Multi-component Prediction Of Rare Earth Mixed Solution Based On GA-ELM

Rare earth elements have unique physical properties such as photoelectromagnetism,and so on.With a small amount of rare earth elements added to industrial products,the quality and performance of products can be greatly improved,with the effect of "turning stone into gold".At present,China’s rare earth separation enterprises mainly use solvent cascade extraction technology to obtain single,high-purity rare earth elements from rare earth symbiotic ore.This method has the characteristics of long process,strong coupling,and large hysteresis,which makes it difficult to achieve automatic control.The market competition requires rare earth companies to take into account both product quality and production efficiency,and efficient detection of rare earth component content distribution is a prerequisite for realizing rare earth extraction process parameter adjustment and automatic control,so it is imperative to study the rapid and accurate detection method of multi-component content in the process of rare earth extraction and separation.In this paper,in order to solve the problem that the component content of rare earth mixed solutions with coexisting color characteristics and non-color characteristics ions is difficult to quickly and accurately detect.Under the conditions of HSI single color space and mixed color space,the multi-component content prediction method of rare earth mixed solution based on genetic algorithm optimized extreme learning machine(GA-ELM)was studied.The main research contents are as follows:1.The experimental comparative analysis method was used to determine the image characteristics of the multi-component rare earth mixed solution.First by experimental,the information of praseodymium / neodymium(Ce Pr / Nd)mixed solution image with color features and non-color feature coexistence is compared with the image of the praseodymium /neodymium(Pr / Nd)mixed solution image with both color features.The difference in the image characteristics of the two rare earth mixed solutions is determined,and then the first-order moments of the H and S components,which are monotonic with the Pr and Nd component content,are used as auxiliary variables of the component content model.2.Using GA-ELM to establish a multi-component content model of rare earth mixed solution.In view of the extreme learning machine(ELM)has the advantages of less parameter settings,high accuracy,fast learning speed,good generalization performance,and the shortcomings of the ELM model initial weights and thresholds are uncertain.The genetic algorithm(GA)was used to optimize the model parameters.The Pr-Nd multi-element content model based on GA-ELM was established,and the effectiveness of the method was verified by the extraction field data.3.In order to make up for the shortcomings of a single color space,a method forpredicting the multicomponent content of rare earth extraction based on mixed color space was studied.First,the first-order moments of the feature components of the rare earth mixed solution image in the HSI,HSV,and YUV color spaces are fused,and then the GA-ELM based on the fused color feature component,the relative red component,and the color vector angle as auxiliary variables.The Pr and Nd multi-element component content models were tested by extraction field data.The results show that the prediction accuracy of the multi-component content model has been further improved.In summary,the mixed solution of rare-earth extraction tank containing color features and non-color feature ions coexist as the research object.The multi-element content prediction model established by GA optimization of ELM parameters can be applied to rapid detection of multi-component content in rare earth extraction process.