Optimum Design Of Distillation Column With Mixed Integer Space Particle Swarm Optimization Algorithm

Distillation column is one of the most important separation equipments in chemical and petrochemical industry. In chemical enterprise, approximately 90% of the separation and purification are distillation operation, which uses 90% of the total energy consumption for separation. Thus distillation column optimal design, reducing equipment investment and energy consumption, is of great economic importance.Since the optimal design of distillation column involves integer variables such as total stage number and feed stage, and real continuous number such as reflux ratio, it is a mixed integer nonlinear programming problem. Based on the current study of MINLP model for optimal column design, this work adds the hydraulic constraint to the model, making the optimization result more practical. The advantages and disadvantages of various algorithms for the MINLP problem were studied in the thesis, and the particle swarm optimization algorithm was selected as the tool for optimal design of distillation column. To improve the efficiency of PSO algorithm for solving MINLP problem, the strategies for dealing with the integer variable and constraints were studied in detail, and a new integer handling method was proposed. A new version of PSO algorithm for solving the MINLP problem was presented, i.e. mixed integer space PSO (MIS-PSO), which integrates the proposed strategy of handling integer variable and the Deb’s way of handling the constraints.A new platform was proposed based on C# programing language. The MIS-PSO is programmed via C++ programing language, called by C# to update the particle’s position which represents the structure and operating parameters of the designed column. These parameters were sent into the Aspen software by C# to calculate the necessary data for objective function evaluation. The new framework reduces the complexity of optimal design and makes convergence easier via the computation power of Aspen. The MIS-PSO algorithm was then applied to the design of a azeotropic distillation column. The optimization design result indicates that the optimized total annual cost (TAC) of 7.08×105$/year is obtained from MIS-PSO algorithm, not only satisfying all the constraints but also being better than those results published in the literature. It is also 19.64% lower than those calculated by the Aspen Consep. Moreover, the column designed with Aspen Consep has not only an inferior TAC than that obtained from the MIS-PSO algorithm, but its superficial gas velocity is below the limit of 0.5 m/s. This means that the MIS-PSO algorithm is good at the optimal design of distillaton columns. The MIS-PSO algorithm was further applied to an azeotropic distillation process with two columns for separation of isopropanol and water. The TAC obtained from the model with the hydraulic constraint is 1.218×106$/year, which is 16.96% lower than the result published in the literature. Moreover, the published superficial gas velocity of Column C-1 is 0.851 m/s, which is more than the upper limit of 0.8m/s and does not satisfy the hydraulic constraint. Therefore, the results obtained from the MIS-PSO algorithm are again superior to those published in the literature.Above work shows that the MIS-PSO algorithm provides a new train of thought for PSO algorithm to solve the MINLP problem for it directly optimizes in the mixed integer space. The MIS-PSO algorithm can be used for the optimum design of distillation sequence and its results are better than those published in the literature before. In the future, this algorithm can be applied to optimal design of more complex distillation column to satisfy the needs of large-scale chemical process optimization.