吸附位点主要集中在α-笼和双四元环。而在ZIF-67和ZIF-318中的吸附位点主要集中在十二元环。C2H4和C6H6分别在ZIF-318和ZIF-20中的扩散最强烈,且C2H4在3种MOFs上的扩散系数明显比C6H6的扩散系数大。