期刊文献+
共找到2篇文章
< 1 >
每页显示 20 50 100
Diffusion Process of High Concentration Spikes in a Quasi-Homogeneous Turbulent Flow
1
作者 Masaya Endo Qianqian Shao +1 位作者 Takahiro Tsukahara Yasuo Kawaguchi 《Open Journal of Fluid Dynamics》 2016年第4期371-390,共21页
When a mass spreads in a turbulent flow, areas with obviously high concentration of the mass compared with surrounding areas are formed by organized structures of turbulence. In this study, we extract the high concent... When a mass spreads in a turbulent flow, areas with obviously high concentration of the mass compared with surrounding areas are formed by organized structures of turbulence. In this study, we extract the high concentration areas and investigate their diffusion process. For this purpose, a combination of Planar Laser Induced Fluorescence (PLIF) and Particle Image Velocimetry (PIV) techniques was employed to obtain simultaneously the two fields of the concentration of injected dye and of the velocity in a water turbulent channel flow. With focusing on a quasi-homogeneous turbulence in the channel central region, a series of PLIF and PIV images were acquired at several different downstream positions. We applied a conditional sampling technique to the PLIF images to extract the high concentration areas, or spikes, and calculated the conditional-averaged statistics of the extracted areas such as length scale, mean concentration, and turbulent diffusion coefficient. We found that the averaged length scale was constant with downstream distance from the diffusion source and was smaller than integral scale of the turbulent eddies. The spanwise distribution of the mean concentration was basically Gaussian, and the spanwise width of the spikes increased linearly with downstream distance from the diffusion source. Moreover, the turbulent diffusion coefficient was found to increase in proportion to the spanwise distance from the source. These results reveal aspects different from those of regular mass diffusion and let us conclude that the diffusion process of the spikes differs from that of regular mass diffusion. 展开更多
关键词 Turbulent Transport High Concentration Spikes Quasi-Homogeneous Turbulent Flow conditional sampling technique PIV and PLIF Measurements Passive Scalar Diffusion
下载PDF
基于元学习技术的变工况齿轮故障诊断方法
2
作者 郭敏 周超 +4 位作者 郑成基 陈鹏 胡国宾 范青荣 朱小红 《机电工程》 CAS 北大核心 2023年第11期1682-1690,共9页
在变工况齿轮故障诊断过程中,存在齿轮运行工况多变、故障样本数据少、数据分布性差异大和故障数据非均衡性等问题,导致传统的深度学习模型通用性差、诊断准确率不高。针对这些问题,提出了一种基于元学习技术的变工况齿轮故障诊断(VWFD... 在变工况齿轮故障诊断过程中,存在齿轮运行工况多变、故障样本数据少、数据分布性差异大和故障数据非均衡性等问题,导致传统的深度学习模型通用性差、诊断准确率不高。针对这些问题,提出了一种基于元学习技术的变工况齿轮故障诊断(VWFD)方法(模型)。首先,采用重叠采样技术,对齿轮的原始振动信号进行了重采样,增加了故障样本的数量;其次,对重采样的故障数据进行了短时傅里叶变换(STFT),将其转化为时频特征图,使其数据形式更加符合模型的输入,以便于后续提取更完善的故障特征;然后,将Inception模块引入到基于元学习技术的原型网络中,以提高其特征表达能力,获取更加全面的齿轮故障特征信息;最后,基于优化的原型网络,建立了各类故障的度量类原型,采用度量分类器进行了故障分类,对变工况下的齿轮故障进行了诊断;为了验证VWFD模型结构与Inception模块引入位置和数量的合理性,设计了一系列对比实验,并对实验结果进行了分析。研究结果表明:与采用其他故障诊断方法得到的结果相比,采用VWFD方法所得到的诊断精度更高,如在相同负载、不同转速变工况类型下的5-way 5-shot实验组中,VWFD的平均诊断精度高达91.26%,而支持向量机(SVM)、卷积神经网络(CNN)和原型网络(PN)的诊断精度分别仅有74.48%、87.22%和89.56%。 展开更多
关键词 变工况齿轮故障诊断 重叠采样技术 元学习技术 原型网络 短时傅里叶变换 Inception模块
下载PDF
上一页 1 下一页 到第
使用帮助 返回顶部