Variable nozzle turbine （VNT） has become a popular variable geometry turbine （VGT） technology for the diesel engine application. Nozzle clearance, which can＇t be avoided on the hub and shroud side of the VNT turb...Variable nozzle turbine （VNT） has become a popular variable geometry turbine （VGT） technology for the diesel engine application. Nozzle clearance, which can＇t be avoided on the hub and shroud side of the VNT turbine due to the pivoting stators, can lead to turbine performance deterioration. However, its mechanism is still not clear. In this paper, numerical investigation, which is validated by experiment, is carried out to study the mechanism of the nozzle clearance＇s effect on the turbine performance. Firstly, performance of the mixed flow turbine with fixed nozzle clearances tested on flow bench. Performance of the tested turbine with the same nozzle clearance is numerically simulated. The numerical result agrees well with the test data, which proves correct of the numerical method. Then the turbine performance with different nozzle clearances is numerically analyzed. The research showed that with nozzle clearance, flow loss in the nozzle increases at first and it reaches the maximum value when the clearance ratio is 5%. Flow at the exit of the nozzle becomes less uniform with nozzle clearance. The negative incidence angle of the rotor also increases with nozzle clearance and leads to more incidence angle loss in the rotor. The low energy fluid formed in the nozzle due to the nozzle clearance migrates from hub to shroud side in the rotor, which is another main reason for the rotor＇s performance degradation. The present research exposed the mechanism of the dramatically decrease of the turbine performance with nozzle clearance： （a） The loss associated with the nozzle leakage increases with the nozzle clearance; （b） The flow loss grows up quickly in the rotor due to the incidence angle loss and migration of the low energy fluid from hub to shroud side.展开更多
Inlet recirculation is proved as an effective way for centrifugal compressor surge margin extension,and is successively used in some engineering applications.Unfortunately its working mechanism is still not being well...Inlet recirculation is proved as an effective way for centrifugal compressor surge margin extension,and is successively used in some engineering applications.Unfortunately its working mechanism is still not being well understood,which leads to redesigning of inlet recirculation mostly by experience.Also,most study about inlet recirculation is steady to date.It is necessary to study surge margin extension mechanism about inlet recirculation.To expose the mechanism in detail,steady and unsteady numerical simulations were performed on a centrifugal compressor with and without inlet recirculation.The results showed that,with inlet recirculation,the inlet axial velocity is augmented,relative Mach number around blade tip leading edge area is significantly reduced and so is the flow angle.As the flow angle decreased,the incidence angle reduced which greatly improves the flow field inside the impeller.Moreover,inlet recirculation changes the blade loading around blade tip and restrains the flow separation on the blade suction side at the leading edge area.The unsteady results of static pressure around blade surface,entropy at inlet crossflow section and vorticity distributions at near tip span surface indicated that,at near stall condition,strong fluctuation exists in the vicinity of tip area due to the interaction between tip leakage flow and core flow.By inlet recirculation these strong flow fluctuations are eliminated so the flow stability is greatly enhanced.All these improvements mentioned above are the reason for inlet recirculation delays compressor stall.This research reveals the surge margin extension reason of inlet recirculation from an unsteady flow viewpoint and provides important reference for inlet recirculation structure design.展开更多
