针对目前基于深度学习的带式输送机煤量检测算法参数量大、难以部署于边缘计算设备及缺乏定量检测的问题,提出一种基于改进DeepLabv3+的带式输送机煤量检测方法。将轻量化网络MobileNetV2作为DeepLabv3+的骨干网络进行特征提取,在尽可...针对目前基于深度学习的带式输送机煤量检测算法参数量大、难以部署于边缘计算设备及缺乏定量检测的问题,提出一种基于改进DeepLabv3+的带式输送机煤量检测方法。将轻量化网络MobileNetV2作为DeepLabv3+的骨干网络进行特征提取,在尽可能保证分割精度的同时提高计算速度;针对煤流与输送带的方向性特征及输送带像素边缘呈现细长条状结构特征,采用条状空洞空间金字塔池化(SASPP)进行加强,并将SASPP模块、1×1卷积与残差结构进行融合,得到CA−SASPP,加强深层次的特征提取;结合CBAM注意力机制实现对特征图中关键信息的加权聚焦。实验结果表明,改进DeepLabv3+模型在平均分割精度仅下降0.36%的情况下,参数量减少了85.58%,推理速度提升至113帧/s,较原方法提高了12帧/s,在保持与原模型相当的分割精度的同时,实现了显著的轻量化效果。基于语义分割结果,通过计算煤量与输送带区域的面积占比,实现了煤量的定量检测,为多级带式输送机的智能调速提供了理论依据。将改进DeepLabv3+模型通过TensorRT加速并部署至Jetson Orin Nano边缘计算设备,实现了煤流图像的实时处理与分析,降低了云端服务器的计算负担,满足了工业现场对实时性和准确性的需求。展开更多
飞机积冰严重威胁飞行安全,为更好地了解飞机积冰的气象条件和云微物理特征,寻找积冰与宏观气象条件阈值之间的关系,本文利用飞机观测资料,结合雷达、欧洲中期天气预报中心ERA5再分析资料,对2018年12月10日河南中东部冬季一次飞机严重...飞机积冰严重威胁飞行安全,为更好地了解飞机积冰的气象条件和云微物理特征,寻找积冰与宏观气象条件阈值之间的关系,本文利用飞机观测资料,结合雷达、欧洲中期天气预报中心ERA5再分析资料,对2018年12月10日河南中东部冬季一次飞机严重积冰过程的天气条件和云微物理过程进行了分析,并对此次积冰强度进行了评估。此次积冰过程是受高空槽、低层切变以及地面冷高压的共同作用,流向河南的低层暖湿空气沿着冷高压爬升,形成上暖下冷的逆温层,促进了该区域内液态水的积聚。积冰区域内液态水含量丰富,而冰水含量较低,液态水含量主要分布在550~750 hPa高度之间。飞机观测数据与ERA5再分析资料结果较为一致,强积冰区在3630 m高度层上,云中温度为-8.2℃~-6.8℃,相对湿度在87%~92%,平均云滴数浓度为499.9 cm^(-3),液态水含量最大可达0.87 g m^(-3),该区域内大小过冷液滴共存,基本无冰晶粒子。通过计算:飞机穿过液态水含量丰富区发生轻度积冰、中度积冰、严重积冰时的液态水含量阈值分别为0.05、0.20和0.58 g m^(-3)。当逆温层强度为1.9℃(100 hPa)^(-1)、垂直运动速度在-0.25~0.2 Pa s^(-1)、水平辐合强度为-3.73 kg m^(-2)s^(-1)的情况下,积冰强度达到中度以上的概率为77%。展开更多
Dynamical Joining of the solid-state metal is the key technology to realize endless hot rolling. The heating and laser welding method both require long joining time. Based on super deformation method, a 7-bar and 2-sl...Dynamical Joining of the solid-state metal is the key technology to realize endless hot rolling. The heating and laser welding method both require long joining time. Based on super deformation method, a 7-bar and 2-slider mechanism was developed in Japan, and the joining time is less than 0.5 s, however the length of each bar are not reported and this mechanism is complex. A relatively simple 6-bar and 1-slider mechanism is put forward, which can realize the shearing and extrusion motion of the top and bottom blades with a speed approximately equal to the speed of the metal plates. In order to study the kinematics property of the double blades, based on complex vector method, the multi-rigid-body model is built, and the displacement and speed functions of the double blades, the joining time and joining thickness are deduced, the kinematics analysis shows that the initial parameters can＇t satisfy the joining process. Hence, optimization of this mechanism is employed using genetic algorithm（GA） and the optimization parameters of this mechanism are obtained, the kinematics analysis show that the joining time is less than 0.1 s, the joining thickness is more than 80% of the thickness of the solid-state metal, and the horizontal speeds of the blades are improved. A new mechanism is provided for the joining of the solid-state metal and a foundation is laid for the design of the device.展开更多
