Study on freezing separation mechanism of salt solution based on Micro-CT and deep learning
[Objective]With the rapid economic development,the shortage of water resources has become a universal problem in the world.The limited availability of freshwater resources forces humans to develop purification methods for saline water.The method of freezing consumes less energy and induces less corrosion because of low operating temperature;however,it has not been widely used commercially because of the relatively low desalination rate.This is because,during the saline progressive freezing process,water molecules crystalize and expel the salt into the unfrozen solution,while a part of the concentrated solution is inevitably trapped inside the nearby ice block to form salt inclusions,which reduces the desalination effect.Therefore,it is necessary to study the relationship between salt inclusions and salinity distribution,which is helpful in understanding the freezing desalination methods from a microscopic point of view and then explore the way to improve desalination efficiency.[Methods]In this paper,a NaCl solution sample with the original mass concentration of 6 wt%was frozen using a unidirectional heat transfer method under controllable conditions in the laboratory,and the freezing temperature was set at-20℃.The upper,middle,and lower parts of the 6wt%NaCl solution ice were scanned using the Sanying Nano Voxel-3000 micro-computed tomography(micro-CT)device with the pixel size setting of 6.65 µm.A deep learning(DL)model based on a 5-layer U-Net in the program Dragonfly was used to segment the micro-CT images.The area proportion of salt inclusions can be calculated,which is the ratio of the area occupied by salt inclusions in the region of interest(ROI)to the total area of the same region.In addition,a batch of samples was prepared using the same method to study the salinity distribution of the 6 wt%NaCl solution ice.Each sample was only cut once in a different position along the vertical direction,and the mass and salinity values were measured using an electronic balance and a salinometer.The samples were selected according to the criteria that the mass errors of both NaCl samples are less than 5%,and the ice sample was divided into 10 layers with equal mass and was numbered accordingly from top to bottom.The average salinity values of the layers were calculated according to the fitting curve between the salt content and the mass percentage.The area proportions of salt inclusions and the salinity distribution were also compared.[Results]The results showed the following.First,the area proportion of salt inclusions in the upper part of the 6 wt%NaCl solution ice decreased from the highest value of 65.67%to the lowest value of 8.51%and then gradually increased from 8.51%to 13.11%.Meanwhile,the area proportion of salt inclusions in the middle part of the ice sample increased from 17.45%to 21.24%,and in the lower part,it slowly increased from 20.80%to 22.20%.Second,on the top of the NaCl solution ice sample,the existence of a high salinity thin layer can be"seen,"and the measurement proved that its salinity is much higher than that of the original solution.The salinity of the second ice layer greatly decreased to the minimum value,whereas the salinity from the third ice layer to the 10th layer gradually increased.Third,the area proportion variation tendency of salt inclusions from top to bottom is relative to the ice layer salinity of the sample.[Conclusions]By comparing the area proportion of salt inclusions and the salinity distribution of the ice sample,it can be found that the area proportion variation tendency of salt inclusions is consistent with the ice salinity,which proves that the internal microstructure of NaCl solution ice is related to salinity distribution,and the combination of micro-CT technology and deep learning is an effective method to visually and quantitatively analyze the internal structure of salt solution ice.These findings will help to improve the understanding of freezing desalination methods from a microscopic point of view and then explore ways to improve desalination efficiency.
NaCl solution icedeep learningsalt inclusionssalinity distribution
万方数据
维普