Horizontal and vertical distributions of δ^18 and δ^13 were investigated in shells of four planktonic foraminiferal species, Globigerinoides ruber, Globigerinoides sacculifer, Pulleniatina obliquiloculata and Neoglo...Horizontal and vertical distributions of δ^18 and δ^13 were investigated in shells of four planktonic foraminiferal species, Globigerinoides ruber, Globigerinoides sacculifer, Pulleniatina obliquiloculata and Neogloboquedrina dutertrei, from a total of 62 core-top sediment samples from the Indonesian throughflow region. Results were compared to modern hydrologic conditions in order to explore potential of proxies in reconstructing fluvial discharge and upper ocean water column characteristics in this region. Our results show that, in the Makassar Strait, both of depleted δ^18 and δ^13 of these four species were linked to freshwater input. In the Bali Sea,however, depleted δ^18 and δ^13 for these species may be due to different reasons. Depleted δ^18 was a result of freshwater input and as well influenced by along-shore currents while depleted δ^13 was more likely due to the Java-Sumatra upwelling. Comparison of shell δ^18 records and hydrographic data of World Ocean Atlas 2005 suggests that G. ruber and G. sacculifer calcify within the mixed-layer, respectively at 0–50 m and 20–75 m water depth, and P. obliquiloculata and N. dutertrei within the upper thermocline, both at 75–125 m water depth. N.dutertrei calcifies at slightly deeper water depth than P. obliquiloculata does. In general, δ^13 values of both G.ruber and G. sacculifer are larger than those of P. obliquiloculata and N. dutertrei at all sites, possibly related to depth habitats of these species and vertical distribution of nutrients in the Indonesian throughflow region.展开更多
面部表情是传递情感的重要信息,是家畜生理、心理和行为的综合反映,可以用于评估家畜福利。由于家畜面部肌群结构简单,因此家畜面部不同区域的细微变化对于表情的反映较难识别。该研究提出一种基于多注意力机制级联LSTM框架模型(Multi-a...面部表情是传递情感的重要信息,是家畜生理、心理和行为的综合反映,可以用于评估家畜福利。由于家畜面部肌群结构简单,因此家畜面部不同区域的细微变化对于表情的反映较难识别。该研究提出一种基于多注意力机制级联LSTM框架模型(Multi-attention Cascaded Long Short Term Memory,MA-LSTM)对家猪时序面部表情进行分类识别。首先通过简化的多任务级联卷积结构实现帧图像中猪脸的快速检测与定位,去除非猪脸区域对于识别性能的影响。其次提出一种多注意力机制模块,利用不同特征通道视觉信息不同相应峰值响应区域也不同这一特性,通过对峰值响应相近区域进行聚类捕获表情变化引起的面部显著性区域,实现对面部细微变化的关注。在自标注构建的家猪表情数据集上的试验结果表明,该研究提出的多注意力机制级联LSTM模型在4类表情的平均识别准确率为91.826%,对比关闭多注意力机制模块平均识别准确率平均提升6.3个百分点,同时误分率也有较为明显的降低。对比其他常用面部表情识别算法LBP-TOP、HOG-TOP、ELRCN、STC-NLSTM,MA-LSTM模型平均识别精度分别提升约32.6、18.0、5.9和4.4个百分点。试验结果验证了该研究提出的多注意力机制级联LSTM模型在猪脸表情识别的有效性。展开更多
早期病虫害精准识别是预警和防控的关键,但是病虫害种类繁多数量巨大,外部形态存在类间相似度较高而类内差异性较大等性状特征,导致病虫害识别仍然是一项极具挑战的工作。为实现病虫害识别分类任务中差异化特征的提取和表示,该研究提出...早期病虫害精准识别是预警和防控的关键,但是病虫害种类繁多数量巨大,外部形态存在类间相似度较高而类内差异性较大等性状特征,导致病虫害识别仍然是一项极具挑战的工作。为实现病虫害识别分类任务中差异化特征的提取和表示,该研究提出一种大规模多类别精细病虫害识别网络模型(a large-scale multi-category fine-grained pest and disease network,PD-Net)。首先通过在基准网络模型中引入卷积块注意力模型,通过混合跨特征通道域和特征空间域实现模型在通道和空间两个维度上对关键特征提取和表示,用以增强网络对差异化特征的提取和表示能力。其次引入跨层非局部模块,提升模型在多个特征提取层之间对于多尺度特征的融合。在61类病害数据集和102类虫害数据集上的试验结果表明,对比AlexNet、VGG16、GoogleNet、Inception-v3、DenseNet121和ResNet50模型,该研究提出的面向大规模多类别病虫害识别模型,Top1识别准确率在病害和虫害集上分别达到88.617%和74.668%,精确率分别达到了0.875和0.745,召回率分别达到0.874和0.738,F1值达到0.874和0.732,试验结果对比其他模型均有一定幅度的提升,验证了PD-Net模型在大规模多类别病虫害识别上的有效性。展开更多
基金supported by Advanced Boost System Development for Diesel HCCI Application of DOE(Grant No. DE-FC26-07-NT43280)
文摘Variable nozzle turbine （VNT） has become a popular variable geometry turbine （VGT） technology for the diesel engine application. Nozzle clearance, which can＇t be avoided on the hub and shroud side of the VNT turbine due to the pivoting stators, can lead to turbine performance deterioration. However, its mechanism is still not clear. In this paper, numerical investigation, which is validated by experiment, is carried out to study the mechanism of the nozzle clearance＇s effect on the turbine performance. Firstly, performance of the mixed flow turbine with fixed nozzle clearances tested on flow bench. Performance of the tested turbine with the same nozzle clearance is numerically simulated. The numerical result agrees well with the test data, which proves correct of the numerical method. Then the turbine performance with different nozzle clearances is numerically analyzed. The research showed that with nozzle clearance, flow loss in the nozzle increases at first and it reaches the maximum value when the clearance ratio is 5%. Flow at the exit of the nozzle becomes less uniform with nozzle clearance. The