The investments of the electro-hydraulic servo system of the mold non-sinusoidal oscillator are great, the modification ratio of the mechanical type is unable to be adjusted online, and some continuous casters suffer ...The investments of the electro-hydraulic servo system of the mold non-sinusoidal oscillator are great, the modification ratio of the mechanical type is unable to be adjusted online, and some continuous casters suffer from server resonance during the casting. A mold non-sinusoidal oscillation mechanism driven by servomotor is proposed and the prototype is produced in the lab, the investment is low and the modification ratio is can be adjusted online, and the stability problem is studied. At first the dynamics model of the servomotor non-sinusoidal oscillation is established, and the kinematics differential function is deduced. Furthermore, based on the harmonic balance method, the eigenvalues of the system are solved; the criterion of the stability of the system is put forward. In addition, the eigenvalues and harmonic with different oscillating parameters are analyzed. Analytical results show that the real parts of the eigenvalues are positive, the system will be unstable, and the resonance will occur when the positive real parts of the eigenvalues are extremum. A foundation is established for solving the running smooth problem and next application of this mechanism.展开更多
基于2024年5月10日和20日在丹江口水库汇水区开展的针对气溶胶和积层混合云的两次飞机观测,利用机载探测资料,分析了水库汇水区气溶胶和云微物理量的垂直分布特征。结果表明,气溶胶数浓度总体随高度的增加呈下降趋势,与白天相比,夜晚近...基于2024年5月10日和20日在丹江口水库汇水区开展的针对气溶胶和积层混合云的两次飞机观测,利用机载探测资料,分析了水库汇水区气溶胶和云微物理量的垂直分布特征。结果表明,气溶胶数浓度总体随高度的增加呈下降趋势,与白天相比,夜晚近地层气溶胶浓度更高,最高达4000 cm^(-3)。中低层气溶胶中值体积直径约为0.2~0.3μm,有效直径约为0.6~0.8μm。气溶胶粒子主要集中在小粒径段,粒子谱为多峰分布,第一峰值直径为0.14μm。云内气溶胶浓度基本都略低于云外,而更强的吸湿增长作用导致气溶胶粒径更大。气团后向轨迹分析结果表明,近地层气溶胶粒子主要来自局地排放,高层粒子主要来自远距离传输,5月10日观测点周围人为排放产生的污染性细颗粒物是造成近地层气溶胶浓度高于5月20日的主要原因。两次观测的积层混合云内的云滴有效直径整体均在25μm以内,云滴数浓度最大值分别为118 cm^(-3)和205 cm^(-3),5月10日和20日云内液水富集层分别在6100 m和5500 m左右,液水含量最大值分别为0.44 g m^(-3)和0.78 g m^(-3)。由于气溶胶粒子竞争水汽的作用,云底附近的较高气溶胶浓度使得云滴浓度出现高值,同时云滴直径减小。小云滴阶段粒子谱为单峰分布,不同高度的峰值粒径略有差异,主要集中在11~25μm,雨滴和冰晶阶段粒子谱转变为多峰分布。展开更多
文摘针对目前基于深度学习的带式输送机煤量检测算法参数量大、难以部署于边缘计算设备及缺乏定量检测的问题,提出一种基于改进DeepLabv3+的带式输送机煤量检测方法。将轻量化网络MobileNetV2作为DeepLabv3+的骨干网络进行特征提取,在尽可能保证分割精度的同时提高计算速度;针对煤流与输送带的方向性特征及输送带像素边缘呈现细长条状结构特征,采用条状空洞空间金字塔池化(SASPP)进行加强,并将SASPP模块、1×1卷积与残差结构进行融合,得到CA−SASPP,加强深层次的特征提取;结合CBAM注意力机制实现对特征图中关键信息的加权聚焦。实验结果表明,改进DeepLabv3+模型在平均分割精度仅下降0.36%的情况下,参数量减少了85.58%,推理速度提升至113帧/s,较原方法提高了12帧/s,在保持与原模型相当的分割精度的同时,实现了显著的轻量化效果。基于语义分割结果,通过计算煤量与输送带区域的面积占比,实现了煤量的定量检测,为多级带式输送机的智能调速提供了理论依据。将改进DeepLabv3+模型通过TensorRT加速并部署至Jetson Orin Nano边缘计算设备,实现了煤流图像的实时处理与分析,降低了云端服务器的计算负担,满足了工业现场对实时性和准确性的需求。
文摘飞机积冰严重威胁飞行安全,为更好地了解飞机积冰的气象条件和云微物理特征,寻找积冰与宏观气象条件阈值之间的关系,本文利用飞机观测资料,结合雷达、欧洲中期天气预报中心ERA5再分析资料,对2018年12月10日河南中东部冬季一次飞机严重积冰过程的天气条件和云微物理过程进行了分析,并对此次积冰强度进行了评估。此次积冰过程是受高空槽、低层切变以及地面冷高压的共同作用,流向河南的低层暖湿空气沿着冷高压爬升,形成上暖下冷的逆温层,促进了该区域内液态水的积聚。积冰区域内液态水含量丰富,而冰水含量较低,液态水含量主要分布在550~750 hPa高度之间。飞机观测数据与ERA5再分析资料结果较为一致,强积冰区在3630 m高度层上,云中温度为-8.2℃~-6.8℃,相对湿度在87%~92%,平均云滴数浓度为499.9 cm^(-3),液态水含量最大可达0.87 g m^(-3),该区域内大小过冷液滴共存,基本无冰晶粒子。通过计算:飞机穿过液态水含量丰富区发生轻度积冰、中度积冰、严重积冰时的液态水含量阈值分别为0.05、0.20和0.58 g m^(-3)。当逆温层强度为1.9℃(100 hPa)^(-1)、垂直运动速度在-0.25~0.2 Pa s^(-1)、水平辐合强度为-3.73 kg m^(-2)s^(-1)的情况下,积冰强度达到中度以上的概率为77%。
基金Supported by National Natural Science Foundation of China(Grant No.51475139)
