Research On Batch Process Operation Optimization Method Based On Process Transfer Model

Batch process is an important processing method in modern industrial production.As an effective way to improve the product quality of batch process,operation optimization is of great significance.The rapid development of social economy and the increasing complexity of industrial manufacturing process also put forward higher requirements for the operation optimization of batch process.Batch process optimization methods based on data-driven models usually rely on sufficient and complete data information to establish reliable process models.For the new batch process which is just put into operation,the insufficient modeling data will lead to the inefficiency of modeling,and then affect the implementation of the subsequent optimization strategy.However,there are many similar processes in the actual industrial process,and these similar processes have rich data information of the old process.Using the process migration strategy to migrate these data to the new process can assist the modeling and optimization of the new process.In this paper,based on the process transfer model(PTM),the operation optimization methods of batch process are studied,the main contents of this paper are as follows:(1)Aiming at the problem that it is difficult to model the new batch process due to insufficient data,the Joint-Y partial least squares(JYPLS)method is used to construct the PTM,and on this basis,an optimization compensation method for the new batch process is proposed.The difference between similar batch processes and the uncertainty of PTM will aggravate the model-plant mismatch,which leads to the mismatch of the necessary conditions of optimality(NCO)of the model optimization problem.The modifier-adaptation(MA)method can effectively solve the NCO mismatch problem,and the self-tuning(ST)method can further compensate the suboptimal results of batch optimization,making the optimization results closer to the optimal value.Therefore,the optimization effect is improved by combining MA optimization and ST optimization.Finally,the effectiveness of the proposed method is verified by simulation of the crystallization process of cobalt oxalate.(2)The optimization method described in research content 1 belongs to the batchto-batch optimization method,which can effectively solve the disturbance information between batches,but it is difficult to effectively solve the mismatch and interference problems existing in the operation period of a single batch.To solve this problem,an integrated optimization method based on PTM is proposed.On the basis of MA strategy,MCC strategy is introduced.MCC divides the operation period of a single batch through several decision points,and optimizes the operation at the decision points to overcome the disturbance problem in the batch,so as to further improve the optimization effect.Finally,the effectiveness of the proposed method is verified by the simulation of cobalt oxalate crystallization process.(3)When the new process data is seriously scarce,it will be difficult to establish PTM.In order to effectively generate new process modeling data set and reasonably control modeling cost,and realize efficient optimization of new process,a hierarchical optimization method based on PTM is proposed.At the upper level of hierarchy,the design of dynamic experiments(Do DE)method is used to generate data sets systematically to build response surface model(RSM),then further improve the operation trajectory by using MA strategy,and take the final optimization result as the initial value of the operation variable of the lower level optimization.At the lower level of hierarchy,the JYPLS model is established by combining the new process data information accumulated by the upper level optimization and the historical data of similar old process to realize the information transmission from the upper level to the lower level optimization,and then the ST strategy is used to iteratively modify the trajectory of the operation variables.Finally,the effectiveness of the proposed method is verified by the simulation of cobalt oxalate crystallization process.The thesis includes 20 figures,2 tables and 72 references.