Prediction Of Component Content In Rare Earth Extraction Process Based On Hybrid Algorithm

In recent years,with the continuous breakthrough and popularization of artificial intelligence technology,it has been gradually applied to various industrial production processes.For the typical production industry,rare earth industry,as a rare resource,it can play an irreplaceable role in some high-tech development fields such as new energy and new materials,making it widely used in aerospace,defense and military industry,etc.Value.In addition,the rare earth industrial process has the characteristics of large delay,non-linear,time-varying,strong coupling,multi-variable and so on,which makes artificial intelligence technology more important.As a part of artificial intelligence,the soft measurement technology can solve some important process parameters that cannot be obtained or measured directly by sensors in the industrial process or the single measurement parameters obtained cannot fully reflect the operating conditions of the equipment and the content of rare earth elements.The problem of not being able to measure directly online.Based on this,on the basis of analyzing the current status of online measurement of rare earth elements and summarizing the deficiencies of traditional soft-sensing modeling methods,a method for predicting the content of components in rare earth extraction process based on hybrid algorithm is proposed.The specific research contents are as follows:1.To ensure the feasibility of the method in this paper,in view of the advantages of the RBF kernel function,such as good linear fitting performance and short training time,LSSVM uses this kernel function.By comparing with RBF neural network,the experimental results show that LSSVM has better performance than RBF neural network.2.The performance of LSSVM mainly depends on its penalty factors and kernel function parameters.The traditional parameter selection method is mainly manual selection(such as trial and error method,empirical method,etc.);and the swarm intelligence algorithm can automatically search for parameters,so that LSSVM performance meets a given Requirements.The traditional grey wolf optimization(GWO)algorithm is a swarm intelligence algorithm.When the objective function is a multimodal function problem,the linearization of the control distance parameter a in the algorithm can easily cause the algorithm to fall into local optimization.In order to solve this situation as much as possible,this paper uses an improved GWO algorithm to select LSSVM parameters.First,this paper reduces the possibility of the algorithm falling into local optimality by nonlinearizing the control distance parameter a;secondly,it introduces the differential evolution algorithm to mutate the population obtained in the later stage of the GWO algorithm,and mutates the mutated population with the GWO algorithm The later populations were compared,and thesimulation results proved the effectiveness of the proposed method.3.In order to improve the prediction performance of the prediction model as much as possible,the integrated learning Adaboost algorithm is introduced,the improved GWO-LSSVM is used as the weak learner in the integrated learning algorithm,and the prediction performance of the algorithm is improved by adjusting the number of weak learners.Simulation The results prove the effectiveness of the proposed method.The research results can provide an effective method for the automatic detection of the component content of the rare earth extraction production process,and are of great significance for promoting the comprehensive automation of the rare earth industry production process.