arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2006.06593physics.flu-dyn
收藏

碰撞液滴的拉格朗日跟踪

Lagrangian tracking of colliding droplets

Reece Vincent Kearney, Gregory Paul Bewley

摘要:

我们介绍了一种新的拉格朗日粒子跟踪算法,跟踪粒子在三维接近接触的轨迹之间的分离。该算法还检测低韦伯数二进制碰撞,导致聚结以及液滴破裂。通过找到一组描述每个物体边缘的圆圈,在二维高分辨率数字图像中识别颗粒。这允许识别在投影中重叠超过80%的粒子,即使对于噪声图像也是如此,并且无需调用额外的时间数据。该算法通过最小化惩罚函数来从三维粒子坐标构建轨迹,该惩罚函数是使用来自四个时刻的信息的与预期粒子坐标的偏差的加权和。这种新的混合算法对合成数据进行了验证,并发现完美地再现更多的轨迹比其他常用的方法。对于平均移动距离小于其最近邻距离十分之一的粒子,碰撞检测的准确率为95%。

英文摘要:

We introduce a new Lagrangian particle tracking algorithm that tracks particles in three dimensions to separations between trajectories approaching contact. The algorithm also detects low Weber number binary collisions that result in coalescence as well as droplet break-up. Particles are identified in two-dimensional high-resolution digital images by finding sets of circles to describe the edge of each body. This allows identification of particles that overlap in projection by over 80% even for noisy images and without invoking additional temporal data. The algorithm builds trajectories from three-dimensional particle coordinates by minimizing a penalty function that is a weighted sum of deviations from the expected particle coordinates using information from four moments in time. This new hybrid algorithm is validated against synthetic data and found to perfectly reproduce more trajectories than other commonly used methods. Collisions are detected with 95% accuracy for particles that move on average less than one tenth the distance to their nearest neighbor.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