| Rare-earth,called “Industrial vitamin”,is great and indispensable help for the development of science and technology.As extraction and separation process of rare earth is a complex system with long process,multi-variable and nonlinear characteristics,the whole production process is influenced by external interference and internal uncertain factors of the extraction process,which caused unsteadiness for control system and damaged quality of rare earth products.So it means a lot to ensure quality of rare earth products to study treatment method of uncertain factors during the rare earth extraction process.This paper considers the cerium praseodymium/neodymium extraction and separation process as the research object,regards the unmodeled dynamics in the system operation process fully.First,the rare earth extraction process is described by combining linear model and unmodeled dynamics.Then,a predictive controller is devised to achieve the tracking of the monitoring level component content to the set value,and a compensator is designed to eliminate the impact of unmodeled dynamic terms on the system as possible to ensure that the rare earth extraction solution obtained at the outlets at both ends of the extraction and separation process meets the process requirements.The research contents are as follows:1.In view of the nonlinear characteristics of the extraction and separation process of rare earth,the accuracy of the established model could not be guaranteed.The first-order Taylor expansion was carried out near the steady-state operating point to establish a nonlinear model describing the extraction and separation process with the sum of linear terms and unmodeled dynamic terms.In the above nonlinear model,gradient identification method is used to identify parameters for linear terms,and extreme learning machine is used to estimate unmodeled dynamic terms.Moreover,slime mold algorithm is used to optimize the initial weight and threshold of extreme learning machine to alleviate the problem that the neural network is prone to fall into local optimum,which affects the accuracy of unmodeled dynamic estimation values.2.The generalized predictive controller is designed on the basis of the above nonlinear model,and the nonlinear term compensator based on the slime mold algorithm-extreme learning machine is designed to eliminate the influence of the unmodeled dynamics on the system as far as possible.The experimental comparison and analysis with the PID controller commonly used in industry show that the predictive controller based on the unmodeled dynamics compensation has better control performance.3.Considering the coupling effect between variables,a decoupling prediction controller based on unmodeled dynamic compensation was designed,and the error weight matrix was introduced to further reduce the impact of variable coupling on the system.Compared with the controller without decoupling link,the results show that the decoupling controller has less adjustment to the extractant and detergent flow and better control performance.To sum up,for the cerium praseodymium/neodymium extraction and separation system,the nonlinear model of dynamic addition of linear model and unmodeled model is used to describe the extraction and separation process of rare earth.The predictive controller based on unmodeled dynamic compensation and decoupling predictive controller designed on the basis of the nonlinear model are effective and has guiding significance for the implementation of automatic control in rare earth separation enterprises. |