negative incidence angle of the rotor also increases with nozzle clearance and leads to more incidence angle loss in the rotor. The low energy fluid formed in the nozzle due to the nozzle clearance migrates from hub to shroud side in the rotor, which is another main reason for the rotor＇s performance degradation. The present research exposed the mechanism of the dramatically decrease of the turbine performance with nozzle clearance： （a） The loss associated with the nozzle leakage increases with the nozzle clearance; （b） The flow loss grows up quickly in the rotor due to the incidence angle loss and migration of the low energy fluid from hub to shroud side.
文摘Inlet recirculation is proved as an effective way for centrifugal compressor surge margin extension,and is successively used in some engineering applications.Unfortunately its working mechanism is still not being well understood,which leads to redesigning of inlet recirculation mostly by experience.Also,most study about inlet recirculation is steady to date.It is necessary to study surge margin extension mechanism about inlet recirculation.To expose the mechanism in detail,steady and unsteady numerical simulations were performed on a centrifugal compressor with and without inlet recirculation.The results showed that,with inlet recirculation,the inlet axial velocity is augmented,relative Mach number around blade tip leading edge area is significantly reduced and so is the flow angle.As the flow angle decreased,the incidence angle reduced which greatly improves the flow field inside the impeller.Moreover,inlet recirculation changes the blade loading around blade tip and restrains the flow separation on the blade suction side at the leading edge area.The unsteady results of static pressure around blade surface,entropy at inlet crossflow section and vorticity distributions at near tip span surface indicated that,at near stall condition,strong fluctuation exists in the vicinity of tip area due to the interaction between tip leakage flow and core flow.By inlet recirculation these strong flow fluctuations are eliminated so the flow stability is greatly enhanced.All these improvements mentioned above are the reason for inlet recirculation delays compressor stall.This research reveals the surge margin extension reason of inlet recirculation from an unsteady flow viewpoint and provides important reference for inlet recirculation structure design.
基金The National Natural Science Foundation of China under contract No.41176044Shaanxi Provincial Technology Foundation for Selected Overseas Chinese Scholar under contract Shaan Renshe[2015]No.1190
文摘Horizontal and vertical distributions of δ^18 and δ^13 were investigated in shells of four planktonic foraminiferal species, Globigerinoides ruber, Globigerinoides sacculifer, Pulleniatina obliquiloculata and Neogloboquedrina dutertrei, from a total of 62 core-top sediment samples from the Indonesian throughflow region. Results were compared to modern hydrologic conditions in order to explore potential of proxies in reconstructing fluvial discharge and upper ocean water column characteristics in this region. Our results show that, in the Makassar Strait, both of depleted δ^18 and δ^13 of these four species were linked to freshwater input. In the Bali Sea,however, depleted δ^18 and δ^13 for these species may be due to different reasons. Depleted δ^18 was a result of freshwater input and as well influenced by along-shore currents while depleted δ^13 was more likely due to the Java-Sumatra upwelling. Comparison of shell δ^18 records and hydrographic data of World Ocean Atlas 2005 suggests that G. ruber and G. sacculifer calcify within the mixed-layer, respectively at 0–50 m and 20–75 m water depth, and P. obliquiloculata and N. dutertrei within the upper thermocline, both at 75–125 m water depth. N.dutertrei calcifies at slightly deeper water depth than P. obliquiloculata does. In general, δ^13 values of both G.ruber and G. sacculifer are larger than those of P. obliquiloculata and N. dutertrei at all sites, possibly related to depth habitats of these species and vertical distribution of nutrients in the Indonesian throughflow region.
