Component Content Multilple Model Predictive Control For CePr/Nd Cascade Extraction Process

With more and more extensive scope of application of rare earth, rare earth has become an importantand indispensable strategic resources. In recent years, with China’s rapid development of rare earth industry,the production of rare earth has reached an unprecedented scale, but the flow is regulated and componentcontent is detected by manpower, the production efficiency and product quality are seriously affected.Therefore, the establishment of model of rare earth extraction separation process, and the design of thecontroller, has become a key step can achieve automation of rare earth extractionThe rare earth extraction process is a strong nonlinear process with the existence of a large number ofrandom, uncertain information, thus increasing the difficulty of modeling and control of the extractionprocess. Based on dynamic characteristic curves of component content of "simulation separatory funnelmethod" and operation data, the model of rare earth extraction process is established using data drivenmodeling algorithm and predictive controller is design in this paper. The concrete research contents are asfollows:1. By taking CePr/Nd extraction process as an example, according to theory of countercurrent extractionthe technological parameters of CePr/Nd extraction separation is design and separation index is determined.Then we use MATLAB to simulate the experiment of a separatory funnel cascade extraction. Componentcontent dynamic characteristics curve of CePr/Nd extraction is obtained.2. The model sets of the rare earth extraction process are establishded using data driven modelingalgorithm. This model sets describe the nonlinear dynamic characteristics of rare earth extraction process.Firstly we determine the number of sub models by using subtractive clustering algorithm, secondly accordingto the least square algorithm we determine the model parameters, Finaly according to the operation processof different conditions we select the most matching sub model as the current model. In order to verify thevalidity of the proposed model set, the component content multi model generalized predictive controlalgorithm is proposed and the curve of component content and flow of the monitoring stage is obtained.3. In order to verifying the effectiveness and applicability of the proposed method, we select the CePr/Ndextraction separation process as the object of study. The simulation results show that the proposed componentcontent multi model predictive control algorithm can guarantee the ends of monitoring stage componentcontent meet the requirements, extracting and scrubbing flow can be adjusted in real time, and realizes theautomatic control of the extraction separation process.