Research On Dynamic Control Method Of Benzene/isopropanol/water Separation By Extractive Distillation Based On Anfis

Benzene and isopropanol are important chemical raw materials in chemical processes and often form ternary azeotropes with water that are difficult to separate by ordinary distillation.Efficient recovery of benzene and isopropanol from industrial wastewater is an important issue in achieving comprehensive resource utilization.There are a variety of separation techniques available to treat azeotropes,and extractive distillation is the most common method to achieve effective separation of benzene,isopropanol and water.Extractive distillation uses an extractant to change the relative volatility of the original compositions to achieve an effective separation of the azeotropes.Extractive distillation is an efficient,economical and environmentally friendly separation process.However,due to the presence of extractants,the difficulty of online composition detection and the strong coupling between variables,the extraction distillation process still suffers from poor extractant recovery,difficult to guarantee product purity and unstable control processes.In this paper,the design and optimisation of a dynamic control scheme based on PID control and Adaptive NeuroFuzzy Inference System（ANFIS）is carried out for the extractive distillation process of benzene/isopropanol/water,mainly including the following.By analyzing the pipeline problems existing in the three-column extractive distillation,the three-column extractive distillation process that meets the actual pipeline requirements is studied.On this basis,the sensitivity criterion method is used to determine the temperature sensitive plate of each distillation column,and a conventional temperature control scheme based on PID control and a dual temperature control scheme are proposed to ensure the stable operation of the three-column extractive distillation.In order to verify the feasibility and effectiveness of the control schemes,feed flow and composition perturbations are introduced for the two control schemes respectively,and the dynamic characteristics are evaluated by integration of the squared error.The results show that the conventional temperature control scheme has a relatively simple control structure,is less costly and can effectively handle ±10%feed disturbance.The dual temperature control scheme,based on multiple temperature controls,can precisely control the internal temperature distribution of the distillation column and solve ±20% of the feed disturbance with better robustness and stability.In order to solve the problems of extractive distillation such as the difficulty of online detection of compositions and the strong coupling between variables,this paper designs a combined ANFIS-PID control scheme based on the dual temperature control scheme using ANFIS instead of the composition controller.The scheme uses the easily measurable distillation column variables as input to ANFIS,instead of selecting the compositions as the controlled quantity,avoiding the need for real-time detection of product concentration.Compared to conventional temperature control schemes and dual temperature control schemes,the combined ANFIS-PID control scheme proposed in this paper not only deals effectively with the effects of feed disturbances,but also has the advantages of short response time and small overshoot.In view of the problem that the internal algorithm of ANFIS is easy to fall into local optimum and the control scheme of ANFIS has limitations,an initial fuzzy inference system is constructed by using the fuzzy C-means clustering method.Genetic algorithm（GA）and particle swarm optimization（PSO）are used to train the network parameters of the ANFIS model,which improves the convergence speed and prediction accuracy of the ANFIS model,and ensures the effectiveness and accuracy of the control scheme.On this basis,the influence of the number of clusters on the prediction accuracy of the ANFIS system is analyzed.Taking the three-column extractive distillation separation of benzene/isopropanol/water as an example,the improved ANFIS-PID control scheme was compared with the dual temperature control scheme and the ANFIS-PID control scheme.The results show that the proposed GA-ANFIS-PID and PSO-ANFIS-PID control schemes show good control effects.The appropriate number of clusters can significantly improve the prediction ability and control ability of ANFIS.In particular,the PSO-ANFIS-PID control scheme is superior to other control schemes in performance indicators such as settling time,overshoot,and control accuracy.The research in this paper provides a feasible dynamic control scheme for the benzene/isopropanol/water extractive distillation separation process,which provides a useful reference for its industrial application and is of some significance for the control and optimisation of the extractive distillation process.