文摘以金属化率为60%的低品位金属化球团为研究对象,使用热力学软件FactSage 8.3分析了其熔分过程,熔渣的熔点和黏度,以及熔渣与耐火材料间的相互作用.结果表明:金属化球团熔分的温度应高于1 515℃;每加入1 t金属化球团进行熔分时,应额外添加43.6 kg碳粉;随着留钢量的增大,金属化球团熔分后的铁收得率升高,但均低于80%,仍需配加碳粉以促进球团深度还原.根据CaO-SiO_(2)-Al_(2)O_(3)相图,确定了金属化球团熔渣调质的目标成分,即每熔分1 t金属化球团时,应添加6.31 kg Al_(2)O_(3),2.66 kg SiO_(2)和12.64 kg MgO,调质后熔渣的熔点为1 405℃;当熔炼温度为1 600℃时,调质后熔渣的黏度为0.164 Pa·s.以调质后的熔渣为基准,随着MgO添加量的增大,熔融热解炉炉衬受到的侵蚀减弱.
文摘面部表情是传递情感的重要信息,是家畜生理、心理和行为的综合反映,可以用于评估家畜福利。由于家畜面部肌群结构简单,因此家畜面部不同区域的细微变化对于表情的反映较难识别。该研究提出一种基于多注意力机制级联LSTM框架模型(Multi-attention Cascaded Long Short Term Memory,MA-LSTM)对家猪时序面部表情进行分类识别。首先通过简化的多任务级联卷积结构实现帧图像中猪脸的快速检测与定位,去除非猪脸区域对于识别性能的影响。其次提出一种多注意力机制模块,利用不同特征通道视觉信息不同相应峰值响应区域也不同这一特性,通过对峰值响应相近区域进行聚类捕获表情变化引起的面部显著性区域,实现对面部细微变化的关注。在自标注构建的家猪表情数据集上的试验结果表明,该研究提出的多注意力机制级联LSTM模型在4类表情的平均识别准确率为91.826%,对比关闭多注意力机制模块平均识别准确率平均提升6.3个百分点,同时误分率也有较为明显的降低。对比其他常用面部表情识别算法LBP-TOP、HOG-TOP、ELRCN、STC-NLSTM,MA-LSTM模型平均识别精度分别提升约32.6、18.0、5.9和4.4个百分点。试验结果验证了该研究提出的多注意力机制级联LSTM模型在猪脸表情识别的有效性。
文摘早期病虫害精准识别是预警和防控的关键,但是病虫害种类繁多数量巨大,外部形态存在类间相似度较高而类内差异性较大等性状特征,导致病虫害识别仍然是一项极具挑战的工作。为实现病虫害识别分类任务中差异化特征的提取和表示,该研究提出一种大规模多类别精细病虫害识别网络模型(a large-scale multi-category fine-grained pest and disease network,PD-Net)。首先通过在基准网络模型中引入卷积块注意力模型,通过混合跨特征通道域和特征空间域实现模型在通道和空间两个维度上对关键特征提取和表示,用以增强网络对差异化特征的提取和表示能力。其次引入跨层非局部模块,提升模型在多个特征提取层之间对于多尺度特征的融合。在61类病害数据集和102类虫害数据集上的试验结果表明,对比AlexNet、VGG16、GoogleNet、Inception-v3、DenseNet121和ResNet50模型,该研究提出的面向大规模多类别病虫害识别模型,Top1识别准确率在病害和虫害集上分别达到88.617%和74.668%,精确率分别达到了0.875和0.745,召回率分别达到0.874和0.738,F1值达到0.874和0.732,试验结果对比其他模型均有一定幅度的提升,验证了PD-Net模型在大规模多类别病虫害识别上的有效性。