文摘Dynamical Joining of the solid-state metal is the key technology to realize endless hot rolling. The heating and laser welding method both require long joining time. Based on super deformation method, a 7-bar and 2-slider mechanism was developed in Japan, and the joining time is less than 0.5 s, however the length of each bar are not reported and this mechanism is complex. A relatively simple 6-bar and 1-slider mechanism is put forward, which can realize the shearing and extrusion motion of the top and bottom blades with a speed approximately equal to the speed of the metal plates. In order to study the kinematics property of the double blades, based on complex vector method, the multi-rigid-body model is built, and the displacement and speed functions of the double blades, the joining time and joining thickness are deduced, the kinematics analysis shows that the initial parameters can＇t satisfy the joining process. Hence, optimization of this mechanism is employed using genetic algorithm（GA） and the optimization parameters of this mechanism are obtained, the kinematics analysis show that the joining time is less than 0.1 s, the joining thickness is more than 80% of the thickness of the solid-state metal, and the horizontal speeds of the blades are improved. A new mechanism is provided for the joining of the solid-state metal and a foundation is laid for the design of the device.
基金Supported by National Natural Science Foundation of China and Baosteel Group Co.Ltd.(Grant No.U1260203)Natural Science Foundation Steel and Iron Foundation of Hebei Province,China(Grant No.F2013203291)+1 种基金Doctor Startup Foundation of Hebei University of Science and Technology,China(Grant No.QD201247)Foundation of Hebei University of Science and Technology,China(Grant No.XL201004)
文摘The investments of the electro-hydraulic servo system of the mold non-sinusoidal oscillator are great, the modification ratio of the mechanical type is unable to be adjusted online, and some continuous casters suffer from server resonance during the casting. A mold non-sinusoidal oscillation mechanism driven by servomotor is proposed and the prototype is produced in the lab, the investment is low and the modification ratio is can be adjusted online, and the stability problem is studied. At first the dynamics model of the servomotor non-sinusoidal oscillation is established, and the kinematics differential function is deduced. Furthermore, based on the harmonic balance method, the eigenvalues of the system are solved; the criterion of the stability of the system is put forward. In addition, the eigenvalues and harmonic with different oscillating parameters are analyzed. Analytical results show that the real parts of the eigenvalues are positive, the system will be unstable, and the resonance will occur when the positive real parts of the eigenvalues are extremum. A foundation is established for solving the running smooth problem and next application of this mechanism.
文摘基于2024年5月10日和20日在丹江口水库汇水区开展的针对气溶胶和积层混合云的两次飞机观测,利用机载探测资料,分析了水库汇水区气溶胶和云微物理量的垂直分布特征。结果表明,气溶胶数浓度总体随高度的增加呈下降趋势,与白天相比,夜晚近地层气溶胶浓度更高,最高达4000 cm^(-3)。中低层气溶胶中值体积直径约为0.2~0.3μm,有效直径约为0.6~0.8μm。气溶胶粒子主要集中在小粒径段,粒子谱为多峰分布,第一峰值直径为0.14μm。云内气溶胶浓度基本都略低于云外,而更强的吸湿增长作用导致气溶胶粒径更大。气团后向轨迹分析结果表明,近地层气溶胶粒子主要来自局地排放,高层粒子主要来自远距离传输,5月10日观测点周围人为排放产生的污染性细颗粒物是造成近地层气溶胶浓度高于5月20日的主要原因。两次观测的积层混合云内的云滴有效直径整体均在25μm以内,云滴数浓度最大值分别为118 cm^(-3)和205 cm^(-3),5月10日和20日云内液水富集层分别在6100 m和5500 m左右,液水含量最大值分别为0.44 g m^(-3)和0.78 g m^(-3)。由于气溶胶粒子竞争水汽的作用,云底附近的较高气溶胶浓度使得云滴浓度出现高值,同时云滴直径减小。小云滴阶段粒子谱为单峰分布,不同高度的峰值粒径略有差异,主要集中在11~25μm,雨滴和冰晶阶段粒子谱转变为多峰分布。
