A convolutional neural network technique for online tracking of the radius evolution of levitating evaporating microdroplets of pure liquids, liquid mixtures and suspensions

In this study, we use a convolutional neural network ( which we trained on theoretically generated Mie scattering patterns, grouped into classes encompassing consecutively small ranges of radii) to classify corresponding experimentally recorded sequences of patterns, enabling the tracking of the radius evolution of evaporating microdroplets of pure liquids, liquid-liquid mixtures, and suspensions. The zip file contains 3 folders: "Convolutional Neural Networks" which contains the MATLAB script for the untrained convolutional neural network architecture, (t7g24d_Modified.m), the trained network (Stage1_Network_Pure_DEG.mat) and its corresponding class information. The class information is a ".mat" file containing 3 sub-vectors labelled "class" (class labels), "Range" (class radii ranges), and "Radius" (class average radii). Finally, the folder contains a MATLAB script "Classify_images.m" for classifying the experimental images sequentially. The 2 other folders contain images (generated for 3 classes and experimental images. The full length article can be accessed from Journal of Quantitative Spectroscopy and Radiative Transfer: https://doi.org/10.1016/j.jqsrt.2025.109533 Or from arxiv : https://doi.org/10.48550/arXiv.2410.08857本研究采用经理论生成的米氏散射模式（Mie scattering patterns）训练得到的卷积神经网络（Convolutional Neural Network），将散射模式划分为若干对应连续小半径区间的类别，对对应的实验记录的散射模式序列进行分类，以此实现纯液体、液液混合物及悬浮液的蒸发微液滴半径演化的追踪。本数据集压缩包包含3个文件夹：其一为"卷积神经网络（Convolutional Neural Networks）"文件夹，内含对应未训练卷积神经网络架构的MATLAB脚本t7g24d_Modified.m、训练完成的网络模型Stage1_Network_Pure_DEG.mat及其配套类别信息。类别信息为.mat格式文件，包含三个子向量，分别标记为class（类别标签）、Range（类别半径区间）与Radius（类别平均半径）。此外，该文件夹还包含用于按序分类实验图像的MATLAB脚本Classify_images.m。其余两个文件夹分别包含三类生成的散射模式图像与实验图像。本研究的完整论文可从《定量光谱学与辐射传输学报（Journal of Quantitative Spectroscopy and Radiative Transfer）》获取：https://doi.org/10.1016/j.jqsrt.2025.109533，亦可于arXiv平台获取：https://doi.org/10.48550/arXiv.2410.08857