Losses in channel flows are usually determined using a frictional head loss parameter. Fluid friction is however not the only source of loss in channel flows with heat transfer. For such flow problems, thermal energy ...Losses in channel flows are usually determined using a frictional head loss parameter. Fluid friction is however not the only source of loss in channel flows with heat transfer. For such flow problems, thermal energy degradation, in addition to mechanical energy degradation, add to the total loss in thermodynamic head. To assess the total loss in a channel with combined convection and radiation heat transfer, the conventional frictional head loss parameter is extended in this study. The analysis is applied to a 3D turbulent channel flow and identifies the critical locations in the flow domain where the losses are concentrated. The influence of Boltzmann number is discussed, and the best channel geometry for flows with combined heat transfer modes is also determined.展开更多
针对现有的风格迁移方法在对水表进行数据增强后导致颜色失真,内容保留不完整等问题,提出了一种基于大卷积核的任意风格迁移算法(arbitrary style transfer algorithm of large convolutional kernel,LKAST)。首先,针对风格图像使用大...针对现有的风格迁移方法在对水表进行数据增强后导致颜色失真,内容保留不完整等问题,提出了一种基于大卷积核的任意风格迁移算法(arbitrary style transfer algorithm of large convolutional kernel,LKAST)。首先,针对风格图像使用大卷积核提取风格特征,保留风格特征的高层特征;此外,通过引入新的损失函数,更好的保留迁移结果对内容的保留;最后,通过两组对照实验验证方法的有效性。实验结果表明,该方法能够在模拟水表现场环境的同时保留足够的内容信息,在仅改变数据增强算法的前提下,单次多框目标检测(SSD)算法准确率提升6.84%,YOLOv5准确率提升6.56%。展开更多
文摘Losses in channel flows are usually determined using a frictional head loss parameter. Fluid friction is however not the only source of loss in channel flows with heat transfer. For such flow problems, thermal energy degradation, in addition to mechanical energy degradation, add to the total loss in thermodynamic head. To assess the total loss in a channel with combined convection and radiation heat transfer, the conventional frictional head loss parameter is extended in this study. The analysis is applied to a 3D turbulent channel flow and identifies the critical locations in the flow domain where the losses are concentrated. The influence of Boltzmann number is discussed, and the best channel geometry for flows with combined heat transfer modes is also determined.
文摘针对现有的风格迁移方法在对水表进行数据增强后导致颜色失真,内容保留不完整等问题,提出了一种基于大卷积核的任意风格迁移算法(arbitrary style transfer algorithm of large convolutional kernel,LKAST)。首先,针对风格图像使用大卷积核提取风格特征,保留风格特征的高层特征;此外,通过引入新的损失函数,更好的保留迁移结果对内容的保留;最后,通过两组对照实验验证方法的有效性。实验结果表明,该方法能够在模拟水表现场环境的同时保留足够的内容信息,在仅改变数据增强算法的前提下,单次多框目标检测(SSD)算法准确率提升6.84%,YOLOv5准确率提升6.56%。
文摘针对不同型号滚动轴承监测信号之间特征分布差异大、故障数据样本少,导致轴承故障精度低的问题,提出了一种基于改进交替迁移学习的滚动轴承故障诊断算法。为了充分发挥卷积神经网络(convolutional neural network, CNN)对二维数据优秀的特征提取能力,首先将一维振动信号转化为二维图像,输入到深度卷积神经网络中学习;其次,为了减少源域与目标域数据间的特征分布差异,提出了改进的交替迁移学习(improved alternately transfer learning, IATL),通过交替计算域间的CORAL损失函数和最大均值差异(maximum mean discrepancy, MMD)损失函数,并反向传播更新各层网络权重与偏置参数,以实现变工况、跨轴承型号和小故障样本条件下轴承特征迁移适配;最后,在全连接层使用Softmax函数对目标域数据进行故障诊断。为了验证该算法的有效性,采用凯斯西储大学(Case Western Reserve University, CWRU)的滚动轴承数据集进行了迁移试验验证。结果表明,与仅计算CORAL损失函数和MMD损失函数等算法对比可知,该算法有效地减少了领域数据之间的特征分布差异,具有较高的故障分类准确率。
