Detection Of Component Content In Pr/Nd Extraction Process Based On Its Ion’s Color Feature

China is one of the countries in the world which are rich in rare earth resources,and with the theory of counter current extraction, the rare earth products satisfied thepurity requirement can be gained from a mixed solution of rare earth. Under theguidance of this powerful theory, the scale of China’s rare earth production andproduct yields have been among the world’s largest. However, there still exist someproblems such as offline analysis, experience control and manual adjustment in theextraction and separation process in China’s rare earth industries, resulting in hugeresource consumption, serious environmental pollution, low production efficiency andunstable product quality. This paper aims at the problem of online detection of theelement component content in the extraction process, and takes the Pr/Nd extractionand separation process in which the ion of Pr and Nd have characteristic color as anexample, then proposes a method based on ion’s color feature to detect componentcontent:1) The standard solutions are prepared in the lab to establish a color library bythe solutions’ images. The mean of the H component respectively extracted from theimages shot in the lab in the HSI color space is taken as the color feature, and then astandard model is established by the component content and the mean of the Hcomponent data corresponding to the images shot in lab.2) The images shot in workshop are collected to establish a library, and the meanof the H component are extracted from the above images. Then the mean of the Hcomponent data of the images shot in the lab and workshop corresponding to the samecomponent content are compared with, and also the reason why there are different isanalyzed. At last, this paper establishes a compensation model to reduce the errorbetween the mean of the H component corresponding to each image shot in the laband workshop.3) We make use of the compensation model to revise the H component content,and then combine the revised data with the standard model to predict the componentcontent corresponding to each image shot in the workshop. The simulation result shows that when the component content of Pr is bigger than10%, its relative error isbelow6%, which satisfies the requirements of the workshop for monitoring andcontrolling.In the end, we combine the image retreival based on color feature with thecomponent content and the mean of H component data in the standard liabray topredict the component content corresponding to each image shot in the workshop,then devise an interface of component conten detection to make the whole processexplicitness and visualizable.